Excellent. Well, thank you everybody for coming, thank you for those watching online. We're very excited about what we have to show you here today. Obviously, we started with our favorite slide. I'll let you speed read this for 30 seconds. Okay. The safe harbor out of the way. Let me just run you a little bit through the agenda here. Welcome to Appian Investor Day 2026. We're very excited to have you here with us in New York. You're first gonna hear from Matt, our CEO. Sanat and Jake are gonna go do a deep dive on our product, platform and strategy. We're gonna switch it up a little bit and introduce some outside voices. We're gonna have a customer panel and also a conversation with PwC.
Mark, our CRO, is going to come on stage and talk to us how he's driving an evolution in our go-to-market. I'll jump in with a little bit more insight into the finances of our business, and Matt and I will wrap it up with Q&A. We have people in the room and online who are new to Appian. Let me provide a little bit of a snapshot. We were founded in 1999, which means that we've been automating processes for over 25 years. Not too shabby.
We crossed $700 million revenue mark last year. We have 140 customers who pay us more than $1 million per year in software. You are gonna learn more about them as the day goes on. We're gonna do approximately $100 million in EBITDA this year. It's a big milestone for us. We have over 2,000 people globally who are Appianites. Okay. We're gonna throw a lot of slides at you. You're gonna have a number of people, Appianites and external, talking at you. What we're hoping that you get out of today at the end of the day is these four things. First, Appian is mission-critical. We want you to come away with deep understanding of how our customers use us, which is for complex, cross-functional, mission-critical use cases in regulated industries.
Again, you'll hear that from us, but you'll also hear that from customers themselves. Second, you've heard us say this before, we are an essential AI enabler. We are that deterministic layer that AI needs in order to make an impact at enterprise processes at scale. Third, we're continuing to drive sales efficiency, Mark in particular is gonna talk to you about this, but we're excited about the progress that we made, and we believe we can do better. Finally, we're gonna talk to you about our growth algorithm to drive profitability per share and the multiple ways that we can do that for the next several years. Okay. With that, I'll introduce Matt.
All right. Good afternoon. My name is Matt, Founder and CEO. I'm gonna tell you what Appian does. Here it is. We're a worldwide leader in complex process automation. Because of that, we're able to bring reliability to AI in enterprise situations. We've bolded a couple words there, process and AI. I'm well aware of which one of those is trending relatively. I'm gonna start with process. I wanna start with process for a couple reasons because I think it's important that you know where we're coming from, and also because I think until you understand what it is we've accomplished in process, the capabilities we've built in process, I don't think you can fully appreciate what we bring and the unique position we inhabit in the modern AI ecosystem. With your permission, I'm gonna start with the basics.
I'm gonna start with process. This is a process. It looks as a flowchart. Looks like a flowchart. Some people call it workflow. That is an apt name because it is about the way work flows through an organization. It governs the flow of work. It starts and ends in circles. It goes through the diamonds, and the work gets done in the rectangles. Process is a concept that was invented for two reasons. The first is so that workers could specialize and do a certain job really well, so you could allocate labor more effectively. The other reason it's invented, and probably more pertinent to our discussion today, is reliability. Process allows you to ensure reliability even though the inputs may be flawed. Someone might make a mistake somewhere in that process, but by the end, the answer is correct.
The way we do that is we validate, we check, we escalate, we remediate, we check with a rule. There's so many things a process does in order to be sure that a mistake gets found before it reaches the output, right? You can think of process as kind of an error remediation system, a tool that preempts mistakes and makes sure that they don't go forward. That's your first clue, by the way, what the connection is gonna be between process and AI, because process was built to give humans a reliability upgrade, and these days we have another technology that needs a reliability upgrade. Right here, process, this is like a kindergarten version. Let me show you something that looks a little bit more indicative of an actual process that we would see in real life.
Okay, we were gonna give you an animation. This is a process. It's probably got 500 and some nodes, and it does claims management. All right? These nodes are not the whole story. Some of them may be sub-processes, which means that when you get into them, they re-reveal another layer of nodes, and that new layer could be equally as big as this one. Could easily be so. All right? A process like this is not the whole story, but this still lets a lot of nodes. Here's another example. We have the flyby. Okay. This is gonna show you what a layer in a process looks like. This one is actually the center of a logistics enterprise.
This is like a ticking clock that coordinates all of the handoffs, all the integrations between a worldwide logistics network. A lot of nodes in there, yet that's not the totality. In both cases, you saw 500 and some nodes. That is not typical of an Appian process. A typical Appian process would be more like 6,000 nodes. All right, now we're going to give you the flyby. A typical Appian process would be 6, 7,000 nodes, that's the average of our average customer. If you go to our high-end customers, the 7-figure customers, you're talking an average of maybe 26,000 nodes versus the 400 or 500 that you just saw a moment ago.
Even now, I'm not showing you the full thing, but I wanted you to get a sense of scale, just how complicated these things can get. Working from the basic example all the way to this is the reality that we actually work with day-to-day. Why would somebody use an Appian process to handle a behavior in their organization? Number 1, if that behavior is complex, that's why they would use us. Secondly, and in parallel, if they needed it to be perfect. If they want absolute reliability, even on a complex thing, that's our job. In order to do that, we have a terrific data layer that informs the process and aggregates information and allows you access to it across the ecosystem and gives you read and write, and it's secure and optimized. It's amazing.
I'm not gonna talk about it very much. It'll come up later, but it exists, and that's another reason why people use Appian process. If you put together those things, the complexity, the total reliability, and the unusually strong industry-leading access to data, you get the formula for handling mission-critical systems. Well, that's it. That's what we're for. That's what we flourish at. It's in no surprise. If you know that much about us, you wouldn't be surprised to find out that 80% of our customers are in highly regulated industries. Our top sector is government. After that, it's financial services, pharmaceuticals, insurance. This is where we thrive, and it makes all the sense in the world, considering the values that we bring to bear. I'd like to walk you through three examples that show how this works. All right?
This first one is gonna be at a major financial services conglomerate. You know this company. I can't say their name. This is fraud management. We do their fraud management, which is to say we review millions of transactions, and we send high volume, very detailed reports to the Federal Government on which the reputation of this leading financial firm depends. That's our role. They used to do this through six legacy systems, a lot of personal time coordinating them all. There was major risks. There was risk of fines. There was risk of crimes happening. This was a very risky thing in the past. We came in. We're handling it. At this point, the gathering of all the necessary data is taking on the order of seconds instead of, you know, just seconds.
They've reduced by 98% the time it takes to process, and they've reduced their financial crime risk by 76%. I don't know how you measure a reduction in financial crime risk. This is their number, not ours. It's a big accomplishment. We're helping them achieve their mission-critical needs. Next example is a global pharmaceutical leader. We're doing their quality control for their medicines, which include debatably the world's leading pharmaceutical product. We are ensuring the safety of those medicines. This process used to be extremely paper and human-intensive. They had to get it right. There's no doubt about that. They used separate applications. They merged it with paper, they merged it with human time, and it was so laborious because they had to get it right.
It was so laborious that it was a common occurrence for medicines to expire while still in the warehouse. They had to get it right. You can imagine why they did not go from that system to just turning some AI agents loose and see what happens. This is the kind of thing you've got to get right. They turned to Appian. We now automate this process, and the work it takes to do a minor check has reduced by 95%, and we have reduced the time it takes to clear a batch by 65%. The accuracy is still there, and now they've got speed. This is a big accomplishment. third out of three. I'm gonna do three examples right now. This is the last one for the time being. This is a major branch of the U.S. military.
We're talking about a provisioning system that we're running, which is to say this is the way that we deliver objects, weapons, data to soldiers in the field. This is the system by which they get it. It used to be there were multiple systems. They were legacy. They were kinda disjointed. They would make errors. There would be redundancy. You'd wait a long time to get what you needed in the field, and by the time you got it, you got three of them, right? That's how the system used to work. They turned this over to us. We've now coordinated it, so it's a lot faster, not just a little. A process that used to take six to eight months is now taking under two days. It's also accurate. The data's better, the quality is better, the provisioning, the accuracy.
It is just a major transformation in the way an exceptionally important process runs for this U.S. military branch. Now, our customer base, this is our customer base. This is who we do our work for. It's not just mission-critical work. It's mission-critical entities that we're providing it to. We have 7 out of the world's top 10 pharmaceutical companies, 7 out of the world's top 10 insurers, 8 out of the world's top 10 non-Chinese banks, every one of the 15 U.S. governmental agencies, every one of the U.S.'s military agencies, the governments of 20 countries. This is our customer base. This is not trivial relationships either. These are deep, valuable, annual relationships. That's the work that we do. All right. Now, that previous slide, that's the endorsement that means the most to me. This is a nice endorsement as well.
These are the world's largest analysts saying that we're a leader in whatever they call our industry. They've got different names for it. I think Gartner calls it Business Orchestration and Automation Technology, which spells BOAT. The others have their own, their own names, but they may not agree to what they call it, but they do agree that we're good at it. We do have the unanimous agreement of the major analysts. Now, with that, I have come full circle and explained the process side of what we do. Please keep that in mind as we go forward into the AI side because it all depends upon what we've accomplished in process.
When we started, all the workers in this process were humans, which is to say that every one of those rectangles where a job gets done, the job was being done by a person. Like, 10 to 15 years ago, maybe 20 years ago, that changed, and it changed gradually over a number of years, but it changed a lot. It got to the point where digital workers were doing the jobs instead of people. You had a mix, you had a team. There were lots of different digital workers. You had Robotic Process Automation and API calls and business rules and of course, artificial intelligence, though this was before the heyday of large language models, it was still AI.
We had all these different workers doing a job in tandem, and we found out right away that it really matters what job you give to what worker. Workers are good at different things. At this point, it's almost a cliché to talk about probabilistic versus deterministic, right? I'm sure you've heard that, like, way too much. I'm gonna have to explain it briefly, apologies, because it's important to what I'm trying to say. Probabilistic is any technology where if you ask it the same question twice, you get different answers. That's totally important because it means it's not perfectly predictable, and if it's not perfectly predictable, it's not perfectly reliable. You've got a technology that though it may be powerful, like AI is very powerful, it is not perfectly reliable.
In the cases that we have been talking about, in those three case studies, in those customers that I put up on the slide, they need total reliability. Just total reliability. This is an essential distinction. It is a profound difference actually between the two, but I think it is largely understood, so I am not going to spend much more time on it. In fact, some people have gone beyond that conclusion, and they say, not only is it obvious that there is a divergence between probabilistic and deterministic, but probabilistic technology like AI will require a deterministic layer in order to work reliably. I agree with that, and I will talk more about it later, but that is where the conversation has gone.
For the time being, I'm going to focus on the different kinds of workers and the way we discovered their capabilities when we put them into a process in the early days. We found that they sort themselves into two categories. One of those, they are reliable. They do exactly what you ask them to do. However, they're not capable of reasoning their way through a hard problem. They do exactly the right thing, but they can only handle simple jobs. The other workers are much more capable. They do reason, but they are not entirely predictable, and also they're more expensive. Like AI and people can do great work, but they're expensive and a little bit unpredictable. What the world is looking for, of course, with the AI revolution is, could we just take AI and kind of, you know, shift it into the middle?
Could we make AI not just powerful but reliable? This is the $1 million question, the $1 billion, $1 trillion question. Can we make AI totally reliable? The answer is yes. Yes, we can. In fact, Appian is doing it. We're doing it all the time, and the key is the process technology that I walked you through at the beginning of this talk. Let me show you how we're doing it.
This is a close-up of 1 node in a process model. As you recall, the work gets done in the rectangles. In this case, we gave the work to AI. What that means is we expect AI to be the worker that does the work. When we put AI into a process node and delegate the work to it, here's how we do it. First of all, we'd be sure it's got a single job. That AI is specialized.
That's the thing that it does. We have a narrow group of inputs that come in, and they're sorted according to the AI's likely ability to do that job. We give the AI a very narrow range of possible actions. Not any improvisation, but a set of pre-collated, curated, audited behaviors that it's allowed to launch, which, while narrow, are still extremely powerful. We're auditing extremely closely, not just at what the AI does and whether we think it's appropriate, but at the net outcome of the entire case. If it touches AI, we look at the total outcome to be sure that AI is not correlated with inferior net case outcomes. Whatever AI does, we're reconciling it with some other entity. It could be another AI run in parallel. It could be a person. It could be a rule. Everything is reconciled. Humans are in the loop.
Other things are in the loop. We're careful all the time about what AI is doing. If we find a systematic problem, any kind of a deficiency, we're gonna change the definition of the AI, or we're gonna route work away from it to take away the work that it's not doing as well on in order that the AI can exhibit top performance on the things in which we delegate it. I mean, look at this, right? It's almost like we don't trust AI. That's a joke. We don't trust AI, but we know that if you put it in with all these restrictions, it's going to give you a great output. You need to be careful. You can't just let it loose. This is the structure that can make AI reliable.
If you've followed the literature, if you've read the conclusions of surveys, then you are aware that the industry has struggled to make value with AI. Study after study has shown that many organizations are getting not just a little, but zero value from AI. It's astounding, actually. I mean, this is the greatest technology of a generation, our economy seems to be having trouble finding value at all out of AI, particularly in high-value use cases, particularly at times that you're making strategic decisions or facing a customer or doing something where you can't afford a mistake, then it's enormously difficult to attach AI. It poses the question, I believe it's the number one question of 2026 in business anyway. That question is, how are we going to make value with AI in strategic applications?
Let me show you, right? Because I think we have the answer to that critical question. This is our AI usage, right? Over the past 9 quarters. I love this growth, and I think it tells a story. Q4 was bigger than all the quarters that came before it. Q1 is bigger than all of 2025 put together, right? This is an exceptional growth story here. As you look at it, please keep in mind who these customers are and what they're doing. Remember, this is 7 out of the 10 largest pharmaceutical companies, 7 out of the 10 largest insurers, 8 out of the 10 biggest non-Chinese banks, every branch of the government, every branch of the military. By the way, you know what they're doing because you saw the case studies.
They're doing the most mission-critical things, and they are the largest, most error-intolerant organizations. This is the cohort that is hardest to move to AI, and this is what they're doing. 40% of them are paying Appian for AI, 40% of our entire customer base, and their usage is going up exponentially. That's how we're answering the question. This is our ARR on the AI tier in our product. Again, terrific growth. I have mentioned reliability as the core reason why you would use AI in a process, and specifically in an Appian process. It is the number one reason. Before I proceed, I want to explain that there are a couple of other reasons why you would wanna use AI in an Appian process. One is our unparalleled access to data. The Data Fabric is really something special.
I'm not gonna talk about it now, but access to data, and for that matter, access to shared assets across the enterprise, Appian is really good at inventing and sharing shared assets. AI, by its very nature, must be bottom-up. You give AI a job like build me this application or do this job, it thinks bottom-up. We think enterprise top-down. There's a fundamental difference, and sometimes it's really important. Finally, you know, there's data access. There's some applications that you just need human attention on. It could be because they're gonna be reviewed by the government over the course of years to be sure they're absolutely perfect, like FedRAMP and IL6 and stuff like that. There's some code that you've just gotta get right.
You know, it's important to know, actually, and I'm not sure there's been enough talk about this publicly, that there's an entire side to the software industry, the application creation industry, where the main cost of making an application is not the cost of the lines of the code, it's the cost of the mistakes if you get it wrong. We tend to work on that side of the business. It's not about the lines, it's about the mistakes. Code could be cheap, but mistakes are expensive. That's the side of the business we work on, and it's an important side. I feel like it's been overlooked in the conversation of the last few quarters. That's important to us. When people say AI is gonna write everything, I think, well, have you ever heard of a SIFMU? Right? A systemically important financial management utility, right?
Not only can they not write their code with AI, they can't even write a spec without the permission of the government and a lengthy public review period, right? No way are they just gonna publish some AI code, right? There's a whole side of the software industry, especially the government, the regulated industries, where what you publish is of utmost importance for its reliability, and that's the true cost. Why, why do I know about SIFMUs and the lengthy regulatory process? Well, 'cause we automate, of course. All right. It's one thing to make this argument. It's one thing to say why AI belongs in a process. It's deterministic, and Appian does all these things. It's one thing to say it's another to show it.
We felt it was essential that we demonstrate our thesis by choosing a solution that had universal applicability and demonstrating it to the world in action. That's what we did. We chose this one. It's called Doc Center, and basically, it's processing incoming documents. Every organization has this problem. Everybody's got thousands of incoming documents that they've gotta read. They could be registrations or submissions or regulatory checkups or complaints or policy changes or address updates or, like, receipts or photographs of the crash to your insurer or whatever, right? Everybody's got a torrent of incoming documents. We thought this would be a great place to demonstrate how AI and process can make music together. We built this product, Doc Center, and that's basically just exactly what it does. It takes all your incoming documents, no matter what shape or format, digital, physical, anything.
It takes all those, it parses them through AI, and there's a complex way we're doing it. We've got multiple large language models and humans working as in a team. It's very kind of predictable. Then you get three things out of that. You get uploads to your databases, you kick off any response processes that you need in order to react to the incoming stimulus, and then third, you make your response. That's it. That's what DocCenter does. The statistics have been fantastic, and I've mentioned a few of them across the bottom here, but this is just hardly scraping the surface. Hundreds of organizations are using this now. Enormous enthusiasm around it, driving a boom in AI usage.
It's been a terrific thing for us, I believe most importantly, it's made our point about how AI and process together can do things that AI alone cannot, because that's the core reason that we're doing it. Take, for example, this life insurance company based in North America, which is using us for incoming document processing. They do 11 million documents per year. It used to be that they were doing manual reviews, and it was slow and error-prone, and they couldn't scale. Now they've adopted Doc Center, and they're processing 600,000 pages per month with a 75% reduction in review time and a 98% extraction accuracy. This is the magic of Doc Center.
There's so many examples, and many of them were on the stage at our show two weeks ago, and it was great to see the success stories that they proclaimed. Everybody on stage was talking about AI and about what we can do in a combination of AI and process, and it was great to see all that success. We use agents. We have a special take on how to do agents. Our agents are smarter, simpler, and safer, and that's the core of our differentiated value proposition. They're smarter because we have access to our Appian Data Fabric, and agents are as good as the data you give them. We have a terrific way of serving data to our agents. They're simple because the intuitive interface by which you create and later define an agent is incredibly straightforward.
You could do it in a couple of minutes, like 10 min. max, and it's just as easy to revise it later on if you want to. That, by the way, is with all the guardrails. That's not like fire and forget. That's with all the modifications and the safety and the tracking and everything. You could do that in 10 min. Then third, it's safer because we monitor everything. In our process environment, we know exactly what every entity is doing, and that's just great for AI. You want that total numerical record in order that you can optimize it, improve it, track it, change it, reroute it, what have you. That's the environment AI thrives best in. All right. Here is a agent example for you.
It's a North America telecom provider, and they're doing wire installation processes for large housing communities. This is an integration-heavy, complex process, and it used to be done very inefficiently. They've got Appian doing it with an agent, and they have 90% accuracy before involving a human. The agent is doing an incredible swath of the job by itself, slashing costs, slashing time. They've got such savings, and it's highly accurate. That's our agents in action. Now, this is a really important point. This is something I wish that everyone understood. In fact, if there was one slide that I wish I could just show in Times Square and get everybody to totally understand, this might be the slide.
My point here is about application development. As you'll see as I go on, it could apply to a lot of things that agents do. In this case, I've drawn a triangle. That triangle represents all of the applications that your business or any business does. It's sorted according to how much reliability you need. Some of it requires only a little bit of reliability, and some needs a lot. My scale for reliability is nines, which is the customary scale for reliability, like 99.9, 99.99, right? That's how many nines you need. Some processes don't need many nines, and some of them need actually an incredible amount.
If personal safety or financials or something is on the line, if you're deciding who gets a job or sending astronauts into space, you need a lot of nines. Vibe coding is good for some applications, but it's not good for every application, and you can't do it if you need a lot of nines. This is a really important thing to realize, and it's just as true for work that AI does as it is true for applications that AI writes. Writing an application is really like just doing the work in advance. You're making the decisions. You're writing the script that makes the decisions instead of making the decisions in real time. Basically, it's the same thing. You're entrusting the decisions, and you need reliability in some cases.
What we've got here is a situation where vibe coding and AI generally covers part of the market, but standalone, it cannot cover the other part of the market. That's where we come in. We're doing spec-driven development, which is to say that we can provide AI the reliability in writing code like we do provide it the reliability in doing work. It's not that different. In both cases, AI alone can fill the bottom of the pyramid, but it takes AI plus Appian to fill the top of the pyramid in both cases. Let me just wrap up by stating the obvious here, which is that the more reliable the job necessitates, the more reliability the job needs, the more likely it is that it is valuable.
In fact, the correlation between reliability and value is so tight that you could basically consider it the same axis, which is why I've just labeled it value. We're not talking about some esoteric corner of the business here. This is actually the most valuable part of the business where AI can't go, where statistics repeatedly show that AI has not gone, right? People are not using AI for this. They don't dare because they can't afford the mistakes. This is where we can take AI. With our technology, we can take AI to places that it cannot go alone. Let me show you more. The mainstream way of developing in Appian is now something we call Composer, which is a natural language development methodology. I could say it's like Claude Code, I'd be more precise to say it just is Claude Code.
Like, we use Claude Code. There are two differences between the way we build an application and the way you would work with Claude Code. Number 1 is that the endpoint of the process in our case is an Appian application, not a code application. That's difference number 1. There's a number of advantages. There's a number of reasons why you would prefer an Appian application. 'Cause it's got a lot of pre-built power. It's very strong. It's got data integrations with our Data Fabric. We'd go on, but there are a lot of great things about using the Appian platform. It's modern, it's updated, it works on all these devices, et cetera. Okay, that's 1 reason. 1 difference is you get an Appian application instead of a code stack.
The other thing that is different between Appian Composer and Claude Code is that before we write the application, we pause. We say, and we show it to the person. We give them a complete and detailed dashboard and say, "This is every last detail of the application that we are about to write. We're not writing it yet until you check this and you agree. Here's every role, here's every user, here's every screen, here's every data table, here's every index, here's every rule. You audit it, you look at it, you share it around. When you're ready to go, then we build it." We have this moment of collaboration, this moment of togetherness and auditing to be sure it's right. Then after we build it, you can go back to that stage anytime you want.
You can go back to that and say, "Okay, tell me the way it is right now, and I'm gonna make a few changes. I'm gonna tweak this rule. I'm gonna change that interface. I'm gonna modify this or that," or you could just make the changes and you iterate. You just cycle again and again through this exceptionally collaborative and ever-changing evolution of your application. Those are the key differences between us and Claude Code. This is an amazingly powerful capability here, and it serves three purposes. The first is if you've got a new application, the first thing you're gonna do is write a spec. That spec becomes the incredibly detailed dashboard. That dashboard becomes your application faster than ever before. Incredible time savings.
The second thing is you take a legacy app, we extract that into a spec, and then we proceed as before. We make a Appian application to replace your legacy application. I'm gonna talk more about that in a second because I think it's an amazing new possibility. Then third, you continuously improve existing applications that were already in Appian. This is marvelous. It's gonna keep our customer base up to date. I'm really excited about that functionality as well. All three of these are game changers. I'm gonna drill into particularly the legacy apps concept for just a moment because it really represents a whole new horizon for us, and it could be, an extraordinarily valuable application of AI technology. Every organization I speak to has the same problem. They have thousands of legacy applications.
They are out of date and redundant and insecure and trapping data and poorly integrated and hard to use and requiring training, and CIOs absolutely hate them, but they survive because it's expensive to replace them, and they're worried about risk, right. The old application may be bad, but at least it works, and if you change it might not work. That's it. The cost of replacement and the risk of replacement are the reasons those applications stay where they are. AI has changed two main things about this. Number one, it's made it a lot easier, a lot more cost efficient to make that change. Secondly, it's made it more urgent to make that change because of applications like Mythos that can security crack into existing applications.
It's one thing to have a modern application that's gonna get a patch in the next month to be sure that it covers whatever deficiencies it may have. These old applications, they're not getting a patch. They're flawed. They've been exploitable for decades. Like, who was it said, 70% of Fortune 500 applications are 20 years old or more. It's incredible how out of date these things are. They've been flawed for all that time, but nobody ever found how to get into them. Now, it's gonna take Mythos a couple of minutes, and they're cracked. The security premise is becoming a major problem. There is an urgency around this that there didn't used to be, and organizations have to move. When they do it, they want to re-platform.
They're going to move a code stack onto a modern platform in order that it is going to be safe and maintained and modern in the future. They want to improve the application, not just translate it, but make it better. Furthermore, they would prefer to consolidate. We can offer all three of those things, and we have. We've been in this market for a long time. We have a track record of being a leader here, though it was in the past a relatively small market. We've got a fantastic track record. We consolidated 500 applications at Hitachi down to one. We saved the Air Force $80 million in the first year by consolidating.
We've done some incredible things in legacy modernization, and now we're ready to lead in the new version and the much bigger version of the legacy modernization market. We have specific advantages here. One of them is that the Platform we port the application to is a great Platform. I'm talking about the Appian Platform. All the capabilities that come in that Platform make it a terrific landing point for migrating your legacy apps. The second is that moment, that pause. That pause where you can collaborate with Claude Code and decide exactly the nature of the application. That's the moment where you can take a legacy app and make it better. You don't have to just reproduce your legacy technology on a new Platform. You can bring it up to date, you can make it modern, and you can do so safely.
That's really an incredible gift. Then third, we're capable of consolidating applications. Like I just mentioned at Hitachi, we're 500 down to one. For all these reasons, we feel that we're a strong player in this market as it grows, and our technology is right up there with the best. Here we're talking about a Fortune 500 insurer that does end of life insurance underwriting and application intake. It used to do this with a legacy portal. It replaced that portal with a secure governed application. It saved a great deal of time by doing the new application writing with Appian AI, with Composer, and that's the reason I mentioned this use case, because we were able to provide them tremendous savings and translate an essential application to a working format, a superior working format in an efficient manner.
All right, what I've now done is complete the entire circle. We talked about process, we talked about AI. We established what Appian's edge was in process, and we understood why that gives us a unique place in the expanding AI ecosystem. AI is not standalone technology. It's probabilistic. It's not reliable enough. There's gonna have to be an AI stack. The AI stack is gonna have to include deterministic framework that makes AI reliable. We're not gonna be the only player in that stack. In fact, when I look around, I feel like every tech company on the globe has a lightweight workflow layer. Like, literally everybody seems to have it. We're at the high end. We're not the only one player at the high end either, but we've got some powerful technology. This is a meaningful, evolving market, and we're exceptionally well placed for it.
We've always claimed a big TAM. We've always been in a big market, that market gets bigger when you talk about the combination of process and AI. I like to say it doubles. Some people say it more than doubles. I think our ability to add value to a customer has definitely doubled and probably more. That's before you mention legacy modernization. I think legacy modernization is just off the charts in terms of potential. We have always fought over the very top layer in an enterprise, the last thing that they're building, like the newest thing, the latest initiative, we have clashes over the very most modern system. You talk about legacy modernization, now you're talking about every system they've ever done.
We're not fighting for a single most new application. Now the mass remediation of all of their out-of-date and probably insecure applications. This is a big prize. It is dawning right now. It is beginning right now. The thing that will set the trigger for this, that'll make a small industry into a gigantic industry is technology. It's gonna be who can deliver this safely, not who's the first person to put an AI on the start line and hit a big green go button. It's gonna be who can deliver this with true reliability, because the cost of migrating an application isn't the lines of code, it's the cost of making a mistake. We are as close to this, I think, as anybody right now, and it's a truly exciting prospect.
With that, I would like to hand the stage to Sanat to talk about our products. Sanat runs product management. Please welcome.
Good afternoon. Great to see you all, and I'm excited to talk to you about our products and the platform that we've built that sets us up for being able to deliver these mission-critical use cases for the most complex customers on the planet. But I'll start with a little bit about myself. As Matt said, I'm responsible for the products here at Appian, and my career has been in complex B2B enterprise software. Before coming to Appian, I spent many years at Oracle and then at Amazon Web Services, helping build very large businesses, working with some of the largest customers on the globe on very complex problems like supply chain systems, logistics systems, manufacturing, financial transformations, CRM transformations.
You know, when I first spoke with Matt, what really got me excited about Appian was Appian's been maniacally focused on solving these really, really hard-to-tackle, mission-critical problems for these customers, and we are really great at it. The reason why we are great at it is because we take this process perspective. Matt showed you this slide a couple times, and I'm gonna drill down into this a little bit. I'm also going to give you lots of examples of why processes and solving processes is so important. If you think about technology investments that the large companies have made, they've spent hundreds of millions dollars, sometimes billions of dollars in technology transformation programs.
If you go talk to a CEO or the chief operating officer at many of these companies or the CFO and ask them, How did you do on the ROI? Did you achieve your objective of transformation? By and large, I think the answer is gonna be, 'We got there part of the way, but we didn't really achieve our objective.' Our diagnosis is that happened because they were automating transactional silos. The real world doesn't work in silos. Real-world processes run across departments, run across systems. They don't really respect those boundaries. You must take a process perspective to say, 'Let me think about the end-to-end process. Let me design it, automate it, optimize it from that process perspective.' That's the approach we've brought to the table.
Appian, for 25 years, has been really focused on bringing that process improvement mentality to all our customer engagements. Again, as I said, we've gone after the most daunting problems these companies have had. To be able to do that, you need a platform, and that platform needs certain capabilities. I'm going to walk you through these capabilities and why they are firstly so critical, and also why they are so difficult to replicate. The first thing you need is you need the capability to design, automate, and optimize processes. When you are doing that process automation, you really need a portfolio of capabilities. It's not just one thing that can solve it. You require a portfolio. We'll talk about that.
For processes, whether it's people making decisions, whether it's systems making decisions, or now AI making decisions, we all have heard about without good data, you cannot make good decisions. That's particularly true of AI. We'll talk about the investments we have had to make in building out probably the industry's best data layer. We call it data fabric. The next thing is, once you develop these processes, what happens? They decay over time, so you need a process intelligence layer. Last but not least, we are talking about some of the most important processes these companies have. You need an industrial-grade platform, the foundation on which to deploy those applications. All of these things aren't, you know, built overnight.
They are easy to, you know, think about and ask for, but they're really, really hard to build. Let's go into each one of these. So the comprehensive automation portfolio. A given business process, and this is an example of, say, an order-to-cash process. You're gonna see so many different systems involved, in some cases 50, 60, 100 systems that are involved. Some of the systems talk to each other through APIs. Some of those systems talk to each other. You know, there is no API, so you've got to figure out a way such as robotic process automation. We have a portfolio of techniques, technologies. Think about those as digital workers that need to come together. We call this ensemble and this approach process orchestration.
What the process layer must do is it must decide how these systems talk to each other. In many cases, you have got APIs, so you've got to have robust API integration, and that has to be secure, that has to be scalable, that has to be very, very reliable. It has to be heterogeneous because, you know, these systems, legacy systems, have been built up over the years, now you need to support sometimes hundreds of different standards. That, again, is not that easy to build. It takes years and years of doing this to really get there. You've got those integrations. In other cases, you have lots of business rules. If you think about an insurance claims process, what is my policy for approving a claim or rejecting a claim? What is the threshold?
These are deterministic business rules, and you need a robust, extensible way of defining those business rules. Then you have robotic process automation for, say, mainframe systems, which don't expose their APIs, so you have to emulate human beings punching keyboards. Then you now have AI, and AI is playing an increasing role, whether it's machine learning-based automation or now generative AI automation. That is becoming an important digital worker in business process. Last but not least, people play a super important role because the types of customers we support in regulated industries, there is zero tolerance for error. People now are writing These are called human-centric workflows, where it's human in the loop, and they are supervising what's happening. The process layer is what's making all of this work flawlessly, every time, every single time.
To bring this to life, let me give you an example of a global, you know, European-based industrial conglomeration. This company manufactures very complex diagnostic technology, medical diagnostic technology. These things can cost tens of millions of dollars. The process they had before, their order management process, where these orders were coming in through email or their reps were sending those in, it took really a long time for human beings to extract all the information, look at the bill of materials in the complex orders, check those against their standard product definitions to make sure that it was something they could fulfill, that it was, you know, compliant with regulations, and then get those orders manually into systems. A lot of swivel chairing going on.
As you might imagine, when you are talking about these complex medical equipment orders, you really don't want to delay those. Not only is there a customer implication, there is also a revenue recognition implication, and this is what this customer was struggling with. Appian went in, and we put in an automated process with a variety of these digital workers to incorporate intelligent document processing to extract information from these very complex orders, a bunch of business rules, as well as then people overseeing what was happening. Finally, the orchestration of APIs so that the data entry got automated. These orders were getting created, and then the whole supply chain started working very smoothly, and they were able to achieve 95% accuracy. We call it straight-through processing.
This was only possible because we had this portfolio of tools available, you know, on our platform. Let's talk about our Appian Data Fabric. This is another capability that we have invested in for a long time, and this is a word that the industry has started using quite a bit. The reason why this is becoming so important is the processes, as we saw, don't really respect silos. They have to work across multiple systems. The effect of that is the data is siloed. You know, in this case, we talked about the order-to-cash process. You have, you know, a CRM system that could be Siebel or it could be Salesforce or it could be SAP. You have an order management system that could be a custom-built system or it could be, say, Oracle E-Business Suite or SAP.
You have a manufacturing system that most likely is homegrown. You've got financials. Your compliance system might be something else. You have these data silos, or a series of data silos, and your process really has no way of pulling all that information together for automation. The typical way the industry solves the problem is, okay, let's go build ourselves a data lake or data warehouse, and we'll spend months and months building that. The data will be together. That gets you part of the way there, right? Because now you have good reports, you have analytics, you have insights, but it still doesn't solve the processes requirement, which is, I need to complete the transaction end to end.
I need to go read from the source system, the system of record, which say for CRM, it could be Salesforce, then I need to write back to it to keep that data consistency and integrity. That is the problem that we solve. The way we solve it is instead of saying, "Give me all your data, I'm gonna go put it in a data lake or a data warehouse," which is what everybody wants to do, we took a very different approach. We say, "We are gonna create this virtual data fabric on top, and we are gonna layer on top of all these different source systems, then we are gonna do the heavy lifting of caching that data into this virtual database." We will do it seamlessly. We will do it at very high reliability and very high scale.
Now you have all this information in one place for the process to make decisions. Guess what? Now AI can use that information to make great decisions as well. Our Data Fabric is very quickly becoming this essential context layer for AI. You know, this is in a modern AI stack, the context layer is becoming an essential component, and our Data Fabric kind of sets us up to do that, and our customers are beginning to use that at scale for that purpose. As I mentioned before, this concept is becoming very popular, so all our competitors have beginning saying Data Fabric. We happen to have a true competitive edge here, hard to replicate, many patents on this technology. I would say there are three unique differentiators here. The first one is we create a semantic layer.
You may have heard a competitor talk about ontology. The semantic layer tells you what is the data, what are the entities, what are the relationships, their descriptions. For example, what is the standard definition of customer in my enterprise? It really is the information model for the enterprise. AI then uses that to navigate that data and then to be able to find the precise data and make the right decision. The next I talked about is the read/write access. What read/write access provides is instead of having to consolidate all that information, we leave the information where it is, and then we read and write at the right time from the right data source. That also is very powerful.
As you can imagine, it also gets you to the fastest way to AI value, because now you are not spending months and months trying to harmonize data into a data lake or a data warehouse. Essentially, you leave everything as is and you layer our data lake or Appian Data Fabric on top. Last but not least, security and access control is paramount. Just like you don't want any employee to go into your systems and have access to everything in your company, you don't want AI agents to have access to all the information. You want to provision just the data that they should be allowed to see. Our Appian Data Fabric has very strong data access controls.
You can do role-level access control, you know, based on roles, you can do that provisioning. Again, this is technology that works at scale with millions of records, it's proving to be invaluable in our AI journey. A way of bringing this to life is to talk about a large Japanese conglomerate. They're again, a global company. Their process problem was they had acquired a lot of companies they wanted to cross-sell and up-sell to those. The issue with that was a given account team could not get access to the right information at the right time. In that case, what they ended up doing was they layered our Data Fabric across 500 plus source systems, they were able to then create that single view of the customer.
With that, they were able to kind of accelerate their process, save a lot of manual work, and they say that their accounting, their sales teams were 50% more productive. Huge outcome. The next topic is process intelligence. You know, when you create new processes, you want to continuously monitor those, and visibility is a huge problem. With our Process HQ product, you can get visibility, real-time visibility to process performance, key performance indicators, where the bottlenecks are in processes. You know, where that process is going to benefit from automation, and then you can go precisely with the help of AI, remove those bottlenecks. What's happening now with AI agents, and Matt talked about this, all the customers we talk to are really struggling to figure out what is the value that AI is creating?
What is the return on that investment? In our case, when you deploy those agents inside a process, then you are able to see are they really adding value? Is it speeding up the process? Is it replacing costly human work? Is it replacing or getting rid of those inefficient loops in the process? Again, Process HQ is a technology that is super valuable. It was valuable before, and it's even more valuable in the age of AI and AI agents. One example that I'll quickly touch on is this Latin American financial institution. They had a process problem where their compliance systems and customer onboarding processes were very slow, and they could not figure out why. They were missing service level agreements all the time. They were using some other automation technology.
We went in, we layered in our Appian Process HQ, and we were able to quickly diagnose for them where the service level agreements were being breached. With that precise information, we were then able to say, "Here is the automation that you need to incorporate." Using our automation technologies, the portfolio approach, we were able to address those bottlenecks. They were able to save 10 full-time equivalent resources, you know, 2,000 plus days every year in that just that one process. Last, I'm going to talk about, and this is an important area, our enterprise-grade platform. The types of examples Matt talked about or I walked through, these are all customers where these processes are truly mission-critical.
The process does not work or does not work fast enough or is insecure, it's a threat to those companies' well-being, right? That's the definition of a mission-critical process. You know, it's easy to prototype with AI. It's really challenging to now deploy it in production, and here is why. For mission-critical processes to be reliable and dependable, you need firstly the scalability. In our case, our platform supports scale. So for example, a funds processing company, you know, processes their 401(k) reconciliation process. That has to happen every month in a very, very tight window of time. You can't afford the system not to be available. That scalability, our customers run billions of processes every month on Appian. It has to be incredibly secure.
I'll talk more about this, but security again, we have 30-plus compliances that are very industry-specific, including working with some of the world's most security-conscious customers, in financial services, but also in the intel community. The reliability You know, our customers require 5 nines availability, the system can only be down for minutes a year. Achieving that type of reliability, you know, is super difficult to achieve from a technology perspective. That's taken us years and years of investment to get there. That's why these customers that we talked about come to Appian. I talked a little bit about scalability, here are some numbers. Our autoscale technology is able to automatically scale 10x to 100x from the baseline.
As an example, there is a healthcare insurer who runs their Medicare enrollment process on Appian. For that, again, in a short window of time, there is a lot of seasonality, so you need to be able to support that burst workload. Appian does that. Data Fabric supports unlimited number of rows of data to be brought in and cached. As you deploy, for example, AI agents at scale, that capability becomes super important. Last but not least, sometimes these processes have tens of thousands of customers for any given app, and you need to be able to support that in a very performant manner. From a security perspective, this is another one where it requires a lot of investment, a lot of experience to really get right.
Most software companies achieve compliance in the first column, which is right at SOC 2, SOC Level 2. It's pretty much you can't do business in the B2B world without having that. There are thousands of customers who have that. As you start going up that spectrum, the field starts to be narrowing down, right? If you want to do business, for example, with the Department of Defense in the U.S., you have to conform to something called FedRAMP. There is a very specific set of controls that you got to respect and you have to prove. That's FedRAMP Moderate. Again, several 100 providers have that. FedRAMP High and then Impact Level 5, these are a very few select set of vendors who offer this.
What you get to do when you get to those higher levels is you get to connect your systems to the network of the U.S. government, of the most secure workloads that they can imagine, right? These are intelligence agencies, this is the United States military, so on and so forth. That's what we have achieved. In fact, very recently, we launched the Appian Government Cloud that is at Impact Level 5 of the Department of Defense. It's FedRAMP High. We were awarded a $500 million contract to be able to now do business with the United States Army. An example here is the U.S., a branch of the United States military. They run their supply chain system for their arms and ammunition on Appian.
In this case, as you might imagine, they had a variety of systems. In fact, they had some of the most complex and diverse technology landscape, and they wanted to consolidate that because these processes needed to work fast. This was mission-critical, right? You know, any delays were putting the mission at risk. It was putting our war fighters at risk. They were able to use the Appian Cloud to be able to deliver this capability. A big reason why they selected Appian was because of our security posture and our security capabilities. You know, hopefully, I was able to convey that these capabilities are not that easy to build and replicate, and that's why these customers work with Appian on solving the most complex problems, and then they stay with Appian for a long time.
Oh, by the way, it also sets us up really to be a fantastic foundation for AI and to provide reliable mission-critical AI. To talk about that, I'm gonna invite my colleague, Jake Rank, to the stage. Thank you.
All right. Thanks, Sanat. Again, my name is Jake Rank. I'm on Sanat's product team. I lead our portfolio for all of our AI and automation capabilities. As you can see, I've actually had a very long career at Appian, working with many of our most demanding customers in the field as part of our customer success team. I've been out there working across industries, seeing these processes in practice, using our platform with our customers on military facilities, in banks, all across different industries. As part of my role in the product department at Appian, I've also worked across multiple parts of our product, building out some of those integrations, RPA, Process HQ, now especially focusing on the AI area. I've seen many different aspects of the product, many different aspects of our customer base and the processes that they use.
We talked earlier about how the way to get value out of AI is to put it into a process, focus it on specific activities, specific tasks where it can do its job with context and governance around the most important and critical parts of your process. Of course, AI is only a part of our full automation suite. The right tool for the right job is a really important concept in Appian because you don't want to use AI where you don't have to. More risk, more delays, more costs. We have a complete suite of automation capabilities, so you can always pick the right tool for the right task. Of course, AI has extended what we can automate, right? It's bringing more value because you can do new things that used to be done by a human.
AI can take those tasks, whether that's a simple single-step task or whether it's an agentic task. I'm actually gonna walk you through three specific ways that we use AI as parts of our processes for our customers. The first one I wanna talk about is simply looking at individual applications of AI. We've been doing this type of AI work for years. We have taken different technologies, whether that's computer vision or machine learning or gen AI. The technology doesn't really matter because what we've done is we've packaged that technology into an easy-to-use capability that customers just drag into their processes. Very easy to configure. Don't have to worry about the technical details. It gets a job done. It solves that problem. It automates that task.
We've designed a full suite of individual capabilities that allow people to easily automate those parts of their processes. Our knowledge about the process leads to our knowledge about the solutions, making it really easy to use. One of our customers, a global truck manufacturer, uses Appian to automate their supply chain and production planning process. That means as they are looking at their supply chain every day, they're getting warnings about which parts might not be available or running low. All the different things about the supply chain logistics. You guys remember a few years ago with the pandemic, how much of an impact supply chain logistics can have and how much of a negative impact it can have on manufacturing's ability to deliver. This is really important to get right.
They were using a manual process, coming in in the morning every day, looking at all those warnings, trying to figure out, "What should we do about this? How should we handle this today?" They had to do it before the production teams hit the floor, ran into certain bottlenecks. Now with Appian, they've streamlined that process. They are now able to come in in the morning, and instead of slogging through a bunch of manual processes and paperwork, AI has already done the hard work. We put that into the process to automatically process those warnings, to automatically process all the paperwork, their planners come in in the morning and can review and simply review and approve everything that they need to do to get the production line running.
That has allowed them to streamline their delivery of trucks by up to a day per vehicle and saving EUR 29 million annually. Matt mentioned this. Another place that we've seen the ability to deliver real value with AI is with document processing. Every process has documents. Documents are a hard way to get value out of. There's a lot of dark data inside a document. You need to extract information. You need to classify those documents. You need to understand what you're getting, whether it's a document, an email, any unstructured text. There are so many opportunities for where you can use document processing to improve the efficiencies and the outcomes for business processes. Here I'm showing, for example, an insurance underwriting use case, but it's every process.
Every process has multiple places where documents are used, and you need to be able to tap into the value that they hold. We created the Doc Center solution to package everything that you need in order to be successful with document processing into one easy-to-use package. With Doc Center, we actually use multiple technologies, gen AI, machine learning, computer vision, layered so that you can get extremely high accuracy with a very low-cost effort, right? Some other technologies might require you to draw boxes on thousands of individual documents or to maintain those templates over time, which really leads to a very high total cost of ownership. With Appian, we're very flexible. We're very agentic, dynamic. We get you to a high accuracy fast, and we keep you there even if you bring on a new vendor.
Even if you bring in a new business partner that might have a different format for that same type of document, we can adapt to those different document formats. Of course, we make it easy to incorporate your document processing right into your process flow because it's all the same technology. It's all the same platform. Just like we can drag and drop in a specific AI capability, we can now incorporate your document extraction process into your overall business process. When you need to route to a human, maybe because it's low confidence or you can detect that there's an error, humans are always part of the process in Appian, so you're never far away from having a human take oversight on a highly critical business process. That's really important to a lot of our customers.
Along with those security compliance, everything stays within the boundary of Appian. You can build as many of these as you want and never have to worry about, "Am I going to be compliant? Am I going to be secure? Am I going to have my data going out to some third-party provider, and do I have to worry about what they're going to train their models on and maybe sell that model to my competitor?" No, you don't have to worry about that in Appian. It's all private. It's all inside the box. You also notice we got recognized by Gartner. They looked at our IDP solution, our Doc Center, and they ranked it the number one use case for automated processing.
That's a good validation from Gartner, but I actually really like the validation that we've gotten from the many customers that are using Doc Center at very high volume, as you saw in that show, though. The massive increase in AI volume is the success that we've seen with these customers deploying, in many cases, Doc Center. Here we have a U.S. mortgage company who's using Doc Center to accelerate their post-closing audit process. Every mortgage that closes has to go through an audit process that involves 23 different document types and an Excel checklist that their users have to go through and check all these different rules, make sure that everything is in order, make sure all the dotted lines are signed, et cetera.
If you guys have done mortgages, I mean, you know, it's like a huge stack of paperwork, right? Different title companies doing different things. It's not all the same formats. They're scans. They're messy. Doc Center handles that with extremely high accuracy, and that's allowed them to increase their processing by 3 x. They had a 45-day backlog when we started that project, and now it's completely eliminated. They were headed in the wrong direction with the backlog, and now we've solved that by streamlining their process. Think about the fact that post-closing audits is only one small part of the mortgage process overall. Every part of that process has documents. Every part of that process is an opportunity for them to use Appian to streamline their business processes further. We recently announced several big enhancements to Doc Center at Appian World 2 weeks ago.
One, we use AI as a second reviewer. When you extract information from a document in Appian, now we also use AI as kind of a second check, a second pair of eyes to look at what was extracted and see, is that a high-confidence extraction? Is that a potential error? If it is, you can route that to a human because, again, humans are always part of the process in Appian. If it isn't, you can have confidence in straight-through processing that document, which increases the value that you're getting out of the automation in your process. We also take that feedback from AI as well as the feedback from human reviewers, and we combine those and use AI to generate automated recommendations that improve your extraction over time. You use Doc Center, and it gets better as you use it.
Helps you stay at a high accuracy even when business conditions change. Again, even when you bring in new partners or new document formats. Let's talk about Appian's agents. Agents in Appian, of course, can think. They can look at data. They can look at the context that they're given. They can take action. They can then look at the outcomes of those actions, reason, learn from what they've done, and then iterate so they can rapidly adapt to different conditions. That's the real power of AI agents. Appian uniquely brings our foundations so that our agents can be even better by using things like Data Fabric. That's not went into details on.
Data Fabric, not only does it bring together data from across your enterprise, so you can tap into the broadest set of information and make the best decisions possible, but it also secures that data so that both humans and agents only get access to the information that they need. You don't have to worry about letting an agent loose in your ecosystem and wondering what data it's gonna process, what data it might leak. You can secure data to an agent just the way that you secure data to humans. Agents also leverage our process engine. Like, just like everything else, an agent can be placed into a process to do a specific task. Maybe it's replacing a human or augmenting a human that's doing that task. Agents can also take advantage of calling processes.
An agent might decide, at this point in this flow, I need to call a deterministic sequence of steps. Instead of letting the agent run amok, run 1,000 tokens to do that, you can simply run a process model. You have tight control. Agents can decide when they wanna be flexible and when they wanna have tight controls. Maybe there's a specific step that requires a particular regulatory approach. The agent can take advantage of a prescriptive process to make sure that happens the right way every time. Of course, that all happens within the context of the controls that Appian provides. The guardrails, the cost controls, the visibility, the reporting, and of course, the human escalations, the fact that humans can always review what an agent's doing and adapt the outcome as they see fit.
Here's an example of what an agent in Appian could look like. Let's imagine that you're getting an email, and the email it contains a dispute for a credit card charge. You can imagine there's many ways that the customer might choose to identify themselves in that email. Maybe it's by the email address, maybe they included their account number. Maybe they attached a document to that email, which is the statement. Agents can adapt to all of those scenarios by looking at what they've got, extracting the information, maybe using Doc Center to extract the information from that attachment at high accuracy, then reasoning about the data that's available. Let's look at Data Fabric. What information do I have? How does that match what I could potentially query in Data Fabric?
Need to try to identify that customer, and I can repeatedly do that until I'm confident that I know which customer this is. We can reach out and use business rules, so we can apply deterministic logic to help maybe route the flow further. We can even call other agents so that you can have a specialist agent, maybe in a fraud review, and one agent can call that agent to make sure that they're specialized in that job, and we specialize in the overall processing. You can call out to other systems to maybe do other fraud policy detections. You can call those processes, process models to take deterministic steps, maybe writing data to other systems or to make a final decision. We overall have taken this nebulous incoming email.
Through an adaptive process, our agent has turned that into a confident, recommended resolution. That's just one part of the process. That can be routed downstream, so you can actually take action on that resolution. It all pieces together into the end-to-end business process. Our agents are accurate and reliable because we bring the right tools and context to bear. Agents are deployed in a process. That means that they're targeted at a specific task. You guys probably all use some form of AI in your daily lives or in work, right? Like ChatGPT or Gemini or whatever. If you think about the context across your entire business, hundreds, thousands of people all putting information into that little chat window, and you don't have any idea what they're putting in. Bosses don't know what their employees are doing.
They don't know what they're doing with the data that comes out of that chat. That is a scary concept to a lot of people, especially in the industries that Appian works in. By putting the AI instead into a specific task, it's bounded, it's controlled, it does the task that you ask it to do, and it doesn't do all the other stuff that you don't want it to do. That's the power of AI in a process. Of course, our AI also takes advantage of our unified context layer. Everything that you build in Appian, every record in our Appian Data Fabric has metadata that describes what is that data, what's that field, how should that field be used, what are the valid values for this field.
That's the information that makes agents so reliable, so able to use the queries and the lookups in Data Fabric. Remember, Data Fabric's accessing information not just in Appian, but also across your enterprise. You're plugging all of your major enterprise systems into our Data Fabric, which means agents are able to use all of that information. We do that not just for data, but also for process and business rules and documents and integrations and other AI tools. Everything that you have in a Appian platform is described in a way that allows agents to be effective using it. It all benefits from the common security, the shared security model, the shared deployment model, the shared compliance, and the shared data privacy so that you know that the information is your data, and it's not going anywhere else.
We have a customer who's a major Australian insurance provider, and they're doing IT case management for complex financial products. They have to adapt to constantly changing business needs. They're working with a lot of complexity and high-end customers, so they have to be very responsive. Their IT case management system wasn't keeping up. They had to constantly define new workflows. Defining those new workflows meant that they had to get people together for hours at a time to decide how to process maybe even an individual request. They've now used AI agents from Appian as part of their process to take that time of planning and implementing new IT workflows from hours down to minutes. Think about the way that that makes their business now more adaptable.
They're able to serve their high-end customers in a more reactive and quick way, which is critical to the way they operate their business. One of the things that we recently announced was broad support for Model Context Protocol. You guys probably heard about MCP. Appian agents can now take advantage directly of any MCP tool that's put out by other customers, by other products in the ecosystem. Many other major products are putting out support for the Model Context Protocol. Appian agents can now directly plug into those. That means our agents have even more access to data, even more access to take action within the enterprise ecosystem. Everything that we have in Appian, all the data, all the process, all the records, all the businessIs also accessible via MCP, via the Model Context Protocol to other agents and other AI systems.
When you build a process model in Appian, that process model immediately becomes a secure deterministic tool that any agent that anyone is building can take advantage of. We are now at a fantastic tool set if you're building an agent outside of Appian, and our agents are even more powerful because we can tap into those same tools across the ecosystem. Now, we made a bunch of other improvements as well that we announced recently, including the ability to take these agents and actually embed them into an Appian UI and into an Appian form. Now you can be working on a form and have an agent assisting you, an agent that can look up information for you, that can take action for you, that can even fill out the form for you.
This is an incredible capability, and you imagine what customers are facing right now with brain drain. Institutional knowledge is walking out the door every day. You can bake that institutional knowledge into the agent in the form of documents of policies and other data that you have in data fabric. When you have a new worker coming out of training and they needed to know how to do that task, they can ask the agent, and the agent walks them through the steps. It helps them get their job done faster, but it also helps them get the job done better. Remember, we're talking about critical systems, mission-critical processes. It's really important that people be able to ramp up quickly, so there's no risk to the overall organization, and agents helping humans does that.
Agents also take advantage of the unified context layer, like I described, and because we can provide feedback, just like with Doc Center, as you use agents, you give them feedback. Developers can give them feedback. LLMs and AI can give them feedback. Even end users can give them feedback. All of that feedback is synthesized using AI and generates improvement suggestions for our agents. You don't have to be a prompt engineer. You just look at the improvements and iteratively improve your process, your agents. We actually have seen, say, an agent that starts out 70% accurate go to 95% accuracy in 30 minutes of feedback. That's a rapid increase that you don't have to know exactly what prompt to type in. You just rate it as a subject matter expert, and it gets better.
Of course, we've taken our AI guardrails and broadened them to the entire ecosystem so that now AI does the stuff that you want and none of the stuff you don't want. Okay. I told you I was gonna tell you a number of different ways that Appian allows you to use AI in the context of a business process to automate the business operations of our customers. We talked about being able to drag AI into a process, being able to do document processing with Doc Center, and being able to use Appian's AI agents. I wanna pivot to the challenges that customers see even applying those AI capabilities because a lot of customers are not AI-ready. Their systems aren't AI-ready. They're being held back, as we were talking about, with all these legacy systems, right.
90% of their budget's going to the legacy rather than to the new things that are gonna move them forward, the things that are gonna differentiate them from their competitors. There's an imperative to get out of this situation, to unlock the secret that allows them to modernize these legacy systems. It's not just mainframes. It's not just, you know, it's COBOL systems. It's things that were coded 40 years ago. It's systems that some CIO picked 30 years ago and that have just lingered and been the core of the backbone of a business process in that company, even though nobody knows how it works. That is such a major risk that we see. It's not just that these systems are old.
It's not just that they cost a lot, and they're not adaptable. They can't adopt modern technology. It's that nobody even knows how they work anymore. That is a huge risk. This is all about risk, and our approach to managing that risk is through collaboration and spec-driven development. Matt talked about it a little bit before. It's not just using AI in any form. It's using AI in a structured way, using AI first to go extract requirements from those legacy systems so you know what the old system actually did.
It's using collaboration on a plan that allows business and IT to work together to make sure that the plan that you're building for the new application isn't just replicating the old patterns but is actually optimized for what's available in the new modern system, that it's actually captured all the requirements correctly, that it's the screens that we want. We use AI to go build the application on Appian where it can execute, where it can run, right? Unlike building in Claude only, we also run the application. It's an immediate platform for you to execute the things that you build. I talked about the extraction, again, so important. You have these old systems. Nobody knows how they work. You've got to go in and find out how they work.
You've got to talk to the experts, you also have to take screenshots. You have to understand maybe diagrams from 30 years ago about how this application was originally built. You can take even those spreadsheets, right? Everybody knows we have spreadsheets that run our business, right? You can take that spreadsheet, you can upload it into Appian Composer. It'll understand how that spreadsheet is used, the data in it, the columns. It'll reverse engineer the process, then it'll present the plan. With the plan, you can sit down between the business stakeholders and the IT delivery, you can say, "Is this the right plan? Let's add some things. Let's remove some things. Let's change some things. Let's look at the screen previews that you're gonna build. Let's look at the data model.
How about the processes? Everything about what we're about to build is now on screen and is there for you to be able to collaborate. You can change things. You can remove things again. It's a very important step to know what you want to build it. That's actually true of all projects in IT, right? They always say requirements is the most important part. That's why projects fail. This design solves that problem. AI goes and builds it. It builds a fully functional, fully production-grade Appian application that has a UI, a data model, process models. It's got integrations. It's got decisions. It's got AI. It's got agents. It has everything that the platform does built in at the appropriate place with the appropriate requirements. We've used Appian Composer at a global insurance broker.
They have a contract lifecycle management process that's on legacy software. They write contracts for $200 billion of annual premiums, and nobody knows how that system works. Like, holy crap, right? Like, whatever. You know, the adage, if it ain't broke, don't fix it. You can understand why nobody wants to touch that system. The fact is, it is broke because you can't modernize it, you can't take advantage of new technologies, and you're at risk every single day. You need a confident way to modernize a high-risk, high-profile system, and that's what Appian delivers with Appian Composer. We looked at their .NET application. We looked at all the requirements. We talked with all the people.
We bring all those things into Appian, build that plan, they collaborate on it, and now they've been able to build that forward with AI to a modern Appian application. High-risk stakes, large dollar amounts, minimal information about how it works today, but still able to be successful and highly confident. Now, we enhanced a lot about the way that we use AI in our platform when we announced at Appian World a couple weeks ago. One of them is we released developer agents. Now, not only can you build end-to-end, but you can also use individual developer agents to delegate step-by-step, which gives you even more fine-grained control over the implementation of your application. It's not just for building new applications. You can go to existing applications and improve them.
You can do everyday developer tasks with the help of an assistant developer agent, which means every part of the Appian lifecycle is now accelerated, and we deliver more value for our customers. We've enhanced a lot of the AI planning capabilities with more document formats, more reasoning. Gap analysis. You can ask, "What did I not think of?" It'll perform a gap analysis against your requirements and it'll actually ask you questions that help fill in those gaps. Again, you're bounding out your plan before you move to development. We talked about our process intelligence layer. We inject automatically all of the reporting, all the telemetry that's necessary when you build an Appian application through AI. It gets all the things that are needed to report into Process HQ.
You can do process mining, you can do bottleneck detection, you can do KPI tracking according to the things that matter to your business. It's all baked in. You don't have to take extra effort as a developer to put that in. It's automatic. Now, we think that Appian Composer and developer agents are an incredible way to build in Appian. It's the future of Appian. It's how we're gonna accelerate the value for customers across the board. We also know that developers sometimes wanna use their own tools. We've taken all the greatness, all the goodness of being able to use AI within the platform and also made it available using the Model Context Protocol to tools like Claude Code, to Codex, to Pyro. Now development shops that wanna use those tools can also build and deploy and run on the Appian platform with production-grade quality.
I've walked through a number of the different things, both on the process automation side, how process and AI together are delivering more value and better outcomes for our customers, as well as showing you some of the details of how we're capturing that legacy modernization opportunity. I wanna thank you guys all for your time, and I think we're gonna move to a panel with Marc Wilson in a moment. Thank you.
Founders of Appian. These days my job is to serve as Appian's Chief Executive Ambassador, which to me is a fancy way of saying you'll find me in an airport if you're looking for me. I travel the world and get an opportunity to meet with our prospects and our customers and get to hear what their issues are, the things that they're trying to solve. The best way I would describe what we're trying to do for them is to help them achieve strategic value in a face of a world where they're largely confronted by a lot of tactical value opportunities, particularly those that are trying to, in the words of their boards, quote, "Do AI." They're looking for more. This afternoon, I have the pleasure of leading a panel with three of our customers. Guys, if you wanna come on up.
We want all of you to hear directly from them about the challenges that they faced and what they've taken on with Appian. Thank you, gentlemen. How do you feel? I'd like to start with some basic introductions. Why don't you tell us who you are, a little bit about your organization, and what your technology priorities are?
Okay. I'm Scott Morris. I'm the Chief Technology Officer of the National Association of Insurance Commissioners. Kind of a mouthful there. The NAIC. We are not a regulator, we support regulators throughout the United States. State insurance regulators. If you don't know, insurance is regulated by each state, territory, and the District of Columbia. Our organization helps them collaborate on policy. My main goal is to help them with technology, and technology that helps them lower the friction to do business with them. Helping insurance companies. We're kind of that hub in the middle. Insurance companies interact with us, and we provide data information to the insurance regulators. Our priorities for the year, strangely enough, modernization is a big key.
I felt in good company when Matt mentioned 70%, so that we definitely are seeing that, 20, 25-year-old applications that we've been working on and will continue to do so. Data and data platform's key for us, as well as improving our overall customer experience, going from a siloed experience to more of a uniform experience. The last thing I would mention is our.
We've been doing a lot of experimentation with AI. We are certainly seeing efficiencies, but the question is, are we really providing value to our members and to our customers? That's where I don't think we're seeing that yet, and that's one of our focus points.
Great. Bob?
Yeah, I'm Bob LeBaron. I'm a Senior Vice President at Neuberger Berman, and the lead technologist for our alternative technology business. We support all of our private markets business across various investment verticals and strategies.
Keith?
Keith Koharski. I lead our global development organization from a digital and technology perspective at Regeneron, a pharmaceutical company. Looking at how to bring medicines to patients quicker. We all know the amount of time, the amount of capital it takes to prove drugs are effective and safe. Really what we're looking for over the past couple of years is really how automation, how digitalization can be used across global development.
Why don't we start out by talking a little bit about some of the use cases you have for Appian? What's an example of something that you're taking on with that technology?
Perfect. I'm happy to start.
Yeah.
You go. One use case that we recently went live with and was actually one of the Appian Innovation Awards winners, a couple weeks ago, was our study code developer. When you create a protocol for a study, you work through what your patient population, what your inclusion, what your exclusion criteria will be. Looking at all historical data, looking at the markets, what countries you're going to go into, the number of sites you're going to go into. That requires a lot of different datasets and a lot of different data sources. Requires some internal data we have, as well as some third-party benchmarking data. That also works across various different groups inside Regeneron. It's not just one function that's working through that. Rather, a function might take a portion of that protocol.
They might give it to another function to fill out their portion and then have that debate back and forth. The Appian application that we went live with helps combine all those datasets together. Helps in a what we call a single pane of glass, but really a single UI, so you don't have to alt+tab between five or six different systems. You don't need to go out to a third-party data set. You make the foundation of having that scientific debate so much easier. Instead of going back and forth on what dataset you need, rather we are looking at the same set of data in a same UI. You also have that orchestration layer and that workflow component. I can do my portion. Sorry. I can do my portion, then I can give it to Scott.
He can do his portion, and in real time, we can collaborate together on that same document, being a much more efficient process.
How would you describe that process before you took on the orchestration-based approach?
It took a lot of manual effort, right? At Regeneron, kinda our really focus on AI is it helps with some of the automation. We don't feel it can take out the scientific debate, right? Wanna make sure we're using AI and use automation appropriately to help provide information, but ultimately we're making the decisions.
That's a pretty regulated process in the grand scheme of things.
Right.
Tell us a little bit about your regulators and how they look at things like this?
Yeah. I think from a, you know, pharma perspective, about a third to half of what I do on my daily basis is GxP, GCP, GLP validated. Different regulatory bodies will give specific guidance of how something needs to be followed. When we go through, when we build applications like Appian, you need the basics and the essentials. You need audit trails. You need role-based access controls. You need to be able to trace down, you know, how a decision was made across what dataset, across what program to have repeatability if you need to make that decision again, if you need to understand what went into that decision.
These are highly regulated, highly well-documented systems, and that's where you need to have an application, a platform as robust as something like Appian, so you have those constraints well thought out.
Bob?
Yeah. We use Appian for a couple of key areas at Neuberger. We use it for our deal closing workflow and also for our fund and investor onboarding platform. We have two main applications, but they're really more a platform of micro applications, I guess you would say. For our deal closing, it handles a lot of the pre-trade compliance checks that we have to do. It also handles a lot of operational setup for our deals. It also handles other checks that we have, such as ESG, SFTR. We have a lot of different checks that we have to do to make sure that we're in compliance.
For some of our investment strategies, we also have built our, what used to be an Excel-based allocation file for allocating the given deal to all of the different Neuberger funds that would like to participate in the deal. That's all now done in Appian. That's been really nice to be able to see real time the allocation status rather than having, you know, a bunch of different versions of the same spreadsheet floating around and having to deal with the version control and see all of the rules that we have in terms of, you know, who can participate, if it's a good fit for them. Does the client, you know, do the LPs or the fund of one, do they have veto rights? Things like that.
It handles a lot of those complexities. On the fund and investor onboarding side, we handle all of our investor checks to make sure that it's kind of a combination of managing the subscription docs, also the initial onboarding steps of are these docs in good order? Have the AML checks been performed, and so forth. It also includes all of our fund onboarding as well as all the entity onboarding.
You know, in private equity, you know, this group may or may not be familiar, but when you are setting up a fund or a product, you're gonna go to market, and you're gonna say, "I'm going to create this fund and gather investors." Your investor base is going to be comprised of, could be U.S. investors, it could be non-U.S. investors. Based on the jurisdiction that they have, you're trying to minimize the tax liability so that people aren't double taxed, triple taxed, and so forth, so that people are paying the taxes that they should owe. As a result of that, you have a whole bunch of different entities that you are setting up.
A single fund structure could have anywhere from one to, you know, dozens of entities that are set up. It's a pretty complex thing that we're working with. As a result of that complexity, both on the, you know, fund and entity side and as well as the deal side, when we were first evaluating our solutions, we were looking for, okay, do we have any internal solutions that we can leverage? You know, we utilize, you know, ServiceNow for a lot of our IT processing and SDLC, but didn't feel like that was gonna be a good fit for what we needed to do. There's nothing off the shelf that you can just go buy because it's such a bespoke process, and we have so many different business lines as well.
Some of those business lines have developed internally, and they've kind of evolved organically as our business has changed. Some of them are like standalone businesses that we said, "That's a good business model. Let's buy that, and that will become one of our investment verticals." For a lot of what we had in the past, the tech stack was Excel and binders. You know, it's been much improved as we've started gathering all of the data and systematically moving it along the chain as we execute our deals.
Scott?
We have a 20, 25-year-old platform that we replaced with Appian. When you're an insurance company, about 4,000 insurance companies in the U.S. that are regulated, you have to go through your regulator or regulators to get a product approved or a rate increase approved. The state rules are different by each jurisdiction. Largely what we had in place originally was about 20, 25 years ago, we automated scanning of documents and then, you know, work flowing those documents. That was all fine and good. Obviously, we wanted a more automated process. That's when we took on to do this with Appian. The two key things we're really trying to achieve here is regulatory consistency.
Somebody mentioned earlier that there's folks that are moving out of the workforce that's in a state Department of Insurance. That's certainly the case. The regulatory consistency was achieved through seniority and longevity of the staff there. We wanna make sure that folks are making good decisions, making similar decisions on these product filings. The second piece is we wanna speed this process up, and we're just beginning to see some benefit from this. It takes about 40 days on average in the U.S. to get an insurance product filing approved. That varies greatly by state, but we wanna automate the process so that becomes faster. Those were our two key goals. You know, we used Appian to build out those workflows.
A lot of it is consistent through, you know, all the jurisdictions, each of those have kind of their own unique needs. That's one of the platforms that we've built out. You know, these are product filings. There's 4,000 insurance companies. They make about 600,000 filings a year. A filing can have, you know, up to 100, 150 files. There's a lot of information flowing. We're just trying to contain that and automate that with Appian.
On that topic, Scott, you know, what have you seen of Appian's AI features that can help with those processes? What are you doing today to look at getting that to go faster with AI?
Yeah, that's a great question. The first thing we've done is automate the intake process. We had, I mean, actual people looking at, okay, they uploaded all these documents, and then they filled out this form with all this metadata. It used to be a regulator. That was their full-time job, was to make sure that metadata matched what was in those documents. Now we're using Appian's AI capabilities, intelligent document processing to pull that information out and automate that process. The other piece that we've seen, using just recently moving to DocCenter, is classification. I mentioned there's around 150 documents in these filings. Turns out humans aren't really good at classifying these documents.
With DocCenter, we're seeing about 98% effective rate there, and that's sped up the process. That's what we're doing today. We are experimenting with more of a capability to review those filings. Every state has what they call a checklist, and that checklist is just your natural language that says, "This insurance contract must have this particular exclusion," or something like that. What we're doing is using AI in a pilot mode right now to go through and parse that out and then show them where it meets their rules or where it doesn't meet their rules. Once again, it's not just 1 set of rules, it's 56 set of rules. That's, that's how we're using AI.
That's great. Bob, I know Neuberger has been leading the charge on this for a while. I remember the presentation at Appian World last year. Could you tell us a little bit about what you're doing with Appian AI?
Yeah. We're, you know, kinda similar to what you mentioned. We use the intelligent document processing in the Doc Center, specifically with our subscription documents. Those documents are crazy long. They're, you know, could be 200 pages, and they're bespoke. The documents could vary based on the jurisdiction if it's Cayman, Luxembourg, U.S., Japan. Those could all vary in terms of their complexity and length, and sometimes even, you know, depending on a fund could have like a very specific page that's added into there. We're pulling in and extracting all of the data from our subdocs. What that's unlocking for us too is the ability to be able to extract the data more accurately so that we essentially have no, you know, 0 human data entry from there.
Also the ability to be able to do checks that were perhaps, you know, not done consistently or could be done on a haphazard basis. There's sometimes questions too that have not been worth extracting certain data points because we might only get a question one a year. Now it's so much easier for us to just be able to add a one data point and extract all that data up front so that we can solve all of those and answer all of those ad hoc questions as we're getting them down the road. We've put in extraction models. Investors sometimes will upload documents where they're supposed to.
Sometimes they'll say, "Here's my 1 consolidated PDF," and be like, "Go and find it." It becomes a challenge for our team to be able to scale. You know, we've always had a very linear headcount between our AUM and our employees that we have to hire. We're hoping to break that correlation so that we can now start to have a higher carrying capacity for each employee as we utilize AI. You know, we're using that really heavily there, so it's been able to extract the data from non-standard docs. It's also been able to apply and do the first wave of reasonableness checks and as we're extracting all of that data.
One of the other things that we really like about it is as you're extracting that data, when you're in the Doc Center, there's two things that we really like about it. First of all, it's just adding a new data extraction model is basically just a configuration within the tool. Now I don't need to have my developers playing like middleman with the business analyst who's actually the most familiar person with the document.
Now a business analyst can go in, they can actually configure the extraction model, and they can test it themselves and make sure that they're getting the results that they want before they hand it off to the development team, so that the development team can take something that's essentially ready for production and they can then just, you know, drop it into the workflow. One of the other things that's really nice about it is the ability to be able to geotag the fields.
This was a really big selling point for us because it's one thing to be able to extract the data, but as we all know, AI can hallucinate and, you know, if you're just getting a blob of the text, if the country says that they're from Germany, how do you know that they're actually from Germany without actually going into the document and seeing it? We have the ability to be able to click. It looks like a little map icon, and it's essentially geotagged within the document. You can click on the tag, and it pulls it right up next to you, so you can see from an accuracy standpoint if it extracted appropriately.
Those are some real key selling features for us as we use it.
Wonderful. Let me ask one last question before we close up here for Keith. You know, it's the opposite side of what we've talked about here. There's been a lot of discussions about how AI is just gonna solve every problem in the world. It's gonna write all of our software. It's gonna make software companies go away, for example. What's your reaction to that?
How does Regeneron think about that on a daily basis?
Yeah. We're really thinking as a hybrid approach. We don't feel, or in my opinion, don't feel, you know, software companies are gonna go out of business tomorrow. I'm sure all you know about the populist approach on the front page every day. I think a lot of the companies that existed, two years ago have a different model today. A lot of the new companies today may not be around in two years. Kind of taking a hybrid approach and understanding, kind of what core software do we need? How are those core software gonna interact with AI, whether it's part of that core software package or whether it's taking that data and embedding it with other data inside of others. For us, the key is scalability.
We don't necessarily wanna make, you know, a lot of different bets, and attaching ourselves to a company and attaching ourselves to let's say, you know, a model where that LLM could be proven, you know, one, two years down the road that there's another LLM coming tomorrow. That's where companies like Appian, where you can use it as a single pane of glass and reach out to model A, model B, model C to bring those all together when you need it. If you do need to make a change, you could easily plug and play a different component, a different model, in easier.
We are trying to take a hybrid approach, understanding what's coming, but also making sure that we are, if there are changes in the ecosystem, we're best prepared, to be able to make any changes we need to.
Very good. Well, gentlemen, thank you so much for your time.
Thank you.
Thank you.
All right. Thanks, everybody. I am Scott Van Valkenburgh. I lead global alliances and channels for Appian. Welcome on stage one of our top alliances, Dan Scott, who's a principal. We'll do some quick introductions. We're gonna talk a little bit about more how does our partnership and driving growth in the market is gonna impact over the next couple of years with AI and the types of things great firms like PwC are doing in the market. Dan, real quick, as we sit down, why don't we do an intro? We've had an alliance for 8+ years.
It's been a while.
It's been a while. We had some amazing announcements at Appian World doubling down on our alliance and the efforts that we're going there. Just a little bit about you, your background, et cetera.
Sure. Dan Scott, I'm a principal in our cloud and engineering practice. Let's see. I think I'm required to say that the opinions that I express here are my own.
Highly regulated firm. All right. With that being said, let's dive in. One of the things I was really excited about was our announcement at Appian World on legacy modernization, several things Matt talked about. You know, the opportunity to really transform the clients you serve, the clients we serve. Share with us some of the use cases and thoughts and how the firm and you and the practice are viewing Appian.
Sure. Let me start with us. We're in the process of transforming the way we do business. You know, we're trying to use AI in everything we do to bring down the cost of delivery to a client. That has brought up the number of opportunities for us pretty dramatically. That's been really exciting. We think Appian is an important part of that, and that's part of the announcements that we made at Appian World around how do we get what we call end user computing, which is Access, Excel applications. They seem boring. They're what enterprises run on. How do we get them off the desk and actually into something that we can then agentify and make more useful?
When you think about that across these use cases, the firm's put a ton of investment, people, resources in building the Appian practice. Are there some examples you're seeing in certain industries where this is driving more or less?
I would love to tell you that it's 1 industry, but it's right now across the industry. We are hiring more people all the time because this is a hotspot and it's AI adjacent. I think there's a lot of folks who are looking at this and saying they have a workflow tool. In a lot of cases, that's Appian for the customers that I work with. They're like, "Okay, I didn't like two things about my workflow tool. No offense. 1, it took me a while to actually configure that process. 2, when I actually configured that process, occasionally I had to have one of my employees actually jump in and do something 'cause the tool wasn't able to do it.
I think at least our clients are saying, "Hey, this is a great mix of deterministic that I can test, that I've been running in production for years, and non-deterministic and actually getting that faster." That's a little bit of a challenge and putting more miles on our team, because we are doing more engagements for slightly less money, although I shouldn't share that with some of my clients. That has created a lot of new opportunities for us in the market.
You're starting to see the shifts. Every day there's a new approach on AI, how are people thinking of adopting it, et cetera.
There's been a lot of the experimentation mode.
Yep.
In this bridge, in this gap from production mode. Your views of what we've been doing together in the past and some of these new announcements and changing that, what's getting that excitement in the firm?
We have a lot of customers who are deploying either code tools, and I love to tell you they're using it a lot better than just code completion on the UI. They're not always doing that there. They've launched something with an end user, and there's a big gap in between that. That has been a lot more challenging for companies to actually build and deploy at scale. There's a lot of discussions that folks have. Should I, should I go with a general agent? I can't name names on general agents, but I installed and have one of them at home. I do not give it any of my credit cards and/or passwords. That is a really bad idea.
While the general agent concept is wonderful, I just put this agent in, the agent replaces an employee. If I do the math in a company really quick, I can count up my savings really quick, and this is gonna be great. Okay, except there's some problems with that. Your controls in a highly regulated industry were not really designed for an agent. Have you fired an agent?
Yes.
I haven't. You know, what do we call agent collaboration? Is that collusion?
Do I have a 4-eyes control able to work? There's a lot of controls that are in these industries today that you have to rethink when you go to AI. A lot of times people are like, "Hey, I'm gonna get myself a subscription to a coding tool." I have many such subscriptions. I will tell you that is not the panacea that it sounds like it is. We have an entire generated code practice that does a very good business helping companies with generated code, but it costs a lot more than I think people understand.
Yeah, and the subsidization of token pricing as well. As you think about the announcement we made together at Appian World and t he investment behind the firm, the concept of the vibe coding that you mentioned, why did the firm put its weight behind with Composer and legacy modernization with Appian?
This is a growing level of interest for our customers. We've been doing over the last two to three years a lot more legacy modernization. I would say AI has put gas on that particular fire. It was a real no-brainer. In a lot of cases, we're talking about moving from legacy code to new code. New code has a lot of the problems that old code has. When I say that, it rots over time. God only knows what was in it. If you're moving to a fully generated code model, again, that requires some structure that not every one of our clients is willing to sign up for.
The idea of taking the business requirements out of that, modernizing that a little bit, through the, 'cause I'm assuming you do not want the same green screens.
No.
Done in Appian. There were some announcements about UI, but I'm pretty sure there's not a green screen mode.
There's no green screen.
Doing a little bit of modernization in that process, and then getting that application in a modern environment that stays modern, that you can buy a subscription for, that's pretty attractive.
The interesting thing as we've collaborated, it's not industry bound.
Nope.
I mean, this topic is touching every industry the firm is serving.
Yeah. It's across industries.
Yeah. We're super excited. When you take that in the discussions you have with clients today, I think Mark had mentioned, you know, the SaaS death. What does this mean? What are you hearing from clients in a broad sense? How does that relate in the way that you've personally seen our collaboration together, et cetera?
Again, I hope not to disappoint you and/or insult you on this stage, we view Appian as an app platform. That's what we use it for. Whether we're using it with clients, whether we're building products to sell to clients, whether using it for ourselves, that's kinda what we look at Appian as. I'll leave it to others to figure out what that means in terms of the SaaS apocalypse.
We don't view you in the same way that we view other SaaS products.
Yeah, I definitely think there's a different view of the platform as a service.
That's right.
-piece. Do you think clients are thinking this way, or is it just everything's so throw AI at it? I know we've had a lot of discussions on ROI.
Well, clients is a very broad term, and we've got clients that are at every part of this equation. We have some clients who were with us with some of our AI partners when they were doing initial announcements. We have other clients who are not quite as large there. We try and meet the clients where they are. We're in the process of going through this same thing ourselves. We try and bring that humility, and we also try and bring the learnings of how do we actually get the team up to speed on this so that you can start to do it. I think I heard it mentioned earlier, we actually called it strangely enough the same thing. You need an AI stack.
If you're thinking that you're just talking about a coding tool, there are a lot of tools you're gonna want around that. You might want a wiki for knowledge management. The internet is a cesspool. You might want a ticketing system. You might want a workflow system, just saying. There are a lot of tools that we use around AI, and a lot of companies have not yet sort of figured out what is their AI stack gonna look like. You know, we have just touched the surface of that. We look at it all the way from how do I wrap that API that I get, you know, all the way to how do I make sure I don't have injection?
How do I make sure that the right data is going to the right folks? You can either bring us in for a lot of those things or in a lot of cases, Appian brings that out of the box to the platform. For a lot of clients, that's really exciting.
Awesome. There's been a lot of discussion about our firms and their business models changing, et cetera. How do you see this in the demand that the things that we do together? What's the general and in the market?
Yeah. I would say there are sort of two buckets of things that we're seeing. We're seeing the same things that we used to do, and we're doing more of those 'cause we can actually bring the cost down. The ROI got better in that. There are actually some exciting new things that either weren't feasible from a cost perspective, weren't feasible from a just not able to do that perspective. Right now I would say we're very focused on bucket number 1, but we're starting to see people dream into bucket number 2. That for me, is a lot more exciting because that means revenue instead of just cost takeout. As much as I love cost takeout, I'm willing to help clients all day long with cost takeout. Revenue is a lot more fun.
Great. A couple last questions to wrap up.
Sure.
One is the firm and Appian have leaned from the beginning of this year incredibly heavily into our collaboration.
our alliance in ways I don't think either side had seen from that. When you think about this motion and the privatization of Appian within the firm, what are the growth ideas and areas that you're leaning more into? I know we talked about legacy modernization and others.
Yeah.
to help support clients.
Look, we have some of our technical practices like legacy mod, which is important to hear, but I think most people know us for our business knowledge. We have a lot of folks who have some great business ideas, either for a service that they're gonna take their advice business and turn it into a service or, that they're gonna work with clients to make their services better. We wanna be able to support them as an Appian practice to get their ideas into a production level application as quickly as possible. While some of this AI means that there are fewer people in our Appian practice per project, I think overall our program is still growing. Even if that is the case, then we will have enabled our ginormous business consulting business, which it's a win.
I'm a partner in the firm, not just in the seat of practice.
I think it's great. If you aren't aware, PwC has developed some amazing solutions on Appian. They're taking it to market, especially in the pharma life sciences space, including Interactions Hub. We just made the announcement on pharma core labeling these complex processes similar to what was spoken, financial services, different geos, regulatory environments, et cetera, and ways to help clients. I guess last to close out, what are you most excited about our collaboration? Are there any areas that you're particularly excited about?
Just one spot. I'm actually excited about the opportunity in this space. I think if there's one thing that we have seen over the last couple months that makes me the most excited is as the BPM leader, we are very interested now that we have made business process management easy with the ability to use AI to generate a process reliably, repeatedly, and then the existing capability that exists within Appian to use agentic. There's a lot of customers that need that, and we're excited to be here to help them.
Thanks, Dan. More to come with our relationship. I appreciate you spending the time with us. Next we'll introduce Mark Dorsey, our Chief Revenue Officer.
Thank you. Thanks, Scott. For a while, I need a clicker. Scott, you got the clicker? Who's got the clicker?
No clicker.
Okay. I can deal without a clicker. First of all, I want to kind of start off by saying, thank you to the customers who spoke today. Thank you for the partners you heard from. Thank you all for taking the time to be here. You all have lots of things you can do with your day. I appreciate you spending it with us. I want to tell you a little bit about me, right? My background. I spent 15 years at IBM. It was a great experience. Early in my career, I was the software sales representative of the year, the top rep out of 7,000. This is early in my career. From there, catapulted my career within IBM, where I actually went and I had many, many, many jobs.
I was asked to be part of the senior executive training program. To give you an example, when I left IBM, I was what's called a Vice President, and it's a Band C executive. To tell you that, what I mean by that was when we acquired Sterling Commerce, and I was part of that team that did that, the CEO came up with the same level I was at. Just to give you an idea of all the experiences. Not being braggadocious, but just to let you know what context, 'cause titles mean different things to different companies. IBM was a great training ground for me. I really learned how to run a business, and I learned how to sell with value, right?
Value is something you're gonna hear from me a few times today because Appian's technology, our platform, develops tremendous value. From there, I got a short stint at Bank of America Merchant Services, Executive Vice President. I went into Oracle. Oracle was a great experience for me. I spent a lot of time at Oracle in a few different roles. I was recruited there by Rich Sciarra, who works for Mark Hurd, and I was fortunate enough to be mentored by Mark Hurd for a number of years before he passed. It was a great experience. I learned an incredible amount from Mark. I get excited also talking about that. One of the key accomplishments at Oracle for me was I was asked to run their cloud business in the beginning.
I took that cloud business from $10 million to, sorry, just shy of $1 billion in a very short period of time, competing against AWS, Microsoft, Google, and other hyperscalers. What a great opportunity that was to really compete, and we had to compete with a, with a great technology and to sell value, kind of what we're doing here today. I small stint at Alteryx, we got acquired, and then unfortunately, I competed. I had two offers at the very end to go to run to be Chief Revenue Officer, but just shy of a $2 billion business, which was more of a run and maintain, or Matt gave me the opportunity to come here.
I looked at that opportunity, and I just kinda said to myself, "Why Appian?" Well, what came out at me, first of all, the product is incredibly valuable. It's easy to use, and we empower business users to solve complex problems. You don't need to be an Oracle DBA to use it. Anybody in this room could use our tool, right? You can solve really difficult problems and run the orchestration. I'm gonna do my best to speak a little bit louder than the sirens out here to kinda help you guys out. We solve really complex problems. You can see from what you've heard today from my colleagues and some of my customers and partners, we are integrated into the crux of the mission-critical problems, highly regulated industries where governance and performance has to happen.
The product is incredibly valuable, and it made me look at this and say, "Wait a second. Can we sell value here? Do we have a great product?" Everything I looked at the product, I actually talked to customers, I talked to some employees here, I went to my network, and the product works. It's kind of like this unfound gem. I looked at the next thing, and I talked about how sticky it is. I started digging into the financials, and I hope you guys do this as well. Our customer retention rate is through the roof. I'm like, "Okay, something's got to be good happening here." When you talk about value, I focus on selling outcomes. Nobody buys software because they feel like it.
They're buying it to solve a problem, to provide a strong outcome based upon a business case. As you kinda heard it today, it really kinda spans, right? Anywhere from DocuSign and adjusting documents to Agent Studio, which you can put in a process, anywhere within a process. That's kind of the secret sauce, is you can take our Agent Studio and put it anywhere in there. Then modernizing applications has been around for a long time. One of the things I really wish some of you could actually see is a demo of our Composer kinda coupled with either like Anthropic or just a demo in front of it. It's pretty amazing what we can kinda do in a very short period of time. When we show it to customers, it really dazzles them.
I had the opportunity to really transform this organization into selling more. Transforming the organization based upon value was really important because, you know, customers don't wanna buy just a new technical enhancement. What are they looking for? I mean, it's important, it's nice. They're looking for outcomes, right? That's what we focused on today. Let me talk to you about some of the things that were happening. It was more of a technical sale. Spent a lot of time kinda talking internally, focused externally, really focused on small deals. I looked at this and said, "What are we doing this for? This is incredibly valuable technology." What do we do now? We focus on value, ROI, business cases, and outcomes.
I, you know, some of the things we're gonna talk about, you know, today, you heard a little bit of it. We kept some of the numbers off of these slides, but some of these use cases are providing tens to hundreds of millions of dollars in value to organizations. That's it. We have a tool that we use. Continue to be customer obsessed. We do the right thing for our customers. We create the value somehow. How do we create that? It's with our services team, it's with our partner ecosystem, or the customers do it themselves. If you were at our Appian World, you would have heard on stage one of our customers talk about outcomes.
What they do is they take the technology people and the business people, they put in a room, and boom, they have a process, and they create incredible efficiencies really, really fast. Executive relationships. I wanna be talking to the people there and the executives and understand what are their problems they're trying to solve. If we can help them, we'll tell them. If we can't, we'll say, "Hey, we can't." We'll point them in the right direction. I don't wanna waste their time or our time. It creates a lot of crystal go big. I gotta say this, last year, we sold more 7-figure deals than we've ever done in the history of our company, and I can tell you that trajectory is not stopping. As you can see, we focus on that.
Let's talk about really what happened. First I had to do is change the team. I'll tell you, last year, during a year of great results, 34 leaders across my organization were added to the team. We improved the team dramatically by doing that. You can see my direct reports, you know, pretty much all changed, right? The next level down, a lot of transformation. I can tell you this happened throughout the organization at all levels. We're bringing in highly skilled sales executives from Google, Adobe, IBM, Salesforce, Microsoft, ServiceNow. These are some of my directs, but throughout the organization, people who can sell with value, focused on outcomes, large strategic deals, aligning with the companies at the senior level. When you do that, your product's incredibly sticky. Really what else do I need to focus on? Pipeline.
We put a focus on pipeline. Why? Because pipeline is the lifeblood of sales. You go into the numbers, you look at the yield on your pipeline, yield at different stages. These are indicators of where we're going. I can tell you this much, I'm not gonna share any numbers, but our pipeline is up dramatically. Enablement. What are we gonna do? We gotta create the strategy, point and aim our teams in the right direction, enable them how to go do that. We hired a world-class executive to run the enablement, and we're having some great results in that. Forecasting. I don't think there was enough operational discipline on how we inspect the deals, how we qualify the deals with the economic buyers to make sure that we're not wasting their time or our time. Pricing.
Last year, I created a, with the work of the team, extended team, what was called an Enterprise Growth Plan. What does that mean? That really unlocked the ability throughout the organization for people to use as much as they want, where in the organization the value's gonna come from. It's interesting. There's a large financial services customer I sat with early in my time here at Appian. We were just training them, they came back, we had a competition to show their Chief Operating Officers. I can tell you right now there's 4 others have been funding since. They found millions of dollars in value at low levels of the organization they didn't even know that was out there. This is a concept, put it in their hand to go.
This has been kind of a fun journey. I really wanna talk to you about the operating process and rigor we bring in. The first thing is we need to go spend time with our customers. I tell the teams, you know, we had a return to office policy. I have a return to customer policy. I want my teams going and spending time with their customers. That's what I want. Expected. People buy from people they like and trust, and we gotta get out there and build that. How do we do that? We focus on building pipeline. We listen to them. We actually do a lot of discovery work. We find out what their difficult and complex problems are. To do that, we have to spend a lot of time listening and learning and see where we go.
I spend time making sure we're having selling at the executive level. You can see that from some of the a few of the executives that were here today. They're getting so much value, they're coming here and speaking on our behalf. Interesting. Large strategic deals, you'll hear that. They can come in many forms. Why are we focusing on that? We're quantifying the value, and the customers understand the value, and they're willing to sign up for this. You'll see that continuing to grow. Standard, advanced, and premium tiers. That's in the advanced tiers we get our AI, our AI capabilities. Right now there's a tremendous amount of inflow from customers understanding our AI. Really, as you probably understood, I suspect many of you read the MIT study that talks about the values given within the process.
It was a softball for us, right? That's exactly where the value comes in the process because the agentic AI here contained, as you heard from Matt and others, within the process. It's not going out there willy-nilly. It's with a lot of governance, it's with a lot of regulations, and it's clear. We focus on that. It's very important to me to make sure that all our deals we qualify to deal with economic buyers because that actually yields up our forecasting, right? There's a lot of, you know, reps sometimes in organizations that they think it's gonna happen, but I wanna go ask them, "Hey, if we can kinda, you know, achieve this for you, can you do this?" We actually qualify the deals. We bring the operational rigor that I was driving at Oracle here.
By why? Because it works. Customers love it. They wanna understand too. They wanna see where things are at, but they wanna understand what's in it for them, and we clearly show them that with our business cases and our ROI. Now, there's a big focus on winning new logos. I can tell you right now, I can't get into the numbers, but that's a 7-figure new logos in Q1. I can't talk to you about what's gonna happen in Q2, what's happening, but I can tell you that in Q1, we won a lot of customers large. Couple case studies, right? This is a large insurance with what's called their star rating. Star rating, you have to have a certain level of rating or you can't be involved in the Medicaid, you can't be involved with Medicaid.
They'll just downgrade it and go to a competitor. We kinda came in, kinda helped them. We made sure that we worked with them. Spent a lot of time with their current processes and actually completely turned around and have increased their star rating at this point. They've decided right now because of our incredible technology and how we're helping them, they right now are currently in the process of moving 100 applications to us. What the slide shows you in the revenue is just the basic growth. When we first started off, I can tell you the number was not that large. They have committed to a multimillion-dollar deal that is ramped like this. Why? They're getting tremendous value. The second thing they're actually doing, 'cause they signed up for an Enterprise Growth Plan.
They're actually now looking at their competitors in our space, getting off of their competitors and going to us. We and our partners are both doing it. It's not just all Appian doing the services, our partners are doing the services as well because it's a lot of work to get off of these, and they're using some of our other tools like Composer to kinda help with this. It's a, it's an amazing flow, but you can imagine the size of this organization. This is not a small organization, and they're getting lots and lots of value. The next one here is talking about a case that happened in the branch of the military. What I love about this deal is the quick sales cycle to the bigger deal.
We spent a good amount of time closing a transaction with them. We closed in Q3 of last year. Of the quick impact that our team provided them, they then 1 quarter later signed up for a deal that was between 10% and 15% larger than what they did in Q3. A multimillion-dollar commitment here. Why did they do that? Again, they did it because of the value we're providing them. You know, the same organization was in Appian last week, sitting, written down, was strategizing on what's next. What happens is we can land and expand in these accounts because we show them the value, we show them the technology. Like in both of these examples, we go in, we develop a proof of concept with our team, and we show it to them, and they love it, right?
We make difficult problems go away with the value and the technology. Can I go to the next example here? This is an airline manufacturer. They have a massive backlog of airline engines. They were using an old, antiquated homegrown supply chain system. Think of all the parts that has to go into building an airline engine. Think of the logistics of getting these parts. Think if a part comes in and it's broken. Think of the complexity of this and the timing around this. They told me every day we speed up their production of their airline line, it saves them I'll just put this way, it's a staggering amount of millions of dollars. I'm gonna refrain from the number. What is this?
The value is incredible we save them. We're now in one airline, one of their lines, and we're gonna go live full production there. We got five more to go. These numbers here, the number of this thing here could be 5x what it is now. That's the same with the previous two examples. These were just three that I picked, to pick to show you today. We'll be 100%, I'm already talking negotiation with this organization to do a much larger transaction. If you kinda wanna get into this right now, what's important to me is rep productivity. When I talk to my sales managers, I talk to them about your job as a sales manager is to get everybody in your team successful.
I'll share something with you. Inside the organization, I think that the managers. I'm trying to get the managers to understand your job is to get everybody to be successful. What I did is I. There's one manager organization, his entire team got to it, so I sent them all to Appian World. They were blown away by that because, yeah. I want my leadership team know it's your responsibility to keep making sure you get everybody in your team to your plan. It doesn't matter if one rep just goes and crushes on the team, they make the number. I want them all contributing. I wanna make sure we help our teams and show them how to do this. I gotta tell you, it's working, we're having a lot of fun from this. Now two things are happening in ramping time.
Are enabling enterprise sales professionals who know how to sell value and large strategic deals based upon outcome. Get my metrics because we're adding headcount. Not a lot of people are doing that these days, but we're doing that because of the tremendous growth. I look at this right now, the ramp I have a rep that was on board for four months. Q1 sold a 7-figure transaction. That is not typical, but it can happen when you bring in enterprise sales executives, pay them well, and actually set them loose because this is what they do. This is in their DNA. They've done it before, we're bringing it. In addition to that, I wanna say that there's plenty of opportunity ahead. I am extremely optimistic. I'm happy what's happening.
There's a lot of good work to do. We are actually driving a sales organization that's inspired. They're energetic. All the new folks that are coming in, they can't believe how good it is. They're like, "How can this tech company hasn't actually gone through the roof with this?" Because they really are excited about the technology. I'm excited about the technology, right? What happens is that we have to continue to focus on the value, the ROI, the business cases, and we're selling outcomes. People don't realize that. The difference is it's a platform. Somebody's gotta build the value. It's not like an application you think in a SaaS apocalypse. It's just not. What are we doing well? We're gonna continue to grow the team. Matt's pushing me to continue to grow the team.
We're not going to get anybody. I want to get the best of the best, and we're fighting to do that. A world-class operations leader. We're focusing on making sure that we're digging into all aspects of this. Why? We want to help the team succeed. My belief is that everybody in sales leadership is to help increase sales productivity and help more of our sales professionals do better. Focus on selling value. I think you've heard me say that a couple times. Could be why? That's what sells. When an executive is going to go make a multimillion-dollar purchase, they need a rep from my organization, account executives, to deliver at least $1 million deal this year. They're building the pipeline for it, and they're aligned with the senior executives.
I really focus on driving more AI adoption. Why? Because it just delivers so much value, right? It's actually working, and the customers love to hear about it. Now, in addition to this, we wanna actually continue to focus on the top of the funnel. That's more pipe out there today. Brought Scott on board. He's one of the people I have. He's been proven success at many companies. Great relationships now. Now we're doing account planning with our business partners and per account, per region. We got strategic partners kind of out there working for us. Our pipeline from our partners is up increasing as well. We're also looking into launching new revenue streams.
In addition to the Enterprise Growth Plan, which has been a massive hit from people because they don't have to count licenses, and they can just go continue to focus on building value. We're starting some to sell pilots around selling consumption because if you think about what we actually do is it's not the easiest thing to figure out how to price our technology. What we do is we meet the customer where they are in their journey. We're having a lot of fun with this because somebody may want this, somebody may want this. We just want to make sure we sell them in a contractual value that actually meets where they are today.
Sometimes people start in with one, they'll a different one, but then they most of the time, they wanna kinda eventually move to an Enterprise Growth Plan because they see the value in that. Really what I'm doing is in a sales organization is focusing on If I was in your shoes, I'd say, "Hey, Marc, what are you focusing on for AI in the sales organization?" We're starting to use AI to kinda qualify leads to make sure the leads that are coming in different kind of avenues, making sure we're gonna be using it for that. We're gonna start on the pilot opportunities in the BDR space and to bring in leads in that way.
We're looking for efficiencies, whether we're already using it with our, an organization, a tool to help us with discovery and to find out how to make sure a lot of discovery work has to happen in the sales cycle, figure out the problems customers are dealing with and to get them. We'll continue to evaluate this. I wanna kinda wrap up by saying is myself and my team 100% believes in our technology. It's incredibly sticky. It's incredibly valuable. If you have any questions for anything, feel free to reach out to me and ask me. Thank you for your time. Serge?
Thank you. Thanks. How'd I do? I did good. How we doing, team? Homestretch. We're almost there. There's coffee outside. If you need to stand back and stretch, just do it. We're almost there.
Really appreciate the patience and the attention. I'm going to talk to you about three things. Number 1, provide you a little bit more context about our ARR growth and how it divides in various different ways. Second of all, understanding our land and expand strategy, where our customers start and how we see them grow over time. Finally, talk about how we can drive sustainable growth in this business. Of course, Matt will come and join me, and we'll do some Q&A. First on ARR, this is the history. We've grown pretty consistently over time, and last year we've cleared $600 million in terms of ARR. Now we're going to double-click at it multiple different ways.
First, looking at it by product, and this is familiar to you guys because we do report cloud revenue. It shouldn't be a surprise that we're predominantly a cloud company and have been for a while actually. You can see that we're roughly 80% of our ARR is in the cloud, and that's up just slightly over the last five years. You know, based on our guidance, that's gonna continue going up.
What I will say, though, is the self-managed part of the business is actually hugely strategically valuable to us because in our highly regulated industries that are 80% of our business, customers want the option to self-manage. They want an option to be on-prem. As data sovereignty becomes a bigger and bigger issue, having that ability to self-manage is actually a strategic differentiator for us. A small part of the business, but very important. The next way to look at it is by industry. Again, here, Matt's talked about it. The big four are roughly 80% of ARR and have been for the last five years. There's a little bit of a mix shift there, so I'll talk about it.
If you look at our financials, on the left-hand side, you see that our financial vertical has consistently grown over time, but the public sector has actually grown faster. Financial services are a smaller percentage of the business, whereas public sector has grown as a percentage of the business. That's not a surprise for those of you who've been following us for a while. We've had great success, particularly over the last 18 months, as the government has focused more on efficiency. Similar story by theater. Our biggest theater still is commercial North America, and as you can see on the left, it has continued growing over time. However, both our public sector, U.S. public sector and our EMEA business have actually added more to the growth. It's more of a balanced portfolio by geography than it was 5 years ago.
This is my favorite cut, maybe. This looks at the contribution from customers who will spend more than $1 million with us versus all other customers. You can see that a significant majority of our business comes from customers who spend over $1 million with us. Those are customers who are heavily invested in Appian technology, have internal resources, have a center of excellence. We are deeply integrated with all their other systems. They use Data Fabric. Also, at the same time, they are using us as a standard application development platform. They are bringing more and more workloads onto Appian, and those are exceptionally valuable and sticky relationships. What we disclosed some of these numbers, but here is a longer history.
The number of customers who spend more than $1 million with us has doubled over the last five years. You also see that it kicked up in 2025, and that's because of the focus that we've moved to upmarket, larger strategic deals, selling with value, stuff that Mark has just talked to you about. We've seen success more recently on that front as well. What's incrementally interesting is that even though we've grown the number of customers and we get more and more customers over that $1 million mark, the average size has actually continued increasing because we don't stop once you're a 7-figure customer. We make you a high 7-figure customer. You see some of those, Mark has shown you some of those ARRs, and we have a growing number of 8-figure customers as well. Okay, that's the story on ARR.
Let's talk about land and expand. First, we've been in business since 1999, over 25 years, but we're still early in penetrating the market. In particular, you've heard us say we belong at the high end, we belong in the mission-critical use cases. Even if you look at the Fortune 500 and the Global 2000, our penetration is still low. 16% of the Fortune 500, quick math, that's 80 companies. A lot of penetration to grow. Even in our key verticals, if you just look at the Fortune 500 in insurance, financials, and healthcare, still a long way to go. Mark has been talking about some of our more recent success when it comes to winning new logos and particularly large new logos, 7-figure new logos. That's the opportunity.
That's the opportunity set at the high end of the market. Still plenty of way to go. The average size of the customer that we're bringing in has grown. This is, again, Mark talked about in the past, we were more focused on volume. We're now more focused on value. We're most focused on selling the value on the sizes of the transactions. That's showing 40% higher average size. What's even more fun is what happens afterwards. This is a composite growth curve of our customer base. What I mean by that is look at every customer cohort in every year that they've made it. All the customers that have made it to year 2, which is all the customers except the ones that we've acquired last year.
All the customers that made it to year 3, year 4, year 5. We can see that our customers grow over time and keep growing over time. My favorite part of this chart is that in years 5, 6, and 7, we're still getting value. We're still upselling. ARR is still significantly growing. In fact, if you look at the incremental ARR for our entire company last year, over a third of it came from customers that we acquired in 2020 or earlier, which just shows sort of the opportunity that we have even in what you would consider a mature customer base. Okay. Now for the drivers of sustainable growth. First, let's zoom out on revenue. You know, some quarters will be better than others, but if you take a look at over the last six years, we've delivered consistent growth.
You see here our 2026 guidance. We're forecasting in the middle of the range, $825 million. We're getting closer to that $1 billion mark, right? Meanwhile, we've continued improving profitability. We at Appian are very proud of this chart. As you can see, we were significantly negative on EBITDA not that long ago. As we focused on efficient growth, as we frankly pruned some of the investment areas where we weren't seeing the right returns, we've seen a significant turnaround, and this year we're forecasting right around $100 million in EBITDA for 2026 at the midpoint of the guide. Similar picture with free cash flow, so operating cash flow minus CapEx.
We were significantly negative not that long ago, as we focused on efficiency of our growth, we've seen significant improvements. What's interesting, these numbers include the cost of our litigation with Pegasystems, which is not trivial. For example, the $60 million in 2025 is burdened by $10 million costs of litigation, which obviously isn't a forever cost. We talk about the Weighted Rule of 40. This is the idea that we weigh our cloud growth twice as much as our EBITDA margin and calculate it at a Weighted Rule of 40. This is a very important metric, as some of us are compensated on it. You can see, two out of the best quarters in the last three years were two out of the last three quarters. We care deeply about this number.
Now we're going to switch gears a little bit and think about how the past translates into the future by OpEx line item, starting with our biggest expense, sales and marketing. In sales and marketing, we've shown significant deleveraging, or operating leverage, I should say, from 43% of revenue in 2023 to 32% in 2025. To help put that in context, we're providing a comp set here. We're looking at software companies that are $500 million-$1 billion in size and then obviously much larger companies. We're more sales and marketing intensive in the mid-tier software companies because we have a long sales cycle and because we're selling a platform. That doesn't mean that we cannot continue providing leverage and generally providing this trend over time.
What's interesting in sales and marketing isn't just what % of revenue it is, but also how do you use that money to drive revenue growth. We think about it in multiple different ways, as you would expect us to. First, this is our go-to-market efficiency metric that we talk about every quarter, and we're proud to say that it's been improving over the last 11 quarters. That is looking at our billings and divide them by our sales and marketing expense. Another way that we look at it is to look at the relationship between net new software ACV, so the new software business that we bring, and divide that by our cash sales and marketing investment.
I think of this as the purest return on your sales and marketing investment, and you can see that we've improved significantly over the last two years. There are multiple ingredients to that. Mark has talked about some of them. The rep productivity has significantly improved. Our reps are ramping faster, and while we're doing all of that, we're actually keeping a close eye on our expenses. That means that we're getting a better return. That means we're getting a much faster payback on our sales and marketing expenses. You've heard me say that we've earned the right to grow our sales and marketing organization after two years of not growing it, and this is the reason why, because we've improved the returns.
Now the goal is, of course, to keep improving returns while growing the sales org, and you heard Mark being very excited about that. Next up, R&D. Here again, you've seen some scaling from 27% to 22%, but we are significantly higher than the peer companies, both our size and the larger ones. Again, this is because we actually have a very broad surface area when it comes to R&D. We are a platform. We're not a single use case. It's a full stack set of capabilities that we're upgrading, and hopefully, after listening to Sanat and Jake speak, you have a bit of a better sense as to why that is. It doesn't mean we take this for granted. It doesn't mean that we see a significant opportunity to have operating leverage at the R&D line.
There's actually two ways we're driving this. First, over a longer period of time, we've been more aggressively hiring in India in particular because of the labor benefits that we have there. You see the jump that we're expecting in 2026. In fact, all of our hiring effectively is happening in India at a significantly lower cost. More recently, I really commend our R&D team for aggressively pushing to use AI in the development process. You see here a measure of engineering productivity. It's pull request divided by cycle time, we're indexing it to the second half of last year. Just in the beginning of this year, we've seen significant improvement. That's not to suggest there's work done.
It's just the promise of using AI to really completely reconsider and reinvent the Software Development Life Cycle. What that's going to do for us is not only help us provide leverage in our R&D expense, but actually for the same number of dollars, deliver more innovation in the market. Now is the time we want that innovation because we're having great success with AI, and we want to keep pushing it. Okay. Next, G&A. We have provided savings here. In fact, versus the median company our size, we're more frugal when it comes to G&A. If you break that down further, not surprisingly, we have a disproportionate investment in information security because of all the use cases and the regulated industries that we support.
If you look at our other G&A functions, so whether it's finance, people, IT, they're actually quite lean. Nonetheless, with use of AI and other tools, we continue to expect seeing operating leverage in this line as well. I'll save the best for last. Stock-based compensation. As a % of revenue, we're far below not just our, you know, immediate peer set, but also much larger companies. You've heard us say this over and over again. We're very careful about dilution. Because we're very careful about dilution and because of the improved cash profitability and cash flow generation, we're in a position to start returning capital to shareholders and actually shrinking our share count. Last week at our earnings call, we announced that we're increasing the size of our buyback after a strong start to the year from $50 million- $100 million.
That puts us in a position to start shrinking our share count. Obviously, this is the average for the year, so the exit run rate is going to be even more. This is yet another way in which we can continue delivering value to our shareholders and including increasing profitability per share. Okay. Let's talk about how we think about our growth algorithm using 2026 as an example. First comes revenue, of course. We're forecasting $825 million at the midpoint of the range, 13% growth. Subscription will grow a little bit faster than that. You heard about all the tailwinds that we're seeing in the market in terms of AI, improvement of processes, legacy modernization. We see a great runway to continue growing revenue. We're not going to $1 billion and stopping there. We're going past that point. Next stop, EBITDA.
Here, this year, we're forecasting just over 100 basis points of margin expansion after two years of, you know, of two years of total of over almost 20 percentage points of margin expansion. You take that 13% revenue growth and just over 100 basis points of margin expansion, you have 30%-31% EBITDA growth. Significant incremental growth. non-GAAP EPS, we're forecasting $1 per share at the midpoint of the range. We have some incremental drivers there. First, we're de-levering. What that really means is just our interest expense is going down while our EBITDA is going up, so more is flowing through the bottom line. Second, we just talked about it, buybacks. We're shrinking the denominator.
That's how you take a 31% EBITDA growth and turn it into 60%, roughly 60% at the midpoint EPS growth. As you think about these drivers, revenue growth, margin expansion, de-levering, and buybacks, all of them have room to run. All of them put us in a position to continue compounding value for our for our shareholders, and what I mean by that is profitability per share. Okay. 140 slides later, we're back at where we started. These are the four things that we're hoping you remember. Number 1, Appian is mission-critical. You've heard our customers, you've seen examples. We talked about complex, we talked about cross-functional, mission-critical, and we talked about working in regulated industries where compliance is exceptionally important and accuracy needs to be high. You heard that we're an essential AI enabler.
You heard from our customers how they're using AI within their processes while still meeting their requirements, which are significant and are not going away. You heard Marc, the success he's had in driving sales efficiency, the excitement that we see, and still the room to keep growing there. Finally, the multiple ways that we can grow profitability per share. If you're gonna take a picture of any slide, please take a picture of this one. You can keep us honest in these four. With that, we're out of slides, but we're not quite out of time. I'm gonna ask Matt to join me on stage, and we're happy to take some questions. I think we should grab some chairs. Sanjit, go ahead.
Thank you, sir, and Matt. I was able to join on this Investor Day. It's been a while since we had one like this. Great to see you sort of lay out the strategy and the plan. I was really impressed with the sort of technology differentiation that you guys pointed to. I wanted to ask a couple of different questions. The first one is on where growth is going to come from. We have these four major verticals that account for 80% of the business. I think historically there's been times to expand beyond those four verticals. It feels like in this age of AI, focus is paramount, and it feels like these four verticals is where Appian delivers the most value.
I just wanted to first for you, Matt, just sort of sanity check, gut check on going deeper in these four as the growth algorithm in terms of penetrating your TAM, versus maybe just going broader? Maybe start there.
Yeah, that's great. We don't think we need to go into new verticals to get terrific growth. We are focused primarily on the verticals that we've been on. You look at the pipeline right now in Federal, I think we've got a growth story right there.
Awesome. I guess my follow-up question would be sort of a CRO Mark-related question. I'd love to understand a little bit about this Enterprise Growth Plan. Like how long of a period of time do customers get this sort of all you can eat? Because after what happens after that Enterprise Growth Plan expires? From a pricing perspective, it's one topic that I wasn't as clear on. It's like, how do you see pricing evolving maybe along with this move up market?
Yeah. I'll take a crack and then Mark can grade me afterwards. The first thing I would say is Enterprise Growth Plan is an all you can eat multi-year plan. Usually focused on our largest customers who are ready to standardize, who are ready to bring a lot of use cases to Appian. That first example of the health services provider and the 100 plus applications that they're bringing, like that's what we're looking for, right? Those are the enabling conditions, if you will. Second of all, it's just license, right? We don't give them infrastructure, so it's not like we're facing some sort of, you know, margin issue with them.
It just really aligns our incentives really well with the customers because we let the contract get out of the way of them really driving usage of Appian, and we really see that happening. What happens at the end of it, we haven't gotten to the end of it in any of them yet, but we structure such that we see continued growth after that point to continue encouraging them to use the platform and growing the usage as long as they see the value. You may have heard Marc mention value once or twice, that's kind of the point. What Enterprise Growth Plan, it's kind of like the cleanest way to discuss value with the customer as opposed to getting lost in the P's and Q's.
More generally, as we think about the pricing umbrella, we have multiple models that we charge. Obviously we have per user, we have per app, we have consumption, we have Enterprise Growth Plans. We charge for certain pieces separately, like infrastructure. The goal is to meet the customer wherever they are in their journey. You know, and I'm sure you guys do this when you talk to actual economic buyers, people who sign checks to spend money on software. P times Q is interesting, but what really matters is the value. Are you delivering multiples of value that you're seeing? Some of the examples that like the manufacturer, the aerospace manufacturer, we deliver multiples of value of what they paid us, so they're happy to pay us.
Whether that's expressed through an Enterprise Growth Plan, that one wasn't actually. Or is it a per user or some other flavor? It actually doesn't matter. The one area where we're particularly excited, maybe not in the very near term, but over the medium term, is the consumption element of AI. As we're seeing customers be more ambitious and having more success with their use cases, they're getting to the point where that initial allotment of consumption will not be sufficient. That's a great opportunity to engage with them, to sell them more AI usage bundles effectively and get them to keep growing with Appian. By the way, when they get to that point, that's a much easier conversation because they are seeing the value. Otherwise, the use case wouldn't be growing.
Yeah. Let me just add to the Enterprise Growth Plan. Really as I mentioned, with the Enterprise Growth Plan, they have two options, right? They both include additional growth for us, right? They can actually certify their usage and continue paying us a CPI plus an increased growth rate on that. They can say, "I wanna keep doing this Enterprise Growth Plan," and we will go back to them with that offer, and we'll actually add a significant growth rate on top of that because the value they're getting from it. One of the things that we're seeing a lot now is them getting off of our competitor technologies and coming to us on this with software rationalization. A lot of customers you're talking about right now, they're figuring out they're consolidating platforms in this space.
Fortunately for us, we've been a really good landing spot for that. Does that answer your question?
Okay, thanks.
Hey. Thank you. Raimo Lenschow from Barclays. like Sanjit, I enjoyed today as well. I have two questions. One, actually it's Mark that I kind of wanted to kind of get involved again as well, since we don't see him that often. If you think about the build-out of the sales organization, I mean, there's been a lot of progress there in terms of making it enterprise ready, et cetera. It's usually a journey, you know, you need to fill a position. Everyone needs to settle down, et cetera. Where are you on that journey, in terms of having it all settled and everything clicking? If you want, you can use like a baseball analogy.
Yeah, no, I appreciate the question. I think it's a really insightful question because, you know, when you're transforming a sales organization, are you at the beginning, the middle or the end? I believe right now I'm in the ninth and eighth inning. The team we have on the field right now is very, very good. You will make a couple small tweaks. Last year was a real focus on driving large strategic deals so that we could actually hit the numbers, drive the growth, and transform the sales organization. You'll see, I mean, that's what happens, right? What we're gonna continue to do. Like we just got a, you know, like, I gotta be careful what I say here. We just literally hired in the last week, 5 very, very good enterprise account executives.
We're continuing to add headcount, and we're making sure that I'm not just hiring people that don't have the skills to do this. Some people aren't gonna be happy because we don't try to get to club, that, you know, that's been sales organization. I think we're probably in the eighth inning of this because now it's just small changes here. We have some normal attrition, which honestly my sales force is very low attrition because people see the art of the possible. They see the money they can make. We have good comp plans, and I figured that's the question I was gonna ask. We pay the teams well, but we expect a lot out of them. I would say we're in like the eighth inning.
Okay, perfect. Thank you.
Thank you.
Then on the product side, if I look at the presentation, it's like there's a lot of interesting stuff like the Data Fabric. I saw OCR as well. How do you think about your right to win? Because like Data Fabric will be very strategic for accounts. A lot of other guys will try to kind of play there in that market. Like think about like, you know, what's driving it for you that Appian will be the one because you're not going to start as the largest vendor. You're going to be a vendor for the client. Similarly for like if I think OCR looks really interesting, but like I always thought that's kind of what the RPA guys were doing. Just maybe talk to that a little bit. Thank you.
Yeah. Data fabric has a few interesting implications. We've always been tempted to spin it off as its own product. I think that when we talk about the AI stack, we've got actually two bids, right, to be part of that. One is we're the deterministic layer and the other is we're the enterprise-wide data source. Interesting, they sort of serve such complementary purposes that sometimes I feel like what we've really got is the yin to AI's yang. They're kind of at the balance where we're kind of filling the vacuum that AI doesn't provide. We will keep it as coherent as possible.
Hi. Pat Walravens with William Blair. Thank you guys for doing this great presentation today. Matt Calkins, something you said at Appian World was just because you can replicate some of this functionality with probabilistic AI doesn't mean you should, right? Something Marc Wilson talked about like just now was value-based selling of the solutions. My question is really how do you present this to your customers when you go out and talk to them? You know, in the context of seeing a number of enterprises flowing through their token budgets this year, how do you go out and show them, you know, the value that you're providing for the cost and what that looks like relative to the kind of risk-adjusted ROI of trying to replicate this with more generalized technology?
It's such an important point that you're making there about the token budget, about the cost of AI, which is frankly the elephant in the room right now, because no one's really talking about the cost of AI because it's not passed on to the customer. Today AI is heavily subsidized, but someday, and maybe in line with the Anthropic IPO or so, someday the price of AI is gonna reflect the cost.
When it does, this is gonna be 10x the concern that it is right now. We're blowing through a lot of tokens right now. People don't feel the pain. When they do, they're gonna be more interested in a portfolio approach. Not every job should be delegated to an agent. Some of them should. If you need an agent's judgment, if you need its, like, intelligent adaption, then yes, it should go to an agent. A lot of jobs should go to a rule or a bot or an API or a process in some other way. We bring the whole portfolio to those moments, and the economizing consumer of digital workers will wish to use a portfolio and create a balance. That doesn't feel like a main driver today.
I mean, you mention it, so it's not totally off the radar, but it's gonna be a much bigger driver a year from now, I expect.
Pat, since you were at Appian World, you probably talked to some customers. Our customers intuitively get this. Some just got it from the beginning. Other went and spent money and got burned. This idea of a portfolio and the right tool for the right job is resonating, that's frankly will get this in the room. That's what gets us talking about value.
Thanks, Jess.
Great. Steve Enders from Citi . Maybe following up, following on that kind of the last point, but I think part of the presentation, there's a lot of focus on application modernization and getting customers to move things. The proliferation of coding tools out there. What's the pitch for why a customer should decide to pick a platform rather than deciding to build custom code utilizing developer agents?
Yeah. Okay. You should use a platform because a platform is a reliable, modernized vehicle that will keep you safe in the future. It's also exceptionally reliable, and we're gonna guarantee that. It's connected to modern functionality like Data Fabric and common shared app applications. You can merge all of your applications onto the modern platform. Basically, and it's all that and plus the speed and the security with which you can make the migration. I think that some code stacks are gonna turn into new code stacks. I don't propose that everything should be converted into an Appian application.
For those that need the most reliability or would benefit from the power we bring or need to be combined with other applications and use common resources, I think we've got a great value proposition for that set of applications when they are converted from legacy status.
All right. Makes sense. On, I guess, the go-to-market approach again, it seems like there's a lot of focus on continuing to drive within the existing customer base and upsell those. How are you kind of thinking about the segmentation between that proliferation within the customer base versus focusing on the net new logos and how is kind of the sales force segmented to target, you know, that, what is it, 80% of the Global 2000 in key industries that you're not in yet?
Yeah. Well, we don't do 100 farmer split. We do value new logos, right? There's a benefit for that. There's a remuneration for that. We see a lot of upside in the logos we've got. We're doing both, we're doing both with the same account executives.
Where you think you'll probably see more specialization is not by hunters versus farmers, although that's a possibility, but more by industry verticals. Right now we do some of that, but we can do more of that over time, particularly as the sales force grows because our rep population is very small compared to the opportunity that we see ahead of us. We're not gonna get there in a day. It's about building a consistent journey, but that's the journey that we're on.
Great. This is Devin Au from KeyBanc Capital Markets. Really great presentation today. Thank you for that. Sorry to follow up on kind of the topic of pricing. I know we talked about a lot focusing on value, but when I look at kind of slide of you guys showing AI usage is growing exponentially, right? I guess the question is, are you guys maybe perhaps leaving some value on the table? Maybe just give a little bit more details on what are you guys doing exactly to capture more of that. Are you embedding more consumption components to capture the upside there?
Yeah. The first thing I would say, that chart was all production. We exclude like, you know, tinkering and proof of concepts and so forth. The growth that you're seeing is real customers using it in production. The last couple of quarters in particular is driven by Doc Center. That's the use case that's the broadest and where we're seeing the biggest traction at this particular moment. It's a journey, right? That's why it's important to think about our AI monetization strategy in three pieces. First, you wanna get customers onto the advanced tier, which gives them access to AI features and production. That also gives you a certain amount of usage that you get to use for that incremental 25%-35% uplift that you pay us. Okay?
We talked about on the call, 40% of our customers has some portion of their ARR estate on the Advanced Tier. The second is we continue to grow ARR from those customers and moving more and more of their estate onto the Advanced Tier. Matt showed a slide that showed Advanced Tier ARR, roughly $100 million in the first quarter, and that kind of continues growing up into the right. Then the third is what you're talking about. If that consumption keeps growing, hockey stick up into the right, more and more customers will get to the point where their moderate sort of amount of consumption that's increased is gonna build over time.
Okay. Just another question for you, Serge. There's more room for improvement there. Is it fair to say you would continue to kind of be in this modest investment capacity phase while kind of expanding that 100 basis points expansion maybe beyond 2026? Is that the right framework to think about?
We have ability to leverage every single one of our lines in our OpEx. While hoping to further improve those returns from that 0.6 needs to be going up as well. We think we can do all of that at the same time while delivering meaningful margin expansion over time.
Tim, thanks for doing this and taking the questions. I wanted to touch on wondering what's driving that. How are you able to implement faster?
In terms of the degree to which we saturate a customer opportunity that's present at the end of one value creation act and can be there, then it does accelerate them. I wouldn't say that it's helpful in accelerating reps. Yeah, on the rep side, they're ramping faster because Marc and team are hiring better. Second of all, 'cause we build real enablement muscles. The time to value, I think you heard our friend Dan Scott from PwC talk about this is because with AI tools, the aperture of what you can do, and that's what, that's what partners and customers are getting. It's an implementation of a project. Well, maybe I can't afford that.
If it's a meaningfully lower numbers, I can stack more of them in my budget and drive further automation and efficiency in the business, whether in the form of revenue or OpEx. That's what, you know.
Here. On the legacy app modernization opportunity, it was encouraging to see the TAM slide and how large of an opportunity that represents. Could you just level set us on how this has been a driver over the past couple years and how much of a step change you expect the AI unlock to drive over the next couple? Thank you.
Yeah. I'll start and then Dan fill in. In that slide of logos of people where we deliver significant value, but that's sort of the old school modernization, frankly, very people-intensive process. Some of those customers were willing to do it because of the value that they saw at the ARR about that. You've seen a couple of case studies of customers who are seeing 20 on that. As that happens, sort of incremental portions of that TAM are gonna get open. Thank you all for coming. Really appreciate it. We know that it's not easy to spend a chunk of your day with any particular company, so we are very, very grateful and you know where to find us with this follow-ups. Thank you.