Good morning, everyone. Welcome to UiPath Investor Day. We are excited to have so many of you here with us from Las Vegas for what's shaping up to be our best Fusion yet. Just a quick reminder, this presentation will include forward-looking statements and non-GAAP financial measures. Please refer to our SEC filings for disclosures and reconciliations. We have a great lineup for you today. We will take you from the vision and the opportunity ahead, to the innovation in our product roadmap, to how we are going to market with our customers and driving value, and ultimately, the financial model. We will bring everyone back together at the end for Q&A, followed by an investor reception. We have a lot to share. With that, I will turn it over to Daniel for the company vision and the market opportunity ahead.
Thank you, Allise. Hello, everyone. It is really great to have all of you here in Vegas. I am incredibly excited about what we are going to show this week. I think this is going to be the most consequential event that we ever put here. You remember a couple of years ago, we were talking about the act two of this company, and now you are going to see this act two in motion, customer stories, huge innovation pipeline that we are going to be showing. That is going to reflect really the transformation of our company. First, a quick recap of what happened since our IPO. We have built a really solid business. Our revenue grew 3.5x since our IPO. We are crossing this year $2 billion in ARR. In the same time, we considerably increased our profitability and the cash generation.
I think I am right to call UiPath a scale company, and we have transformed our go-to market operating model. That gives us a solid foundation to build on our second act, to build on the future. I think it is very interesting to see how we got there and what is in front of us, and just to put things in perspective. If we think of the automation in general, automation of business processes, that was really a big theme from, I think, the introduction of computers. Maybe that was the biggest theme. What was very interesting that happened around 10- 12 years ago, this RPA was becoming an emerging trend. You know what was the cool thing about RPA? It could have connected disparate systems. In a way, many companies had in-apps automation stories.
RPA, I think, was one of the few successful technologies that can cross the boundaries of silos. I think the story of API automation and RPA grew kind of in the same time. Today, I think they still, both technologies, have a strong place in the automation toolkit. What is going on today? It is the advent of AI. I think it makes even much more clear to all the enterprises that they can reap enormous benefits by bringing AI and automation together. That raise really the question, how this is possible? What this presentation will show you, and what Raghu will go into much greater level of details. Orchestration is an essential feature foundation mechanism to deliver AI at scale into enterprise processes. I am not talking here about Copilot types that increase personal productivity.
I mean, how you can bring AI in the context of enterprise processes and really harvest the huge increase in productivity that is basically the promise of AI. What we are doing ourselves, because I want to make sure that I make clear, orchestration, it's a fundamental trend. That's not necessarily associated with us, UiPath. Orchestration, it's a big trend. I can make a statement that I'm staying behind it. You cannot really deliver AI in enterprise without orchestration and governance to basically offer the guardrails around AI. But how we are doing in this war, how do we capture this trend? You see, it was kind of a natural expansion to us. Even I said it multiple times from the beginning of the inception of our company. RPA was, for us, really a gate to get into the big enterprise game.
We got there via RPA, but then we extended it. Our goal was always to provide a full platform to offer end-to-end process automation. This is what we basically announced. If some of you have been here around 2019, when we showed our big platform, and we call it the Tapestry, that included not only RPA, but included APIs, included process and task mining, because discovery is a big thing, and included the Intelligent Document Processing. Then two years ago, we really announced our foray into orchestration. I think it's important that we make this distinction. Process orchestration, it's not the same thing as agentic orchestration. When people say agentic orchestration, they mean a swarm of agents that have the same goal, and they can act with a lot of agency to pursue that goal.
When we say process orchestration, we mean literally complex processes, and you have an orchestrator that has to coordinate between different type of actors, which are humans, different type of agents, different type of automations. You need to allocate machines to run these automations. You need to allocate credentials. You need to offer security guardrails. You need to offer very complete governance. This is basically an orchestration platform. We made a really bold move two years ago to build our orchestration on a new workflow engine. I think our approach to orchestration is really showing a huge technology leap. Because as we said a few times, the core engine that offers the durability of our orchestration is an open source technology called Temporal. That is here for a while. It's extremely resilient and powerful technology. It's a big shift in a way.
We said our orchestration is like generation 3 technology of orchestration that is out there. I think we are the only kind of scaled company that has built an orchestration engine that offer business process orchestration on this Temporal engine. So that gives us a really good technical advantage. You can see, nowadays we have, I think, one of the most complete orchestration and automation platform in the business. This platform, it's really required. Not necessarily our platform per se, but a platform that offers orchestration and automation is really the foundation into deliver enterprise AI. It's not only me that is saying this. I have Gartner that just released, I think a few days ago, their new Magic Quadrant in business orchestration and automation technologies. Guys, look here.
To me, what makes this Magic Quadrant special, if you look at this, besides the usual suspects, you see IBM, Microsoft, SAP, AWS, Salesforce. They participated in this. They competed for this Magic Quadrant, and you can see where they are positioned. Then look where we are positioned into this one. I think it is a huge achievement for us. Given that our orchestration is relatively new, it even more speaks to the level of sophistication that we brought into our platform. You will see going forward this level of sophistication. I think it is also very important to mention that we are not only a leader in this BOAT, but we are a leader in RPA, we are a leader in IDP, and we are a leader in test automation. All of this contributes to our leadership in BOAT.
This BOAT is an overarching technology approach that really offers everything that is needed to do end-to-end process orchestration and automation. It is also interesting to I want to get more clarity. Why this orchestration and automation is It is so powerful, and it is fundamental in order to deliver AI.
One of the first argument that comes to mind is if you look at the recent, I think, challenges around AI agents that got outside their boundaries and they attack different enterprise systems. I think at this point, it is becoming clear to everybody that this technology is not ready yet. I am just putting a swarm of agents, I will give them a goal like improve my invoice processing, and they will run the business. I do not think any sane customer is thinking in doing this. In the same time, the real question is, how do you benefit of the power?
This is also an extremely powerful technology, and nobody, I think, is denying that AI is a huge secular trend that will change industries and our way we work. To me, our platform facilitates this paradigm. That is really what resonates a lot with our customers, where we put AI in the context of the business. We give AI really the understanding of the business, but let them propose, let them draft, let them suggest how we should run, and put people in the seat where they decide. Then upon people's decision, our framework, our orchestration can run the automations in a very predictable way. Look, I think when I am speaking with customers, I bring to them a series of points that I think resonates with everybody.
One of the interesting point here is when you have work that can be described in rules, and this work should be extremely reliable and predictable, why you are not using software to run that work? I think it is kind of a no-brainer today, and everybody starts to understand. Whatever you can put in software, you put in software. It is 100% reliable. I use this word, exactness, which is so powerful.
A payment, think about, you need to make your payment. Even if AI is going to be 99.9% accurate, it is still you need to be 100% accurate. It is not enough for regulatory perspective. The fact that the tool can do something, it does not mean that you must use that tool for something. I think right now it is also important to think of this. There is an asymmetry, very interestingly, into deploying AI in enterprise.
Deploying AI agents, it's actually not less complicated or faster than a year ago. But deploying determinism, exactness, it's way more faster because we are using coding agents in design time to print automations. And we get the best of both worlds. Automations work with exactness, and they are tokenless. It's impossible to run something that requires token as fast and as predictable with something that is software and runs purely deterministic. So that leads me to more like thinking, what's really the problem with the enterprise AI adoption? Because I know you'll tell me AI is vastly adopted. Everybody have CoWork or Copilot. Again, this is personal productivity. I mean really AI that works autonomously in the context of enterprise processes. I think there are many pilots, and I think most of them kind of fail to graduate to production.
The real problem of AI in enterprise is not that AI is not reasoning well, because I have many reasons to believe that AI can reason better than most people. The main difference between an AI model and a human is that AI doesn't learn on the job. Now you tell me AI has memory. Yes, memory is not the same as learning. AI will need a manual. It's like every process, every business you have will have a full detailed manual of how your business operates. No company in the world has this manual, and AI needs this manual. This is really the biggest difference today, and this is the reason that AI can fail completely unpredictable.
And look, this is what we call, this is the missing. It's the process context. I know that many people are talking about enterprise context, but I think we want to be precise. And in our world, we are bringing here to the picture a way to capture the process context, which is the way that the work happens in operations, in business processes. But now, why this information doesn't exist? Why someone cannot create this manual? Look, the problem is that information is in so many places. Think about it. It's in Slacks, messaging, emails. It's in different system of records. And it's also in people's mind. And many people try to capture this information, and you use very expensive business consultants. But the problem is this information gets stalled extremely fast because business evolves very fast.
Even in the automation world, you need to keep pace with new regulations, new improved operating model. Not only that you need to give a manual to AI, you need to keep this manual up to date for the AI to run really work. This is basically one of the biggest innovation that we are going to introduce here. And again, Raghu will talk in details about it. It's what we call the Map of Work, which is really our enterprise context around our process context that the AI needs to operate in the context of enterprise processes. My favorite metaphor is to think of this. You have the maps and orchestration and automations are the rails.
And that gives someone that has to go from point A to point B, needs to have both in order to If you go and you have a huge system of rails, but you do not have any map, you do not have any information how to use that system, you are going to be lost. These things, they have to coexist. We are introducing a way to capture the information from existing systems and keep it up to date. If you think, this is basically complete our platform. We have the foundation that offers security and observability and governance, and we have tremendous investments in task automation. We are a leader in process orchestration as well. Now on the top of it, we build this Map of Work that gives the AI the keys to your enterprise in order to operate confidently and secure.
This slide is kind of busy. What I want to try to show you here is this platform that we have built, it is incredibly complex. It is not something that you can write codes in two afternoons. This is a platform that is capable of running predictable your enterprise workloads. Guys, this is huge runtime that we are talking of it. Any automations that you print in the world runs on this runtime with scalability, with failover, with predictability built in. That is a huge offering. It is 10 years in the making, thousands and thousands of customers that Aiven It is huge number of bugs that we fixed every day on this in order to make. Guys, we have hundreds and maybe more than 1,000 engineers building this platform. We are also built with Cloude Code. That is not enough. You need to have the validation of customers.
You need to scale it. You need to see it to withstand the storms when someone attacks it. It is a huge effort that we put it underneath. This is why right now we believe we are operating actually in a much larger TAM. Because this business orchestration and automation, it is really the foundation to deliver AI in enterprises.
Again, I am not saying it is our approach unique and the only one, but I think I can make this prediction that most of the successful AI implementations in the context of enterprise process will have to have a foundation that offers them the orchestration and automation, and also the Map of Work that we showed here. That give us basically some solid arguments why we can win in this world. Look, number one, we have built, always said our platform is agnostic. It is open, and it is secure.
What this means is, you feel many workloads happen across multiple systems. I keep hearing about customers. Look, if I have to orchestrate between Salesforce and Epic, for instance, both have some kind of orchestration engine, but that will require to put the data from one into another or make some compromises. Many people will feel comfortable having an agnostic orchestration and automation engine that connects equally well to the system of records that are out there. This is what we offer since our RPA era. This is a platform that can really deliver right now the most complex end-to-end process automations, which I think it is very few companies can really like this. Gartner Magic Quadrant proves it. It is a very rare occurrence in the business to have such a powerful platform. Also, very importantly, we are not destructive to the enterprise tech.
We respect the investments that people have in their enterprise tech, and we integrate there. We are a citizen that work very well in the existing enterprise stack. Having the power to deliver a horizontal platform that can easily scale into different verticals, hey, this is a tremendous advantage. I'll make a very quick remark here. You can go into a particular process, build completely the entire stack without having a horizontal platform. That doesn't scale to the next process. Having a horizontal platform gives you the certainty of the reliability built in this runtime, and you can print verticals much faster on this one. That was, in a way, always our model.
It was land and expand, i t makes it same easy for an enterprise to adopt in a vertical, and we are coming with a few vertical approaches that you'll see later, and then expand to different other verticals. Of course, I mentioned before that coding agents creates right now this asymmetry, and in a way, this is one of the best marriage between AI and automation. AI fixes the two biggest flaw of automation. Number one was how difficult was to implement, actually to discover what are the processes best to implement and discover all the exceptions. We are fixing it with our Cartographer agent. To print the automation, you need huge investments to implement. Now it's basically almost free, thanks to AI. Also the maintenance of automations. Because always people say, "If any change happens into an upstream system, the automation will break."
Yes, that's true, but nowadays, for every exception that an automation raise, I can bring a coding agent, analyze, and fix it on the fly. Basically, we fixed all these two flaws of automation, and that's help us deliver really tangible ROI. Again, I'm talking about in a personal productivity space, AI is great, but many people told me that many customers told me that the only measurable outcome was that their employee have more time to walk their dogs. In our world, it's a measurable outcome because we reduce the time to process invoices. We reduce really the time the work is delivered in a measurable way. You'll also see it later. With this, I'm concluding my session. Next will be Raghu and then Ashim, and they will talk. I'm really happy to have Raghu here on the stage. Please, my friend.
I want to I think I said it a few time, I think it's great that you see him in person, but this guy was instrumental to bring orchestration into UiPath and he's an amazing leader for us, and he change really our approach in building software. So, man.
I appreciate it, Daniel. Thank you. All right. All right, so I'll share a little bit about our product approach, building on the points in orchestration that Daniel made, and really share with you our use case-oriented verticalized selling approach. I think it is a big strategic shift that we have been making over the last year or so that I think we are beginning to realize value for it. Then, of course, we will talk a little bit about AI exactness and how they play together. All right, so here is the agenda. For time, I will probably just skip this.
We will do orchestration. We will go into how we use Map of Work to print, build these verticalized use cases. Then we will talk a little bit about. We will share with you a few product announcements we are gonna make tomorrow to the world, but you'll get a sneak peek and a preview today.
Then we will talk about our thesis for why we are building a little bit on the points that Daniel just made. All right. So I'll start with a survey we had commissioned a few months ago, which is two out of three large enterprises with over billion dollars in revenue are using AI agents in some facets of their work, but only about 30% of them are orchestrating those agents. The challenges are on data quality, the challenges are on enterprise access integrations. Then, of course, on compliance and change management as well. The largest gaps that these surveyed customers saw was orchestration. The individual tasks, the individual agents may do okay, but when there is a handoff involved between these tasks to solve a larger, more compound complex process problem, that's when these integrations and gaps begin to surface, and this is where orchestration comes in.
So why orchestration? Daniel explained this a much, but maybe I will use an analogy to explain this even further. Orchestration is the difference between owning a fleet of trucks versus owning a logistics company. The trucks do the work, but it is the logistics company that defines the routes. It is what determines when to replace a truck when there is a fault. It is what determines when to reroute the trucks when there is a calamity or a natural flooding occurrence or so what have you.
Let us start with analogy to how work happens in the enterprise. People perform tasks. Tasks are done by your robots, tasks are accomplished by agents or by humans, by APIs, you have document tasks and so on. Then orchestration is what ties these tasks together to drive the overall process along. The problem that we see in the enterprise is failures and handoffs.
Failures and slowdowns happen during these handoffs because you are missing this orchestration. Remember I said only 30% of these enterprises have deployed orchestration. The problem is further compounding because more and more agents and more and more tasks are being built in the enterprise. What is really missing is this ability to orchestrate these tasks across the set of constituent capabilities to drive a larger process outcome. So orchestration is what eliminates these handoffs that I talked to you about and gives our clients a global view for how a process works. So you can optimize something end-to-end. You accomplish that ROI, of a larger process. You reduce costs and reduce cycle times. UiPath Maestro is that layer for us. It is the one place that coordinates and governs all of the tasks, your robots, your agents, humans, and everything that goes into building and managing complex process.
Think of Maestro as that control plane for the enterprise. Once you build your agents, it is what drives those agents to eliminate those handoffs. One of our largest financial services clients cut their onboarding time for clients from 12 days to one day, not by adding more automations, but by simply encapsulating the existing sets of tasks that they had behind orchestration. Hundreds of our customers now have deployed our Maestro in production across thousands of implementations. I want to maybe double down on the point that Daniel made earlier about us, UiPath, being the most complete orchestration platform in the world. We offer a fairly comprehensive product suite across three distinct product lines, as you see here. The reason we do this is because the TAMs for each of these product lines are different. The buyer is different, and the builder is also different.
We compete with different vendors across each of these things. Maestro Case is our approach for dynamic case-driven work. I will talk a little bit more about that. Our BPMN is where work is a little bit more structured and has less variability. It is a capability we have had for a couple of years now. Finally, we are launching Maestro Flow to the world, which is a more developer-centered, AI-native approach to building out complex business orchestrations. Three different buckets. This helps us expand our addressable market significantly, and Ashim, in his GTM section, will cover a little bit more about how we think about our overall TAM. What is common across these three layers is we have built a modern, durable execution engine on Temporal, as Daniel described. The capabilities are AI-native, meaning coding agents are used to build and manage and operate and govern these artifacts.
These are modern AI-native products that we have built over the last couple of years. Let me double click a little bit on Case. The buyer for Case tends to be case managers. Case tends to be used for more dynamic business processes. A little bit, I will have Mounish talk to you a little bit about what our unique approach to a modern AI-native approach to case management is. This is used for very dynamic processes, very complex loan origination or fraud investigation type use cases. Now onwards to business for Maestro BPMN. This is where, as I mentioned earlier, processes which are slightly more structured, have a little bit less variability, tends to be what clients use this for. Many of our clients are BPMN practitioners, so they love the fact that we offer a BPMN-based orchestration system.
This is the thing that customers call out when we demonstrate this with them is UiPath offers that one product experience where not only can you define the overall model for how your process works, but implement every single constituent part all in the same product experience. This is powerful. Your business person defines the BPMN. Your individual developers come and takes that BPMN and start implementing the tasks. These could be your RPA-based tasks or agents or have humans in the loop. Most recently, we have announced the launch of Maestro Flow, which is our developer-centered experience, which really targets that true developer, coding agents especially, is gaining a lion's share of how automations and orchestrations are built in the enterprise today. Daniel showed this, but it is worth stating again.
Our work on this in the last couple of years with Maestro has culminated in analysts, including Gartner, recognizing us as a leader. Just last year, we were a challenger. I think we are the only vendor that seen this size of shift in the year. Our investments here is going to continue to be at the same pace as we have over the last year, and we will double down on orchestration as a salient differentiator and transition from an RPA-only company to a both company, which includes RPA as a salient considerable part. Here is a slide that shows a little bit about how a large percentage of our $1 million-plus clients are using Maestro in production. We have seen this trend accelerating, and we expect this to continue over the coming months.
A few anecdotes here, which represents the transformative value that our clients are seeing, with UiPath Maestro. The slide speaks for itself, but massive cycle time reductions, massive cost savings, orders of magnitude. The thing I will highlight, I use these examples only because Maestro exceeded the customers' transformation expectations with this. With that, I want to invite Mounish on stage. Mounish is going to show you how Maestro Case, which is really what is used for the most complex dynamic business processes. Traditional case management solutions can be very cumbersome to implement, and Mounish will show us how our unique AI-native approach helps address and build these cases in a modern way. Mounish, take it away.
Thanks, Raghu, and hi, everyone. Let me take one real customer example to showcase it. One of the top insurance companies is using Maestro to completely reimagine their claims process for the agent to get up. It used to take them 24 weeks to process a claim, and the reason? Claims are dynamic, exceptions-heavy, involves multiple stakeholders across departments. With UiPath Maestro, they are able to close the same claim in just eight weeks. Let me show you how.
Okay. We switch. It is a different screen.
Yeah. This is Dana, a claims officer. She currently has 41 open claims in her queue. But notice the shift. Most of them are progressing on their own. The reason is orchestration. Orchestration is what is driving them forward. For Dana, this is a transformational experience. Instead of chasing 41 claims, she is just focusing on three that require human judgment. Let's click into one of them. Here, you can see a 360-degree view of the claim, the data from different systems, where it is in the overall process, and what is pending on her currently. If she needs more information, she can just ask the conversational agent right here, and the conversational agent will respond with the claim's context. Currently, we can see that this claim is in medical evidence and approval stage. But how did it get here? What's under the hood?
Now we switch to Maestro Case, our AI-native agentic case management solution. It takes that complex, messy process and converts it into a structured observable system. You can see that the claim progresses through stages, and within each of those stages, there are different tasks. RPA for those legacy systems, API for those modern systems, agents, both UiPath agents and third-party agents, and human-in-the-loop decisions like Dana's. Now, if I zoom out, you can notice that there are no edges connecting any of the stages. There is no predefined order. The reason is the case manager that is sitting on the top.
Mounish, speak a little bit about the salient differentiator that we have with traditional legacy case management systems.
Right.
Because I think the task that you showed, where the task can be very complicated. Enterprises run mixed estates. You need RPA to gain access to that system that was deployed in 2005, and then maybe an agentic approach to a more modern system. Also our approach to case agents and how we dynamically creation that process. Speak to that.
Great question, Raghu. Our biggest differentiator here is the case manager that you are seeing on the top. It is a specialized AI agent built for long-running processes, and it has only one job here, which is to determine what is the next best action for this claim so that it reaches towards resolution. Currently, if I zoom in, it is pending on one thing. It is waiting for input from Dana so that it decides what needs to happen next. Let me go back, click into the claim, and open that particular task. We can see that the claim is currently waiting on new medical information. Let us upload the new medical records that have come in and add a note. The interesting part here is the new medical records that have come in are completely different from the ones that we initially submitted.
What this means is the claim needs to take a different path. Let us see how the case manager handles it. Once I submit, let us go back. The case manager is always listening for business events, and the document I just uploaded and the comment I just added were exactly those. You can see the case manager thinking and deciding what needs to happen next. What it does is it uses the event payload that we just submitted, goes through the entire claims context, what has been completed so far and what is pending. Combine deterministic rules with agentic reasoning to figure out the next best action. You can see the different stages that are being fired off. Everything—
Mounish, is it fair to say that the case agent is like the enterprise, it is like the harness that we put. It has the stages, these tools, the models, and automations are skills that it uses to make progress.
Exactly, Raghu. It is a specialized agent built for long-running processes. It is all governed and auditable. If I scroll down, I can see the entire execution trail, the different decisions that the case manager has made, the different LLM calls and tool calls that it made to come up with those decisions. When I click into each of those decisions, I exactly see a reasoning behind each one of them. This is what makes it unique and differentiated. It is AI native from the ground up. This is applying intelligence at the process layer. If I zoom out, all of this might look very complex. The question is, how did I build it? Let me switch to my development environment. This is where I used UiPath for Coding Agents, where I provided my requirements, and UiPath for Coding Agents used UiPath skills to generate this entire solution.
The case plan that you see, 14 stages, 38 tasks, 400+ nodes within those tasks, and 200+ rules to govern the case agent that we just saw in action. Not just that, it also created the business app that we were seeing for Dana. What used to take months of development now happens in days. That is the shift.
This is where Daniel's point earlier about AI proposes.
Yes.
You are the developer. You're still in charge of the developer's decisions, and automation executes. Really is playing out. Now, this example, is this a demo or is this literally what you used to build this?
This is what I literally used to build, and it took about 46 minutes to generate this entire case plan. Then I deployed it in our orchestration software, and then the orchestration is what is executing this durable process.
How long would it take without, A, the case agent that you described, and B, UiPath coding agents?
I think without UiPath coding agents, it would have taken at least a month for me to come up with a case plan, build all the tasks within the case plan, and then define the rules that are required to govern the case agent.
That is awesome.
Now, to sum up, there are three things that we showed. Number one, an AI native way to build complex orchestrations using UiPath coding agents, an AI native way to run them via the case manager agent, and number three, govern and operate them at scale on Maestro. This is how one of the top insurance companies is seeing transformation benefits from Maestro. They are not alone. Customers across verticals and industries are seeing the same.
Thank you.
Thank you.
Thank you, Mounish. This is amazing. We strongly believe at UiPath that in 2026, this modern, AI native case construct is the way how cases will be built and processes will be executed and managed. Thanks, Mounish. All right, actually, let me go back a slide. Daniel introduced the concept of the Map of Work earlier. As he mentioned, process context is scattered around everywhere in the enterprise. The map brings that structure, the meaning, so that agents, humans, can reason and understand over the process and have the same understanding of how a process actually works in the enterprise. The map is living, meaning as the business changes, as the client's business changes, the map also changes and keeps there. The map is 100% owned by our clients. It is governed, meaning changes are audited. It means that they are versioned.
Let's talk a little bit about what the map has. It has structured knowledge. It has the case plans, what Mounish showed a minute ago. It defines how work is structured and how business processes move from the beginning to the end. It has policies and rules. What's allowed, what are the thresholds, who approves what? That type of thing. Then it has the ontology. An ontology is really a way of saying that there's a shared description for business objects that our clients use. Things like customers, things like their orders, things like their invoices and their relationships. Having this ontology allows every system, every agent, to understand the business in the same way.
This is how you take all of the unstructured scraps of information with the Cartographer agent, which I will talk to you about in a minute, and convert that to something structured that agents and humans can reason together. The next bit, the next piece in the Map of Work is what we call operating knowledge. This is what experienced people in the enterprise know. This is institutional knowledge. These are the people who provide trinkets. These are the people who provide worked examples. These are the guidance that people carry. Now, when these experts leave you, many enterprises lose that knowledge. But with the Map of Work, we build the systems and capabilities to capture as much of this operating knowledge as possible, as you will see.
Finally, the Decision Ledger is where the decisions and judgment that human makes and the rationale and the reasons for it are captured. This becomes the goldmine which our specialized agents can use to drive that continuous improvement for processes. This is a magical thing that the Map of Work can provide. Not only do you have that shared enterprise context or process context, but you can use that process context as a living, breathing artifact that improves as your process knowledge improves and drives continuous improvements to how your automations are built in the enterprise. All right. You might wonder, how does this map get drawn? This sounds like a very complicated thing. It really is not. I think I will tell you why UiPath is uniquely poised to solve this very complicated process context problem.
How do we go from this unstructured, siloed information, I will even use the word chaotic artifacts of information, into something that is more structured? There are two answers. First, a person. We are defining this new role that we call the Cartographer, someone who owns the map, someone who curates the map, and someone who keeps the map up to date in the enterprise. Now, you might wonder, is this a new hire? No, it is not. Our clients already have business analysts. They know how the business runs. They already know how to translate business requirements and business lingo to technical teams that do the actual implementation. This is the business analyst role elevated. What elevates this business analyst to be a Cartographer is this new product we are launching called the UiPath Cartographer.
This is the product, i t is really a Copilot that guides the Cartographer to pull all that context and give the structure that helps drive a shared understanding of processes. The Cartographer, the human, is in charge of defining SLAs. They are in charge of defining the outcomes.
The UiPath Cartographer, the agent, the Copilot, assists the Cartographer in gleaning that context and helping create this map in a manner that is shared. Now, the Cartographer alone cannot possibly map a very complicated business process. What Mounish showed is not something one human being knows about. The UiPath Cartographer natively supports collaboration across multiple stakeholders, and this is critical. We have learned over the last decade how businesses actually manage and map their processes, and we have taken that understanding and brought it into how we built the UiPath Cartographer. In addition, our customers, business analysts, they think about process re-engineering all the time.
It's not about just mapping how processes work, but they are the ones redefining how processes should work to get that next level ROI. UiPath Cartographer has inbuilt capabilities to guide the cartographer to help assist with process re-engineering. Take these couple of concepts. You drive process re-engineering, you map all of the work in one shared artifact, and then you drive a collaborative experience with which you glean and gain this enterprise context that helps you drive this transformative value that mapping can help provide. I put this all together so I think we can all reason through the massive opportunity we at UiPath see in communicating with our clients and having shared this vision with them. Step one is cartography, is what I just described to you, what Daniel described to you a few minutes ago, is we capture that process context with UiPath Cartographer.
Step two is you take UiPath, what Mounish showed you. The map is fundamentally written in a way that agents can reason with, which means that even UiPath for Coding Agents can take the map and build out these automation artifacts. This is transformative. Our clients are finding that both the examples that Mounish showed and a few others from our clients, 40- 50 minutes of work where the agent goes off on the side and just builds these things out is truly transformative, and you'll see a few demonstrations of this over the next couple of days. Then step three is Maestro. Maestro, in its various incarnations, orchestrates these very complicated processes in a governed, auditable, and safe and secure and resilient way. Finally, I think step four is an important step that we're also adding to our product vision.
That is this concept we call the Decision Ledger. Human decisions, human judgment is captured. Whenever a human does an override to a policy or makes a decision, we capture that in this thing we call a Decision Ledger. The inputs in the Decision Ledger are fed back to the map to drive continuous improvements to a process. We have specialized agents that can reason over the artifacts that are in the Decision Ledger and assist the cartographer in curating that information and drive this improvement. So that is the cycle. You capture with the cartographer, you build with UiPath for Coding Agents, you run your complex process with UiPath Maestro, and then finally you improve the processes with the Decision Ledger. The Map of Work at the very heart of all this. So you're probably thinking, how do customers get started?
It's almost never from a blank slate or a blank map. I mentioned to you earlier one of the things we have embarked upon over the last year is this verticalized use case selling. As it behooves us, we've built about 100 or more of these pre-built maps that we'll show you in a little bit. Now, where you start depends on your process. There are some processes that run about the same way in every company within an industry. They have very little variability. For those, we've actually completely pre-built out the map, but we've also built out all of the automation constructs that go with the map and curated business user specific experiences. But many do not. Many processes have much in common, but there's enough variability that we can't possibly pre-package them, and we support that, too, and I'll explain in a little bit.
There are a few processes that are completely bespoke to a company. That is what maybe makes the company as unique as it may be. There is kind of three places to start with the map. The first and the best way is what we call the pre-built maps with UiPath solutions. Some processes, as I mentioned, look about the same way in every customer in a particular industry. We have drawn that entire map, and we have implemented all of it. Every single agent, automation, any automation artifact that is required for us to operate the process, we have built it out, and we have built curated experiences for the business user. These are our turnkey products.
What you see in this slide is a little bit of an eye show, but what we want to tell you is that we have built these verticalized processes for these pre-built maps for a very few industries. Healthcare, which is where we have some significant strength in, financial services also, and utilities in a few cross-industry use cases. We have chosen these use cases and industries very specifically. A, we have the strengths in these industries, like I mentioned. Our teams have done TAM and SOM analysis for where the highest growth and revenue opportunities for UiPath is. We have used that to select this list. Finally, we had customer demand within our own customer base that we have used to see which ones to seed these pre-built maps with. Now, as I mentioned, these solutions are completely built by UiPath.
Our clients manage their business, and we take care of the rest. There is only a few configuration artifacts that they have to provide, things like their rules, things like their policies, for which we provide curated experience, and they capture this. Martijn is here on stage, and he will walk us through how this actually works. I want to maybe leave you with one important point before I hand it off to Martijn. We picked a process we call the source-to-pay, specifically because we expect most of you here to be reasonably familiar with it. You could learn about how the platform makes these pre-built maps with UiPath solutions possible. We have more complicated processes, especially one on denials, which has, I do not know, 400,000 plus payees, payer policies across 100 payers. The reason we did not pick that process is because it would be far too complicated.
We would probably spend more time explaining the process than really the concept of these pre-built maps. With that, I will hand off to Martijn to use this simple example as a demonstrative use for how we think about pre-built maps, UiPath solutions, and printing these processes. Take it away, Martijn.
Thank you, Raghu. Let me set the stage here. At Cobalt Ridge, the source-to-pay process was a complete mess. Maverick spend was out of control as people kept bypassing procurement and were not making use of the negotiated discounts and contracts. This also led to a host of problems downstream in the accounts payable department. The team was being swarmed with invoices that are riddled with exceptions. Most of these problems originate at the beginning of the process, the request intake. What about you, Raghu? Do you ever create any purchase requisitions?
Several, and not as this.
I hear you, man. Creating purchase requisitions in a legacy SaaS procurement system can be a painful experience. Let me show you what we do in the UiPath source-to-pay solution instead.
I think you have to switch the—
I'm not sure. Should I switch?
No, hang on. Yeah, do it.
Okay. I should be seeing my demo screen. Awesome. Okay, let me show you how we do this in our source-to-pay solution. We make buying as easy as asking. As a manager in Cobalt Ridge, what I can do now is ask what I want. I want five laptops, four developers. In the background, the solution is using the business ontology to understand my intent. It knows what I need and who I am, and it's using that to gather this data from our business systems, from our SAP, from our Coupa, from our Workday. That's not the only thing we're doing. We're also checking my request against the Cobalt Ridge purchasing policy to make sure we only surface the products that I'm allowed to buy in this situation. This laptop looks great for my use case.
I confirm the shipping address and boom, I'm done. It's that easy. From here, Maestro will take over the approval process. It will start reaching out to the users that need to approve on Teams and Slack. We get a response fast and keep the process moving. How about that, Raghu? It made the purchase requisition for us.
Yeah. So Martijn, this looks almost too simple. I am assuming I am getting the investor that has seen this and like, "Man, I have already seen this. What is going on here?" So show us what is chunky, what is the power of the platform underneath that makes it possible to build something like this in an afternoon of configuration?
Yeah, absolutely.
Yeah.
So like any UiPath solution, the source-to-pay solution comes with a completely pre-built Map of Work. So here you see that visualized. It comes with all of this out of the box, a case plan for source-to-pay, a business ontology describing the concepts, business rules, and policies. It is all there from day one. That is not the only thing. It also comes with a complete user experience, agents, dashboards optimized for all of these personas based on user experience. The only thing the business team now needs to do to make the solution their own is to configure the last mile, their own rules and policies. Want to see how that is done? Let us take a look. So here, as a process owner in procurement, the only thing I need to do is to upload my own purchasing policy.
It's a 30-page document describing everything about who can buy what at Cobalt Ridge. The solution already understands the principles of source-to-pay. So it used that to translate this written document into a set of deterministic rules. Here's the rule we just triggered for our laptops, describing what kind of laptops can be bought for engineers at our company. Run this rule 100 times, you will get the same result every single time. Let me see. Okay, so it was real easy to get this set up as a business user.
Now let us move over to the accounts payable team because they were being overwhelmed with incoming invoices before, but now they are back in control. Here' s a process owner, I can see all of the invoices that are coming in. I can see what my team is doing, but also what the agents are up to.
They are doing a great job, actually. 88% of invoices are now being handled completely straight through. Here is the one for our laptops, for example. You see the agent picked it up. The system picked it up, I should say. It extracted the data from the document, matched it against our systems, performed all the necessary checks, and posted it for payment. No human action required whatsoever. Hey, wait a minute.
There is one invoice there that is blocked right now, so the only thing that matters now is how fast we get it moving again. Let's see if we can resolve it. It is an exception of maverick spend. We are missing the purchase order, and also there is no requester listed on the invoice, so I cannot start the approval process automatically now. Luckily, the agent used business ontology to find out who could be the requester.
It is suggesting it might have been Kate. She is in similar requests before in the last six months. It is also coming with a set of suggested actions. We could go to the expert, Dana, or go to Kate directly. I will do that. Again, Maestro will chase Kate on Teams right now to make sure we get this invoice unblocked fast. That was the exception.
Now let us move over to the Decision Ledger. You heard Raghu tell you earlier, this is where we are capturing all of the decisions made by the end users. We are using that to optimize the solution over time. Here I can see all the exception rules, and this is the one we just triggered when a requester is missing. It is not performing well. 42% of the time the people chose the suggested action, for the rest, they didn't . How does it improve?
The ledger already provides a suggestion of what we can do. It proposes that if there is one likely requester, we should go directly to that requester instead of bothering me. That is a good improvement, saving me precious time.
Martijn, from the business standpoint, what is the ROI that a customer sees with this?
Great question. Raghu, all of our solutions. Let me stay here. All of our solutions come with a set of pre-built KPIs that matter for this process. For source-to-pay, it is all about reducing maverick spend and increasing our touchless rate. I can see here that the process is performing much better than it was before we got started with it. We also saved over $300,000 already, so that is an outcome. Here' s what we just saw. The source-to-pay solution for Cobalt Ridge started in days instead of months. It allowed them to get fast ROI and outcomes that matter. That is essentially what all UiPath solutions do. Over to you, Raghu.
Awesome. Thank you, Martijn.
Thank you.
All right. As you saw from Martijn, the ability to use the UiPath platform and the concept of a pre-built map to make wholly built, complete products as solutions to customers to solve their process use cases that they really care about. They did not buy the components, they did not have to hire developers to build these things out. We had pre-packaged maps, fully built automations for them to go use, and that is what you saw. All right. What do we see here? All right. On the other end of the spectrum, where a customer may have a very bespoke process that is uniquely just their own, our customers can build their own maps. I introduced you to the role of a Cartographer. The Cartographer would build out these maps from scratch, of course, with the Cartographer agent doing the heavy lifting.
So there is another way for clients to use Cartographer and the maps. There is a middle path, and we are proud to announce Process Atlas. A Process Atlas is a collection of maps for a set of processes that are common enough that we can give it some shape, but have too much variability for us to complete those maps. So we give the maps the domain knowledge, which is drawn by experts, and then your Cartographer adds that institutional knowledge to complete the map so that it captures what is unique to your implementation. This is a long list of pre-mapped processes that we are going to launch with Process Atlas tomorrow, actually. We chose these processes, again, very deliberately based on our assessment of sums and times and where we found the best product market fit for our existing customer base. All right.
With that, I would like to invite Anvita, our product management leader, to walk you through how customers realize value with Process Atlas and Cartographer. You will see Anvita show how our history and strengths with mapping and building business processes really sets us up to collaboratively build these first-class assets and build out these maps in ways that we believe only UiPath can. You will see through the demonstrations how that works. So Anvita, please take it away.
Perfect. Thank you. Let us turn to yet another complex long-running use case. Large manufacturing equipments, when they break down, need to be covered for insurance, need to be contained for a repair or replacement, and that costs manufacturing companies a settlement amount. Cobalt Ridge, one of the illustrative manufacturing companies, needs to pay about $42 million a year in coverage cost. Here is the problem. Their mean time to resolution, call to close, is about 11.4 days, and their target for next year is under four days. Here is where the problem really lies. Here is how a single case of warranty resolution gets operated end to end. A version of this for every case, and thousands of them operated by Cobalt Ridge every single day.
There are people across teams that are contributing at different stages of this process, and no single person or team within Cobalt Ridge might have seen this entire picture. You just heard about the concept of a Cartographer. Me, as the Cartographer on this engagement, is tasked to map how this process runs today, find how a reengineered process should be implemented so that we achieve our business goals of reduced call to close time and reduced coverage costs. Let us see how we do that. UiPath Cartographer allows my ability to build and maintain this map with an accelerated time to value. Here is where I begin, UiPath Process Atlas. Hundreds of complex processes mapped by expert attestations with deep domain knowledge across the core seven industries we operate in.
T hey are all mapped across business areas and personas, and f or my work, as I pick warranty resolution in the aftermarket and service business area, I can see a whole map built about how warranty resolution operates within a manufacturing industry. Sample business KPIs followed by the complete case plan. Multiple stages from intake to triage, then doing the diagnosis, and eventually doing restoration and closing the case.
Along with conditional paths when a case goes sideways and evidence that never arrived, having engineering exceptions to deal with or substitutions of parts that need to be reviewed or product quality escalations need to be managed. Every stage marked with a human role on who should be taking decision at every single stage, along with standard business ontology of which are the typical systems and business entities that go in managing this process. It is all underlinked and underpinned to SLAs across stages for the entire process.
It all underpins with standard coverage rules and policies that govern the entire process. This becomes my starting point. As the Cartographer on this engagement, I build on top of this to complete this map, to give it context about how we at Cobalt Ridge will add more information on how we manage our systems, our handoffs, which systems do we track, so I can complete this map and build out my map for warranty resolution.
The 45% number at the top left there, I think it indicates the domain knowledge, the starting point that our experts, our attestations got us to. There is a sea of red here. How does the cartography agent help you, the Cartographer, into accomplishing the— How to prioritize it from the set of actions that it can possibly take to help you improve that number to a higher number?
You are absolutely right. The 45% coverage got me from the Process Atlas skills, the baseline of how a process typically runs in an enterprise setup. A bunch of reds indicating clarifications that I need to provide. My starting point with UiPath Cartographer is to initiate a conversation providing it all the manual discovery documents that I have had. Steps of process interviews, workshop transcripts, sample architecture diagrams, even full sample claims with hours of hand-annotated description about how lines that were down for hours were dealt with.
This is literally the scraps of unstructured information that we are going to provide structure for?
Essentially.
Yeah.
I can simply prompt the UiPath Cartographer agent to document what I provided it. Where the sources contradict, highlight it for me, and I continue to be in charge of approving all of those deviations. One such that it will highlight will be about when there are discrepancies observed on processes that were not observed within the UiPath skills. I continue to spar with the agent to provide it an input on how to continue building this map based off of how we at Cobalt Ridge operate this process. It also highlights inconsistencies, what it observes between the context I provided and the skills it already has. One such where it mentions a couple of stages were missed out in my base map. Claims denial and claims withdrawal. Upon my confirmation, it is able to add that into the Map of Work.
It continues to hydrate it with the context relevant to my enterprise. A few iterations out, we reach at a higher coverage. 45 moved up to about 68%.
Yeah. A lot more greens, but still some reds. I am still curious, how does the agent help you prioritize from all of these unfinished tasks?
UiPath Cartographer is a goal-seeking agent. It starts by listing out what the process KPIs are at the moment as a baseline, so it can figure out how the process reengineering solution should work. 11.4 days, our call to close time, the baseline has been mapped, and this becomes an anchor for UiPath Cartographer to provide reengineering solutions. It built out the entire case plan based on my manual discovery documents. It was also able to add the additional stages it observed as missing from my skills but is relevant for my company. It also looked at mapping all the individual data entities for business ontology, along with the score systems that we at Cobalt Ridge touch in. Helios CRM, AssetVault systems, and more.
It brought in all the rules that we at Cobalt Ridge operate across individual stages and steps within this process, along with mapping individual departments and people and personas of who does what and at what stage. As Martijn noted, there is still a bunch of reds. At this point, me as the cartographer, I need input from my business teams. There are a bunch of rules and exceptions that I need inputs from my business teams in the warranty resolution department to confirm how some of these exceptions have been handled. UiPath Cartographer supports that collaboration natively. Let us see a quick example. Here is one exception on how the UiPath Cartographer mentions that high-value claims over $10,000 need a VP approval, but there are open questions on how these exceptions are really handled. I need input from my business teams to give me insight on that.
Ideally, show me how they do some of these exception handling. From within the tool itself, I can initiate a handoff. The agent is able to draft out a sample question with evidence attached and the explicit question that I need, and I can simply create a handoff to my warranty officer, Scott, in this case, who can receive a notification on any communication channel for Cobalt Ridge and his teams. This provides me to really scale building this map of work through collaborations with SMEs. A few days out, as I start receiving inputs from business teams, all of those inputs get captured as feedback coming within individual sections that the map continues to build.
The agent is able to take all of those inputs, so it can add it to the map, redlining the sections that were updated based on the SME input that I received and is able to continue hydrating the map based off of all the inputs with SMEs that keep coming to me over the next few days. Just like that, mapping scales for me. From the 68% coverage, I reach a number which I feel confident about to hand it off to my developer team to go ahead and implement the re-engineered process. What used to take me months of manual discovery, chasing teams, identifying what the exception handling paths are, has shrunk to weeks, if not days, with UiPath Cartographer. This all resides in a living, governed asset within the Map of Work.
The case map built out with individual stages capturing how work happens today, and a re-engineered view of how work should happen, anchoring to the expected benefits that we started with. Layering on what the baseline of meantime to resolution of warranty is and what is the FY 2027 target. This is how cartography scales building process automation opportunities at scale.
That is awesome, Anvita. Thank you so much. As you saw from this demonstration, our strengths and heritage with understanding business processes allows us to use Process Atlas to help cartographers like Anvita map out the most complex business processes. Not only map it out, but also re-engineer them because of, again, our experiences with what the best ideal use cases and processes should be. All right. UiPath Cartographer is generally available. We are going to announce that tomorrow on main stage. UiPath Process Atlas, the 100-plus skills that we have to hit the ground running, is also going to be announced as generally available tomorrow. UiPath for Coding Agents is generally available. This is what helps our clients automate their processes, but manage it across the whole life cycle from build, deploy, govern, and even continuously improve, as you saw in a couple of the demos.
An important observation that we found is customers of ours are using UiPath for Coding Agent are finding about a 59% increase in their developer productivity. They are finding that those same customers are finding about 2x to 5x increased deployment velocities. The throughput of their developers is increasing almost by an order of magnitude. I want to transition to our test section now. You saw Cobalt Ridge in action. You saw cartography, mapping of work, and building out automations with coding agents. What this really means is the power of cartography and the power of coding agents really accelerates the time to value for our clients. They can get a lot more done. Lots more automations get built, lots more agents get built, lots more things get orchestrated, lots more lines of business applications are also built.
These systems are all built on mission-critical applications like you see here, ServiceNow, Oracle, SAP, and what have you. The resiliency of these applications are fundamental to ensure that all your line of business applications, your automations, do not break. Because if they break, obviously your business comes to a cross. What we are finding happen is because of the proliferation of these coding agents, the volume of what developers are able to produce has increased manifold. More code, more change, more automations. This is causing test debt. Test teams are struggling to keep up. The gap is showing in application quality. It is declining, as you can see here, even as developer productivity is improving.
And for that reason, at UiPath, we are taking a novel AI-centered, agent-centered, coding agent-centered approach to providing the power of automation, the power of coding agents, the power of transformative AI value, just as developers have it to building software, to also testing software. And for that, Ingo, I would like to invite Ingo, our VP of product, to walk us through Test R factory. Ingo, take it away.
Thank you so much, Raghu.
The clicker is yours, old man.
Here you go. Raghu is always spot on, ladies and gentlemen. To put it plainly, software development is accelerating, but our ability to test isn't. That means we are creating risk faster than we can test. And that is exactly where UiPath Test Cloud comes in. That is our market leading offering for all things software testing here at UiPath. And today, we want to take you inside one specific capability of Test Cloud, the Dark Testing Factory. Because the Dark Testing Factory is the solution to fix the challenge Raghu just outlined for us. It is where software testing stops being something you run and starts running for you. And that shift towards more autonomy in software testing has to happen. And you do not need to look too far to see why. Just look at how software testing typically happens today.
First, testing isn't just one task, it's a whole suite. Designing, automating, and executing tests and everything in between. In many organizations, much of that is still done by hand, entirely manually, or at best supported by traditional rule-based automation. That is the manual grind that typically slows us down in software delivery. While rule-based automation helps, it only gets you so far because not every task in software testing can be reduced to a fixed deterministic rule. That's where agents come in. Think of conversational agents like Delegate or Autopilot, and suddenly more becomes possible. Here's the thing, one agent here, another agent there, these are all disconnected pockets of agentic intelligence. That's not enough to keep pace with software development. The real shift happens when you start connecting these islands of intelligence into governed autonomous flows.
Flows that spin up and tear down entire test environments, that execute test cases across stages and analyze results, all autonomously while humans stay in control. This, ladies and gentlemen, isn't a future vision. That's what our customers are already doing today. The lights are already dimming. Our customers are making their testing more and more autonomous, or progressively darker, if you like. Every step up in autonomy brings another big jump in testing speed. That's the journey we've seen our customers take so far. The Dark Testing Factory brings that journey to its next level. You can think of it as autonomous testing on steroids. Here autonomous testing is no longer a point solution. It becomes an operating system, Raghu. An operating system that allows our customers to turn their testing from something they run into something that runs for them.
Ingo, this is amazing. I share your intuition. How does this work? Let's show that.
Absolutely. Let's see it in action. Here you go.
Thanks.
For this, ladies and gentlemen, let's go here into UiPath Test Cloud, that is the place where you design your Dark Testing Factories. Let me show you one we specifically built for SAP S/4HANA here at Cobalt Ridge. Let me start where the work actually begins. What you're seeing here is the blueprint of the factory. You see how work enters the factory, how it flows through, and how work gets done. Remember, you do not run the factory. The factory runs itself. How does this work? The factory is constantly listening to anything that changes your SAP application. That could be a new transport entering SAP, or a user story closing in Jira, or a commit directly coming from your deployment pipeline. A change is flowing into the factory, and from there, the factory takes it over.
It creates a plan to test that change, breaks that plan into actionable work items, and then distributes those across all the tools available to the factory. Raghu, that's the big picture view of how the Dark Testing Factory operates.
Ingo, do you mind double-clicking on one of these, like maybe the risk assessor, and actually show what the value that provides? Also maybe speak to our investors here on some of the differentiators here. What was the pre-Dark Factory cost of developing the risk assessor capability and post-Dark Factory we have learned?
Absolutely. Let's first dive into the risk assessor, and then let's focus on the value add. Let me zoom into one of the tools the factory has available to get its job done. Let's dive into this one, the risk assessor. As you can see, this tool is not a single agent or a single automation, it's a broader agentic workflow. An agentic loop, if you will. This specifically is here to identify the business risks identified or associated with an incoming change. It also consolidates those risks with the factory's broader risk model to determine what needs attention first. Here autonomy is always risk-driven. Here is where it gets interesting. You can even see these agentic workflows at work. For example, here is the risk assessor currently busy processing an incoming change.
You get all the live execution details, and you can see exactly where the factory stands. You can literally see the factory thinking and working in real time. Importantly, those agentic loops do not improve themselves. At the right moment, it's either an agent or it's a human, as you can see, that jump in for review and approval. If a person is needed, well, they are looped in right here in the inbox, where they, for example, provide missing knowledge to the factory, they help clarify an exception, or they help the factory to make a trade-off. That is the role of people in the factory. Humans do not run the factory, they teach the factory.
Every time humans teach the factory, the factory improves its memory. This memory doesn't just get better from what people teach it also gets better by learning from its own mistakes. This is what turns the factory into a self-improving system. Raghu, this is what turns the factory into a system that gets a little smarter, a little more autonomous, or a little darker over time.
Yeah. Ingo, I totally see us bringing the concepts that we've learned with developing software to testing software, to testing Dark Testing Factory. Now tell me, just as in development of software, the human is in charge. AI proposes, but the human actually is the ultimate decision-maker. How does that work? How do you manage costs? How do you manage governance controls, governance? How does that work?
Absolutely. That's exactly it. Autonomy without control is not the goal here. That is why the Dark Testing Factory is not only deeply embedded in our platform, it's also deeply embedded into our AI trust layer, where you can, for example, decide how much autonomy you want to grant the factory and all of its components. You can also define cost and budget controls down to the very last level. You can also, of course, wrap policies around how the factory operates. That means the factory doesn't just act autonomously, it acts securely, too. Not even that is the goal, ladies and gentlemen. We are not making software testing more autonomous here at UiPath just for the sake of it. We are doing it for a reason.
What this is all about is helping you to answer one question we all have to answer at the end of the day, are we ready to ship this software application? As you can see, the Dark Testing Factory helps you to answer that question with evidence. No gut feelings, no intuition, facts. The factory does not just help you to answer that question for one single application, but for every application in your entire enterprise. That gives you the confidence you need to ship fast at scale without leaving quality behind, Raghu.
Thank you, Ingo. Thank you very much. We are at time for the product section, but I will quickly recap. I think what we showed you is our momentum with orchestration. Hope you saw that, and we can take questions a little bit later. You saw how we are going all in this model we call the Cartographer, with the cartography agent to build out the Map of Work, which then is fed to coding agents to truly transform how process context is captured and how automations are built and managed in the enterprise. With the continuous learning loop, where you can continuously keep this map up to date, but also your automations up to date and ever-improving.
Finally, I think as more software and more code is being built in the enterprise, UiPath Test Cloud enables us to actually rein in the risk that comes with more software being written. This is our product strategy. This is our vision. We hope to launch it to the world tomorrow, and we are very excited for it. Thank you for joining us. Then I think I hand it off to Ashim, I think. Do we have a break? We have a break now. Thank you.
We will now take a short break. The event will resume in 15 minutes.
[Break]
All right, everyone. We're going to get started with the second half of the session. We had technical difficulties in the first half of the session, and we just wanted to let everyone know that the first half will be posted to our investor relations website in about 30 minutes. With that, I'll turn it over to Ashim, who's going to run us through our go-to-market strategy.
Let me get my picture up quickly. Thank you, Allise, and thanks for everybody to come here, to Las Vegas. I was just thinking about it. It's been almost 10 years for me at UiPath, and when I look around, it's amazing to feel like you're talking to friends who have talked with each other for so long. As you hear the company story, I'm just thinking as I was in the break of just reflecting on Raghu's pitch. The sophistication of the platform versus 10 years ago is incredible. I remember meeting Daniel somewhere around 2017, and there was really a box on the page called RPA. Today, when you're looking at Cartographer, you're looking at agents, you're looking at orchestration, it is incredible about how far we have come.
In some ways, during the nine years, we've had both a period of evolution and a period of revolution. What I would say for UiPath is right now we are doing both simultaneously. For the next 25 minutes or so, I want to show you how that's transforming over to go-to-market and to show you what we're doing, how are we approaching our customers to capture the opportunity. The opportunity. I think Daniel talked about $142 billion TAM. That is one frame to see the massive opportunity that's sitting in front of us. What we see, you can see in the back, you'll meet a lot of our sales leaders here.
Over the last nine years, we've actually seen an unprecedented level of activity, whether that's POCs, services, driving thing orchestration into production, customer inquiries, or just the sheer amount of work in enabling our sales team to be talking and responding to the customer requests that we see. That is a little bit of execution in terms of just the incredible way that the teams have become deep with our customers, but it is also because of the confluence of tailwinds that we see before us today. Some of these I know that we know already, right? Agents. For the last two years, I think everybody has seen the swarm of agents coming, have been predicting it, and in some ways, we've overestimated it.
What we have underestimated is today there is an incredible call for governance, an incredible call to be able to orchestrate those agents in enterprise processes. What I do feel is something that is often left under the covers but is just emerging is the tokenomics of it all. I remember about a year ago, Daniel talking about why would you use an agent when deterministic automation can do the same job? Today, 30 times higher costs. You hear CFOs coming out and CEOs coming out about how budgets are being exceeded. That is a real call for the RPA portion of our platform, the deterministic pieces of our platform. The AI wave not only has a great pull for us in terms of activity of orchestration and agents, but also into the deterministic parts of our platform.
When you look at what's happening in the world around engineering and what we're able to produce and how fast applications are able to come through, the application spaghetti that went through enterprises, that is only going to proliferate. Not only does that have an orchestration impact for us, an opportunity to be on top of those systems and processes to bring order and efficiency, but it also allows us to do and really grow our application testing business that you see. When you look at our platform, there are multiple vectors of growth. The one thing that is interesting is that in some ways, the technological moat is becoming shallower with the pace of innovation that is there within the engineering realm. But what is very different is the need for depth and expertise.
When you look at expertise for us, whether that is in our people or whether that is in our product, UiPath has been investing in this for the last two years. You'll see that throughout this presentation, whether that's Cartographer or as we go through the go-to-market areas in which we were talking. When you look at all of this together, what is this activity that we're hearing about? What are customers asking? They actually aren't asking about agents. They're not asking about RPA. They're not asking about the technology itself. When they ask about what UiPath does, it's very simple to say that we solve the hardest problems for long, complex workflows that every enterprise is battling with. That is a battle that has been there since I was an intern at General Electric in a manufacturing plant.
It is the same problem that existed when we IPO'd, when we had a broad story around our platform, and it is the same story that exists today. What is interesting as you look through the pages here is you can see the breadth of what we're able to do, right? Daniel talked about the openness of our architecture, model agnosticism, the ease of use that we are able to go and put into enterprises. It means that we can conquer a broad set of problems. What is often missed is in the last two years, internally, we have been transforming ourselves to be super deep in specific domains and verticals. This slide, probably if you go back to our last Investor Day, it was just giving you a demographic of our ARR.
Today, verticals is the way that we are organizing, the way we are going to market, the way that we are approaching our customers, and the way that we are training our teams. I look. Eric Bouchard, who you are going to hear in the panel in just a little bit, in our customer panel, he has been in the financial services area.
I would not call him an automation expert per se, but he is super deep in financial services, right? Joe Taef runs manufacturing and our summit business, which is our broad industrial base. He is talking about procure to pay with CFOs. He is not talking about the core technology itself. That happens as you go further into the processes. This is not just about a technology area, right? Raghu talked about the technology components, Cartographer, getting deep into this, Atlas, very specific processes that we are going.
This is how we are now organized, how we are running, and how we are training and recruiting teams. Brandon Deer, who is our Go-To-Market COO, he owns enablement for us. When you look at enablement is as much about the breadth of our platform, but it is also beginning to drive depth across every function, sales engineering, sales, customer success, services, across the board. That depth translates to our customers. Where we are fortunate, sitting here at this moment, is all the work that was done by the leaders that preceded us in UiPath. We had an incredible period where we were able to penetrate and move into the deepest customers, and we are still winning them today, which I will talk about in a second. When you look at this, we have nine of the top 10 energy utilities.
When you look at financial services, 70% of the top institutions are on UiPath. That has a twofold effect. One, we get a ton of information as we are deepening our expertise by listening to our customers. When you look at Atlas, that is not our technology team in the back room or in the back office, so to speak, hands on keyboard alone. That is our FDEs deployed in our top accounts. That is our teams and our product managers having access to this customer base to hear. What it also means is we have an incredible starting point of credibility and a starting point of workflows that we can go and drive against. When you look at this, though, our strategy still has three components. We are still a land and expand business.
I think a lot of times when you look at companies, they have moved into this area of giving up on the land, but going through expand. If I am candid enough, the hardest question that we always get or the most frustrating question is around our customer count. We do not look at the total number of customers as the primary metric. We look at the quality of that customer base. When you look at the quality and you look at our financial services sector, through very intentional deep analytics and through making sure that we are extremely focused, we said federal credit unions above a certain asset value. Those are the type of logos that we want to go through. You will see that translate into Hitesh's section in terms of the economics for the company.
You can see the amount of increase that we are getting on new customers that are starting out with us at $100,000. And you're looking at that because their ASP, even if they don't reach that $100,000 right away, the average ASP around our new logos has doubled. I can't resist the CFO part of my side. That is efficiency, right? Go to market unit economics, the cost of acquisition can go down, and every dollar that you're putting in is returning more, both on a dollar value basis as well as a time basis. That is something that is not there sitting in our headquarters. That is something that is there in every vertical leader that they're thinking about. They're going after and targeting intentionally the ICP, the ideal customer profile. That' s where we're channeling our marketing dollars. That' s where we're channeling our people.
Expansion is still going to be the fuel that will drive the company. As we think about our ways to approach our customer base, we really started by acknowledging one really important fact, and that is our customer base in itself has expanded. When we first started, I think UiPath's roots were line of businesses. Over time, we invented the center of excellence, so to speak, within many customers during that era. So when you look at a certain period of a time, the center of excellence became our primary customer that we would sell through, that we would talk through. They're still huge importance to us. We are still well connected. When you go around and meet customers during this conference, you're going to find many COE leaders. Many of them have broadened their scope, including AI capabilities.
What you're also going to see is as we go into orchestration, we are now moving in and talking to AI architects. As we go into agentic and governance, our profile of who we are selling to is changing. As we're deepening with the vertical solutions that Raghu talked about, line of business buyers. We're back into talking to supply chain leaders. We're talking to procurement leaders in the office of the CFO. We're talking to chief risk officers, whether that's with WorkFusion or other products that we have. When you think about what that means for us, that is an entirely larger market and a new set of budgets that we can go after. As all things within software, you can go and say, "Okay, I know the customer," and you can unleash the team.
Daniel was very conscious to say, "What are the three motions that we want to invest in?" It gives us focus. It gives us prioritization of where our capital will flow, and it also informs us in terms of what are the types of skill sets we need to bring into the customer. So we have three motions that really are driving our growth that you're going to see some of the impact that that will have or that has had when you see Hitesh's section. Those are, one, orchestration and automation, really building on the foundational base that we have of customers and centers of excellences and moving up the stack. I'll go through that in a minute. The second is selling verticals, and the third is application testing, which you saw Ingo's incredible demo that he gave everybody here just a few minutes ago.
I want to start with the horizontal. Very specific in terms of a customer journey. You can land with a simple use case. That can have deterministic automation. What is very interesting for us is when you go to public sector in Europe, deterministic is in. They do not have high-risk appetites to go through. Landing with deterministic in many regulated industries is actually a structural advantage to us versus a disadvantage. It is an accelerant in the discussion versus a deterrent. When you land with that customer, our biggest thing is land with the right use case. You saw that with the ASPs and the quality of the new logos. Land the right customer, land the right use cases.
Then you move up to the stack to say, "How do we go into higher quality workflows that we have?" In some ways, moving from task to process, what Raghu talked about, that is embodied in that motion that has been there throughout for our customers. What is newer is continuing to now scale from the deterministic side to the other parts of our platform. If I have a workflow of 20 components, now you can go and say 10 of those are deterministic, five of them could be agentic. When you put it all together, it needs orchestration, and you start moving up to the stack. When I think about customer metrics, customers greater than $100,000, when you listen to our earnings, they are growing substantially. Customers greater than a million dollar, we continue to see double-digit growth across those categories of customer bases.
One of the reasons is because we are providing higher and higher value. It is where we are focusing on, moving from task to process, moving from process to impact across everything. In doing that, 90% of our million-dollar-plus customers are pulling multiple elements of that platform together. This motion is actually our second nature. It is very much into our DNA. When you look at this from what the impact is from a customer base, you can look at a key insurance company that we have within our portfolio. They landed and built a multimillion-dollar foundation that was built on deterministic automation. Claims, underwriting, these are things that we were already involved in. As they know our technology, as they know our team, now they are expanding into IDP. They are expanding into Maestro. What is also lost is their deterministic base is also expanding.
This chart is actual data, and what you can see, I would just point to two things that I think are super important. One, you can see the continued expansion, even on the deterministic side. The second point is, when you hit into that area where you could get into that enterprise architecture for orchestration and broaden the scope in terms of AI, it has twofold effect. One is you get a pretty good surge of growth just from pulling in the new products and new capabilities for the outcome that is there. What is lost is look at the curve for the deterministic automation. That is the tokenomics coming into play. Customers know us. They know when to use deterministic automation, and as you get into larger workflows, it pulls in multiple elements of our platform. The second piece of our motion is verticalization and driving vertical solutions.
I give a lot of credit to the Americas sales team around what they drove over the last two years and our international base with Alexandra and Matush, who are also here in the audience. We kind of said, "Look, for the last seven years, we know the processes that are driving the highest value in approaching it in a horizontal way." When you see seven of your banking customers going after loan origination, you can go and say, "How do I start selling loan origination right off of the bat?" What that does is it flips the curve in a different way. It packages it from multi-steps into a single step and a single outcome.
When you do that across the industries, and you infuse our company with good expertise, the right analytics, you can actually go and put this together in a way where you get multiple vectors of growth. Financial crimes. Organically and inorganically, we approach this, right? We are getting incredible response from our banks. Here we are selling to our chief risk officer. Healthcare RCM. Office of the CFO, which Raghu gave you the procure to pay example, and you are going to see that actually in the next slide. What does it bring for people? It brings them a fast time to value. So whether it is the components or a productized solution, they are able to plug it in and get ROI in a much faster way at a much lower TCO. The second piece is you are able to differentiate because we have invested in industry experts.
We have people who have lived their life, both on the product side and on the field side, invested in understanding what a healthcare provider has to go through, what problems are they solving. We are able to infuse that as we iterate with the customers. The third piece is it gives us flexibility on pricing. When you are pricing per widget, it is very different than when you price per outcome. When you are able to say, "You are going to save $100 million," it gives you a much better discussion when it is something tangible and it is something that you can implement within the next 90- 120 days, and you do not get shifted over to procurement. It is a real line of business leader discussion in which we are facing. So let us make it real. This is this year. Fortune 500 automotive company.
You can see they were a good deterministic automation company, $100,000-plus customer. They continued to scale. They expanded with their platform. Then what the team went in with is they said they actually met with the CFO, and they said, "Let us show what we can do around your toughest processes." Procure to pay, no matter how many vendors are out there, you do not hear efficiencies of really transforming procurement for major companies. Manufacturing, retail, that is 80%, 70% of what a finance operations sometimes has to contend with. We were able to show the solution that was built by our product and engineering team. The CFO, he did not have to be technical.
He didn't have to ask, "Can you tell me the difference between agentic and deterministic?" He went and saw that that crate set solves my problem and my need, and it allowed us to price and move that customer from a great customer in our $100,000 club to a million-dollar-plus customer in terms of where they are today. That motion is incredibly important to us. That's the second motion. The third I would call an adjacency, but it's really hard because it's not parallel play. It's synergistic play when you're talking to the CIO or when you're talking to the CLO, CEO, or CFO of a company. You're able to go and bring other value drivers for what we have.
With application testing, Ingo showed you the incredible. We are, to me, leaps and bounds ahead of what I would consider a lot of legacy applications within that space. When you sit down and say, "Here is modern technology to plug into a problem that has been there, where there is still a ton of manual testing or inefficient applications," we are landing application testing, focusing on specific problems. SAP S/4HANA migrations. We can go and land in that area. It develops a credible landing point for our application testing business. Then from there, we can expand meaningfully into other applications or other areas. Then lastly, it is re-energizing our GSI partnerships. GSIs, application testing, systems implementations, this is a massive issue that they are dealing with today.
As much discussion as there is with AI, we're forgetting how many companies are migrating from on-prem to cloud ERPs. Application testing has a massive tailwind for us to be able to go and take advantage of that. GSIs know that, and that brings into a scale play for us within GSIs. That is another vector of growth for our customer base. Here, I'm not going to go through the orange and blue, the deterministic side. Move them with AI products. AI products pull forward deterministic. That part we talked about earlier, but what you can see is there is another burst of expansion that comes from application testing. Now you're not competing against an automation vendor. We are solving multiple problems for our customer.
That is hard to compete with for one-on-one competition, and it deepens our differentiation, and it gives our sales team another avenue of expansion and our customers another avenue of value for where we are. All of that comes with incredible demands on delivery. I would say, as much as we have transformed our go-to-market, we are transforming our delivery centers as well. Forward deployed engineers, I know that's become a term in the industry. What I would say is we are authentically evolving our FDE model in two ways. One, bringing forth our services and our engineering teams to really make sure that the first wave of implementations go well. How many implementations have fallen short in the AI realm? Falling short on ROI, falling short on expectations of implementation timeline or delivery.
Our delivery method is improving every single day, and this is an area that we will continue to invest in. The second piece is our partner ecosystem. I remember three years ago, when you walked around, you can see and talk to our partners, and you felt a little less energy, you felt a little less connectivity. I am excited for you guys to meet our partners right now because they are a channel to market, both on the implementation side and within specific segments of where we have. Professional services is a differentiator for us because being a smaller company, that linkage and the ability to move with speed enables us for faster and faster implementations.
If I go back to those curves that we showed you in terms of when you put forth deterministic to AI to test, the faster that we can get those in production, it gives us the next set of bursts around expansion. We take delivery, and the focus on adoption has been a major focus for us for the last 12 months, and we will continue to double down into this area. That is our strategy, our three motions: expand horizontally, expand vertically, and the vertical depth with vertical depth that we have. Frankly, take the opportunity of what is a differentiated application testing platform and deepen our roots with our customers, and being able to expand the value that we provide our customers. How do we operationalize the strategy? This is a standard pyramid chart.
I know for everybody who is sitting in here, I am sure every single event that you go to, you are going to see this. The difference for us is in that strategic and core business. What you usually hear is large enterprise, commercial. For us, what we have done over the last 6-1 2 months is really look at our strategic accounts and say, "We are going to change the way we support and deliver them to be very specific, tailored." It is not about a ratio there. It is about the skill sets that you are needed. If you are a healthcare provider and you are going through a massive implementation with an orchestration, we probably need more people to help on the engineering side to ensure the proper governance, et cetera, that is moving into those areas.
If you are attacking a case within procurement, we want to make sure that the product team that is associated with it is really working through that area and supporting them in the right way. The top level of this pyramid is not about pyramids. It is not about ratios. The top part of this pyramid is driving value and expansion through really customized support.
The middle level for enterprise is where we are running programmatic plays across our customer base. When you listen to our sales leaders and you interact with them, that is where they are putting, and we talk about our ratios of deploying enterprise reps in terms of how we are approaching it. It is where we are going through the three motions in a programmatic way, aligning marketing, aligning delivery, so we can scale across the thousands and thousands of customers that exist in that space.
Commercial, it is all about ROI. Commercial is important to us, but how we support it, how we deliver it, channel support it with the right approach and go-to-market and the right digital support. We cannot physically be in all of the locations, so building our digital capability in this area is very critical for us. You can see that we have expanded the concept of segmentation around our customer base to the global footprint. Three points of excitement for me on this stage. The first is we are truly global. We have capabilities in Australia, in Northern Africa, in India, in Korea, in Europe, in Germany, going across, and obviously in the United States and Latin America as well. As we talk about operating leverage later, the foundation of where we are playing has been set.
What we're doing now very deliberately is saying, where do we want to channel from a prioritization standpoint, our investment? What are the largest markets to go after? So it's not about segmenting your customer alone, it is about segmenting the geographies. That has a specific area of making sure that we can channel the right amount of investment to win in the markets that we need to win in. Just the last two areas, pricing. Our pricing is always evolving in a certain way relative to the market, but we have a foundational set of principles. We price the platform, we price consumables. Our consumables are really around our orchestration and our agentic capabilities, and we price robots and users. All three of them have tangible value to our customers.
As you look at that really has protected us from this question around attrition of a user base. Because really, robots is the lion's share of that robots and users base, and our pricing is moving towards pricing the platform and the consumption of what we are selling to a customer. They have the ability to understand what they are buying from that perspective. Lastly, our partner ecosystem. It is vibrant. It will continue to improve. We are never satisfied with it, but you can see that we have kept one of the calls that has been there from our partner ecosystem is to be stable. This is a partner program that we launched two and a half years ago or two years ago. We have kept it very stable amongst our partners. What is great about it is we are also developing deeper and deeper partners.
Partners like Genzeon, healthcare providers, deepening with them. That is very important as we are moving and evolving our program, and testing is bringing a new suite of partners into our base, which is super exciting. We will continue to invest in technology and go-to-market partnerships. We are more focused, we are deeper, and we are more intentional in terms of where we are going. That element of focus is something that we are never satisfied with. If you sat in our meeting two days ago, you would say, "Hey, we want to be even more focused. We want to be even deeper with these partners." Going across go-to-market, what I hope that you will see as you listen to our customers is that customer centricity, the focus, and the energy around higher value processes and outcomes is what is really driving our company.
With that, I am super excited to bring to the stage Eric Bouchard and an incredible customer panel. I have sat in your seats. I have been on the outside of the software world. When a vendor is saying something, it means something. But when your customers are saying something, you know that that is true, and I am super appreciative for the time and the expertise and frankly, the quality and the caliber of customers that we are able to bring on the stage today. Eric, I turn it over to you.
Thanks, Ashim.
You okay?
Yeah.
I am looking at the—
Okay. All right. Well, first and foremost, thank you all for joining us today. I thought we could start out with just a round of introductions, if you don't mind introducing yourself, your organization, your role, and then we'll jump into some Q&A. Srini?
Good afternoon, everyone. Glad to be here in this panel. Srini Nanduri. I'm the Vice President and Head of Data and AI at PPL. We are an energy organization, a 100-year-old organization, and really looking at a transformation of AI across the board and how do we build the utility of the future. If you look at it, my team really is looking at AI and automation, how do we help our customer, our field, our grid operations, and really helping our employees having those AI tools so we can serve our customers and really serve the community that we serve best.
Jairo?
Jairo Quiros. I'm SVP, Global Business Services. I'm also responsible to drive automation and AI across the enterprise, leading the center of excellence for automation. In my global role, this is what we do day in, day out.
Hi, everyone. I'm Pramod Dibble. I lead AI and automation for USAA Bank. USAA, you're probably familiar with them, have an insurance company, property and casualty, a life insurance company, and a bank. As we start to think about how we deploy AI and automation across all three of those lines of business, that's really where my role gets involved. With a heavy focus on anti-financial crimes, that would be anti-money laundering, bank fraud, customer disputes, and escalated complaints.
Hey, good afternoon. Glad to be here. I'm Paru Puttanna with Voya Financial. I'm a Senior Vice President, heading the architecture as well as the agentic AI delivery. As you guys know, Voya Financial, we're a Fortune 500 in retirement employee benefits and investment management. My focus is going to be across the board, helping the organization both in the architecture space as well as having the agentic AI and automation delivered.
Thank you, all. I thought maybe we'd start out talking a little bit about the journey, your automation journey over the years, and more specifically, how that's changed in the past two years. Jairo, maybe I'll start with you and the journey Equifax has been on from its inception, but with a focus on really what's changed in the last couple of years.
Super. I think at Equifax, we've been committed to improve the way we work. 10 years ago, actually in 2016 when UiPath was ramping up, we also made partners with you guys. Since then, we've deployed hundreds of automations across the enterprise. Our role is global, so when you think about the breadth of the type of work that we've done with you guys, going from deterministic automation to integrated GenAI to adoption of intelligence automation, and all the flavors that we discuss here today. When we think about agentic AI, for this past year, I think the impressive aspect of how the technology has evolved and the relationship also has evolved as a customer of UiPath is that we strategically have partnered, to provide our feedback to help build our requirements within the product.
I think I speak in the name of Equifax, but I think it's a common sense that you guys have been open to integrate the new features into the product. In this case, particularly over the past couple of years, AI has been an inflection point for us, of course, as many companies. Our focus has been not to really prove the technology works. It's actually thinking about how do we integrate a new operating model, how do we redesign the processes now that we have the technology available in order to do so. Utilizing capabilities such as Maestro and the orchestration has been key to us for the last year or so as we're designing the future around automation at Equifax. Yeah.
I appreciate that, Jairo. Paru, I thought maybe we'd ask you a similar question. Obviously, past couple years we've been very engaged. Maybe talk a little bit about that journey, where Voya was maybe three years ago, how that's changed in the past couple of years, and the things that we're accomplishing today.
Yeah, absolutely. I think like everyone else, I think they've jumped on the GenAI and agentic AI journey in the past few years following the industry trends. We also learned from some of the initial GenAI experiences, build versus buy, what is the best way for us to move forward. It became very natural of using some of the UiPath capabilities, to name a few. We are heavily using the platform, using the Communications Mining, Maestro for orchestration, IDP from document processing perspective, and definitely the unattended and serverless robots. Then recently we've also started to use the Coded Apps. I think we are touching all the capabilities of the latest and greatest from UiPath.
It's been a great journey from the past couple of years and where we are headed, and we are using UiPath's capabilities for our strategic use cases and implementations.
Paru, maybe staying with you. You're using the whole platform. What are some of the use cases that you're most excited about? How have you organized some of those things? To the extent you can touch on it, what sort of value are you getting with some of these workloads in production?
Yeah. We view AI as not necessarily a technology initiative. We are using it as a catalyst to really transform or reimagining how we serve our customers and our partners and also to have a cultural shift for our employees, equip our employees to do the better work. From the value perspective, some of the use cases I'll say, I'll come to the value next. We have multiple use cases which are in production, whether it's the claims processing, analyzing the fraud, and also improving some of the business processes automation. From the value perspective, I would say, I think the value comes in, the way I look at it's a multiplier of you look at the model capability, multiplied by your workflow, multiplied by trust and multiplied by governance. It's an exponential factor of the human ingenuity.
I think if any one of those factors are zero, you're not going to achieve the business value. The way we look at it, how can we reduce the business processes or reducing the time, and we'll also improve from the accuracy and serving our quality of the customers and keeping the customer experience and customer obsession in mind, then automatically your business value is, you will gain the business value. Yes, I think from the finance perspective, you do have to show from the ROI. I think these are the factors that we are including, when we are providing the business value.
Srini, maybe over to you. I know we've obviously spent a lot of time with Paru and some of our forward deployment engineers and insurance experts. We've been talking a little bit about some of the opportunities within utilities and specifically needing cash. Where are you most focused? Where are you guys getting value? What are you most excited about?
Yeah, I think as most of us on the panel, we started our journey with task automation. We looked at employees and said, "What are the tasks that you're working today? Can I automate those tasks?" We soon found out that the business doesn't work in that way. It's a business process that you need to look at from a transformation perspective. We stepped back, we worked with our business and really defined what does the end-to-end meter-to-cash process looks like. This is when you set up a new meter, the billing system that needs to be onboarded, and then you finally get the bill, right? One of the things we found out as part of the billing process is, it's called the high-low issue.
Or we build an agentic solution on high-low, where you get a bill at home, you look at the bill and go, "Why is my bill high compared to last month?" You don't know why. That's a simple question that you should be able to answer. But as a company which has a legacy around so many systems, we are still in the process of transforming some of the legacy systems to the cloud. We found out that that is a pretty complex question for us to answer, where we need to take a question, we need to look at multiple systems, we need to go manually find out why the bill is high by touching the billing system, by accounting system, and some of those areas. There's a lot of that steps that needs to be followed to actually make that happen.
And this is where, one of the great examples we have, we automated the whole thing. We call it the agentic solution around high-low bill analysis, where we have agents actually look at the billing processes. The RPA bot actually go pull that information, provide that information to the agents, and Maestro is really kind of orchestrating that end-to-end process. Looking at why is the bill high? Is the weather related issue, is it something else? And really can provide a much more detailed information back rather than a simple automation that you would have done. And this also gives an example of how do you look at that end-to-end business process that you can really transform rather than those individual tasks. And now the process becomes really intelligent, and then you can fully automate the process as we go through.
This is one of the great examples of looking at the end-to-end business process, identifying those tasks, and really automating some of those steps as part of the meter-to-cash process, where the whole process becomes intelligent.
And Srini, maybe continuing on that vein, I think you said that you guys had mapped 50+ of these processes within the organization. Are you looking at these and at what point are you saying these aspects can be deterministic, this is where we need agents?
Yeah.
How are you making those decisions as you go through? Then, how is that resulting in a build and maybe your use of Maestro orchestration to stitch that together?
Yeah, I think that's a great question. I think the way we have. We normally start off by not looking at whether that's an agent or an automation bot. We actually take a step back and say, what's the business process that you're trying to automate? Is it something where you need to bring some reasoning into the mix? That's where the agents are really good at. If it has to go as a regulated organization, we need to make sure there are certain aspects which are highly deterministic. When we are pulling information from a billing system, when we are adding details back to the billing system. So there's a lot of SOX and other compliance that we need to go through.
We have to look at it as an end-to-end business process, identify those different tasks that you need to follow as part of the process. Some could be agents, some could be deterministic RPA automation BOTs, and some could be human judgment too. At some point of the billing exception, the agent or RPA is not able to solve the problem, the Maestro kicks off that task to a human. So the human actually goes into a workflow, figures out what happened. So there's a human judgment as part of it. This is where really, I think the future of automation would be a combination of those three, as was said before in the previous presentation. It's really around orchestration, around humans with agents, with RPA BOTs.
You need to bring all the three things together, and based on that, you can figure out which is the right process for the right task.
So maybe staying in compliance and regulated industries, Pramod. You are using some of our financial crimes and compliance vertical solutions today. Talk a little bit about how you guys are actually using that at USAA and any ROI or value that you are able to share.
Sure. I am going to avoid getting my hand slapped by the lawyers, so I am going to stay a little vague about figures here. Let us just say that we are comfortably in the black on the engagement. We started as a WorkFusion customer, and of course, with the acquisition, now we are part of the broader UiPath family.
There are a couple of processes that if you are a bank, you must do. If you do not spend your Friday evenings with a glass of wine and the Patriot Act, or maybe Bank Secrecy Act, then you might not know that you are required to file a suspicious activity report, which is called a SAR. A SAR is, you have observed something in a customer's behavior that you need to then refer to the federal government . This is one of the highest risk compliance processes that exists in banking.
99% is not good enough. It has to be 100%. If you missed a few, that means you are going to get a consent order. Part of a consent order means you cannot sell checking accounts, and that is bad. We try to avoid that as much as possible. We are leveraging WorkFusion's solution to perform some of the research that goes into compiling the packet that then goes to FinCEN, over to the federal government. In the before times, this process took about a day or a day and a half to do one. We do a lot of these because we are a big bank. That is one space where we are leveraging this. It is a combination of a variety of different modules. One of the advantages of using a vertically integrated solution is, I could go out and build this, probably.
The issue is, it is a complex multi-step process, and there are different technology modalities that are baked into each one of those sub-processes. What is going to happen here in reality is that my business is going to have an idea, and we are going to talk about it for a year, and then we are going to go build code about it for a year, and then we are going to validate it for six months. The problem that you have solved two and a half years later does not even exist anymore, not really. That is where the vertically integrated solutions really speed up your time to value. Secondary process where we are using this is, you are required to review all of your high-risk customers once a year.
You risk rank your customers based on what you think they're about to do, and then every year you look at them to make sure that they didn't do those things. That process takes about half a day. Again, in the before times, now substantially less than that.
Jairo, let me come back to you. You have a lot of big technology partners and vendors. Why UiPath? Where you're using us, why us versus others? What capabilities? What's driving you guys to make those decisions?
I think when we think about technology, I think we don't start there. We think about the problem that we're trying to solve. And thinking about, what my peer customers here said, our focus is internal. But it's also business-led, meaning that when you think about the big data issues in the world, the fact that we're a data company, the fact that when we think about how do we manage data, it's got to be secured, it's got to be well-governed. I think when we started our journey, it started because of the same reason we continued to pick you guys, and that is that UiPath provides a very reliable, secured governance around.
Even with the Maestro and the orchestration that we're talking about here, I think that's one thing that really stands out, is the fact that you have not only a platform, but you have a way to change your operating model internally in the company. So when you think about what you guys are providing to companies, making sure that we're teaching agents, teaching deterministic automation, teaching human in the loop, that continues to prevail as an advantage. And you mentioned multi-technology. Yeah, we are a huge global company, and we will leverage as much technology as we can being a technology company. But I think the fact that the way you have designed your product, which is open, and the way that we can integrate with other technologies and tools that we use allows us for us to be way more effective.
Now, the differentiation is, in my opinion, there is no other company that provides that orchestration capability that is also governed and secured in the way that we have the observability, the traceability that we need in order to operate in a regulated environment. I think we all have that in common, which is we got to respond to somebody else as far as the use and the proper use of the data that we have. That is one of the elements. The other element to us is that you guys are also investing in making sure that we are way more effective, meaning from a center of excellence perspective in terms of the resources and the time that we have in order to execute this automation. So time to market is very important.
Whether that is an internal finance use case or maybe an external product data problem that we manage. One example is that, through these capabilities that you have offered, we have been able to build automations very quickly. So we go from six months to maybe six weeks deployments. We take each data problems that we used to have, and then we create a solution in order to not only reduce the time that it takes but improve the quality. One example is that we are extracting data from multiple sources that are publicly available in different regions of the world where we didn't have the scale, we didn't have the technology. So it is a highly manual process, and it is also demanding very high quality standard. So with these new solutions that we've deployed, we are now extracting more than 400 data points and very huge unstructured documents.
We are now closing the gap as far as the time that it takes for us to build, package that, build it as a product, sell it out in the marketplace. So huge advantage to us, a lot of the capabilities that you have built.
Hey, Paru. Coming back to you. Similar question, right? A lot of options, a lot of ability to build versus buy.
Yep.
How are you making that decision, why UiPath in specific areas within claims fraud? I would be curious to understand.
Yeah, absolutely. Completely agree with what Jairo was talking about. I would say with UiPath, the journey started almost like 9- 10 years ago with the RPA and automation. Having that pretty strong foundation, when we started to look at the mention, all the capabilities of the platform that we are leveraging, I would say I think it made sense to start when we learned from initial learnings also from the GenAI and Agentic. So said, we want to take advantage of the Maestro orchestration capabilities and how we can connect to multiple systems because we got to connect with our internal systems as well as some of the external SaaS vendors for us to be able to make the decisions from Agentic and then hand over to the automation.
By looking at all of those areas and we said, I think this is where UiPath made more sense for us to be using for some of our use cases. As we are expanding, I think we are seeing more and more opportunities to leverage. I also want to call out that I think that from past couple of years, it has been the strategic and strong partnership and collaboration between the two companies. I think that is the key differentiating factor, because I think without that, I do not think we will be successful in implementing the use cases, and we do need that partnership and collaboration to work these use cases to be successful.
Yeah. I could not agree more. Pramod, you touched on this a little bit around the ability to build versus buy, right? Obviously, with what we are doing with financial crimes, prepackaged solution, maybe go a little bit further. When you think about the things that are happening around some of the vertical solutions as you guys look into the future, how are you thinking about when to build versus when to buy?
Yeah, thanks. The technology has changed a lot in the last two years. I think that's the worst kept secret in the room. We will have both approaches in our strategy going forward. We will build, and we will also buy. We will build when we need a point solution that's relatively small in scope with strong guardrails around it. This is how you control for token cost, this is how you control for hallucination, and this is also how you deploy small, relatively compartmentalized automations to solve critical business problems. When you think about buying, it's really about how are you going to be leveraging a suite of technology that is more complex than you can realistically build within a commercially feasible timeframe.
That's where the solutions that we've deployed with WorkFusion, and now with UiPath, were able to help us solve some of the more critical problems. Because they don't just include data compilation or machine learning or AI solutions. They need to be able to control for all of those factors and then orchestrate across, and some of my peers have talked about orchestration. It's such an important aspect of what we're talking about. That and data accessibility. Data is always the long pole in the tent for any automation program. So that's how I think of build versus buy. Where I'm really excited to see is how the frontier models continue to evolve over the coming years, to see how that balance then recalibrates as the quality of the technology that's widely available changes.
Well, these are my questions. I'd like to throw it out to the audience and see if you guys had any questions that you wanted to ask the customers up here. I think we got a mic coming over to you guys.
Thank you. I appreciate all of you sharing your experience at this point. I would love to hear, I know you guys have been customer for UiPath for a long time, how has your deterministic task automation flow been changing since when you adopted it? You guys talked about AI investment. What are the areas of AI investment that you are right now doing, and how is that being funded?
Big, broad questions. Do you want to—
Yes. Go for it.
Maybe I will repeat your question. I think your first question is, I think, how is the deterministic automation changing? I think the third one I heard about the investment in AI. What was your second question?
What are the areas of investment, and what kind of investment you are doing, and how is that being funded?
Okay. The deterministic, I would say, definitely there are multiples of these automations in place for several years because the automation, RPA, is in this space for a long time for both financial and manufacturing, all of these industries. I think how we are looking at this AI coming in is how can we bridge or combine both together? Because the way that we are leveraging is you are using for some of the AI for your intelligent decision or your reasoning, all of those things that are GenAI and Agentic, and then bringing your deterministic for execution of the flow, as well as I think connecting to different applications. How am I bringing those automations? I think that's where there is definitely a lot of opportunities, and I think we are also going back to some of our deterministic that's already in place.
How can we really take a look at those from the AI mindset? From the investment perspective, all of the past few years, it's been more of looking at, like, I think I heard Jairo say, but we're looking at the smaller automations or where there is a point automations. Now we're looking at more from the transformation perspective. I think Srini mentioned this as well. How do you look at the big picture? Unless if you do not do those transformations, you will not have those business values. From the funding perspective, that's how we are looking at how can the transformation programs that we can bring in, where we can include the AI as part of those transformations, whether it could be a customer experience that you are changing. You are reimagining how you are serving your customers.
That's how we are looking at funding for those AI initiatives.
Any other panelists? I'd be curious.
The one thing I would add there is deterministic, which is building the bot has always existed, right? You take a specific task, and you automate it. What we are also finding is sometimes the rules are pretty complex for you to make a bot or make an automation, and that's where you actually are increasing the surface of how many things you can automate using Agentic AI, right? Like with agents actually combining with bots. We're starting to see the trend where if you look at the architecture of Agentic AI, the bots with RPA bots, which is going to be doing a lot of deterministic flows, which are pretty straightforward. But you can actually expand that surface territory and bring in some agents that can combine with it, and then you bring humans in the conversation and really do the orchestration end-to-end.
And that's where you can really automate the end-to-end processes, right? Which was pretty expensive in the past.
And if any of you are willing to share, if not, you don't have to. The funding, is it coming centrally? Is it coming from line of business? Is it a combination of the two? I think that's likely where the question was coming from.
A couple up there.
Yeah, please.
I think you asked about the evolution, right? When you think about the evolution, and my colleagues here, they talk about the technology and UiPath as a product and the evolution. We started with GUI animations, with RPA. Now, it's like a whole ecosystem, and the deterministic integrates very well with the APIs, with the agents and whatnot. I think the important aspect there also is the way they have envisioned the product, which is when you think about agents and you think about bring your own model approach. For instance, just one example, allows companies like me to integrate with, for instance, GCP for GenAI, right? You have the agent, you have your own LLM, and then you orchestrate the entire thing. When we started this journey, it was more of self-funding, right? Fighting for processes that you could automate and then self-fund.
I think today is way more strategic. It's not like point investments. It's more central when it comes to AI. I think many companies were in that path. In my reality, I have to do both, right? I got to be able to demonstrate that whatever central fund resources I have, they deliver the business outcome back to the units they are serving. Yep.
Yeah. I wanted to ask, you guys are in regulated industries, and I guess, how much was governance, compliance, AI safety, a decision-making factor in investing in UiPath versus other functionality, technical reasons, et cetera? How important was the governance and compliance angle?
I'll speak to that one. Extremely important. It's not a secret that USAA had an anti-money laundering consent order in 2022, and it currently has an open bank consent order. So extraordinarily important. I want to build on something my colleague just said here, where I think what you will see across the industry is more centers of excellence being funded centrally. That doesn't just solve for the question of where the dollars are coming from. It also solves for the governance piece of this. Because at one extreme, you can imagine that every data scientist and developer has Cloude Code available to them, and they can ship products on their own, and that would cause chaos. On the other hand, you can imagine a singular swim lane where all AI projects go through it, and that's too much.
It's going to be somewhere in the middle. Like my colleague here, we will have some centralized funding, but we need to justify the investment from a return perspective.
Well, I know we're at time here. Again, thank you all, not just for this, but for the continued partnership. I'm looking forward to what comes. Thank you.
Absolutely. Thank you.
Thanks.
Back to you, Ashim. Yeah, thanks, buddy. Hitesh, if you wouldn't mind coming up for the fun part.
Thanks so much, Eric. All right, what an incredible story. I'm very, very pleased when I hear our customers talk about their journey, especially how they're deriving their value and more of the scale, starting with task automation and in a sense, adopting the entire platform so they can solve the complex problems that they wanted to solve for their companies. So very impressive. Good afternoon, everyone. I know I have met many of you over the last two years. Really thankful to Ashim for bringing me along in the last couple of years. I'm really looking forward to spending more time with each one of you at our investors' reception as well. Over the next 30 minutes, I really want to spend time on three things.
I want to really highlight all the hard work that I've seen in the company over the last five years in terms of how we have transformed our top line. Focus on some of the growth levers both Ashim highlighted, Raghu highlighted in terms of platform, Ashim highlighted in terms of go-to-market motion. The opportunities that we have in front of us on how we can continue to grow the business. Lastly, also talk about from an operating efficiency standpoint, how can we continue to improve our operating margins, improve our free cash flow margins, and at the same time have the discipline of returning the cash back to our shareholders. Let's get started with the most important thing, is our customer base. Over the last five years, we have grown our customer base.
If you look at it at the time of our IPO, we were at $925 million, which is now grown on an estimated basis by the end of 2027, to be more than $2 billion. What that really tells us is our customers grow with us. As you heard throughout the day today, as customers move from task automation to process transformation, it provides a significant opportunity to continue to expand with them. One most important fact here is we don't have concentration in one particular industry or one particular geography. If you look at our entire customer base, it comes across every single industry vertical that's out there. Not only that, you heard from the panel right before me, is especially in those industries where the customers are running highly complex processes, regulated enterprises.
Those are the industries where I think Raghu also talked about our vertical products, our vertical use cases. These are the ones which is resonating extremely well with our customers. We have more than 50% of our customer base that is in this industry. Banking and financial services, manufacturing, public sector, again, high volume of human manual work, and healthcare. Each of these industry verticals where the customers are trying to solve those high-end complex problems, long-running processes, we are seeing a higher CAGR in each of those industries. Not only that, even apart from those other industries, we continue to see expansion through some of our vertical solutions such as Office of the CFO and others. In addition to this, as we continue to double-click on our customer base, the cohort that is really meaningful for us is our customer base where our ARR is more than $100,000.
We have more than 2,600 customers in these cohorts. If you see, as these customers continue to expand, it meaningfully contributes to our ARR base. Especially our customers more than million dollars, more than $5 million, they represent a significant portion of our overall ARR base. Our customers greater than $100,000, that is a white space. These are the customers where we have an opportunity to continue to push our entire business orchestration and automation platform, which will allow us an opportunity to continue to grow further. Not only that, when you further double-click and look at our top 25 customers, our top 100 customers, this is an interesting fact pattern here, is between 2020- 2027, our average ARR for the top 25 customers has gone up from $2 million- $10 million, 5x .
Our top 100 customers, during the same duration, has gone up 6x . This is a small subset of the 2,600 customers that I highlighted. I can just extrapolate the opportunity that we have front of us, in terms of continuing to expand the rest of the 2,500 customers. This is our land and expand model. Ashim highlighted on this. This is a very important one for us. We have a very good land and expand model. Our platform provides us multiple opportunities to go and land now. What you've seen is, in the past, we used to land with RPA. Now we have ability to land with our vertical solutions, vertical use cases. We have opportunity to land with testing. Our buyers are different within the customer. In the past, we used to just go with COE.
You saw in Raghu's slide, now we are able to go with the AI architects. We are able to go with the line of businesses. Opportunities to land has expanded substantially, and then once we land, our ability to go and expand the entire platform and help our customers solve their problems has also grown up significantly. One of the important metrics is when you look at our top 25 customers, the lifetime value, they have grown 62 x from the point we landed. Now imagine the opportunity we have, especially as the platform continues to expand in some of the go-to-market motions that Ashim highlighted. That really helps us continue to expand our relationship with our customers. This is what it was in the past. Ashim did highlight on multiple vectors of growth for us. There are three things that he spoke about.
The first one is taking our existing install base, especially the customers who were primarily adopting RPA, and moving them up to the entire platform, as we call it as Business Orchestration and Automation platform. The way we define this is our entire customer base that has not only adopted RPA, but the customers who have also invested in a meaningful fashion in our AI products. Once we move the customers to the business orchestration and automation, we are seeing stronger durability, and I will cover some of those metrics in more detail. In addition to this, one of the other drivers that Ashim highlighted and also Raghu showed this morning is several industry vertical use cases, industry solutions. These are resonating extremely well with our customers. This allows us to land the greenfield opportunities. It provides an accelerated time to value.
We just heard from a couple of customers here on especially USAA, how it allows them to not only put the technology in production, but also start deriving value. So that is the second area on how we are focused on in driving our growth. In addition to this, Raghu highlighted earlier today, what you heard is as customers focus more on build, it provides an incremental opportunity for us because the customers need to test more. So there is a natural opportunity in front of us with our Dark Testing Factory, as Ingo highlighted today, that we have a significant opportunity where we can continue to expand further. Every time when we move our customer from RPA to our entire orchestration and automation platform, where we are able to help them solve their process transformation problems, it actually helps us drive our ARR growth.
This has helped us grow our ARR base by more than 50%. At the end of Q2, our ARR from our AI ARR was more than $250 million. Not only that, this is also helping us improve our dollar-based net revenue retention. So when you combine the ARR growth with our disciplined execution and our ability to solve more complex and long-running processes for our customers, we are clearly seeing it is helping us with our dollar-based net revenue retention. If you further double-click on our customer cohort, where the ARR base is more than $100,000, we are already seeing a higher DBNRR of 114%. Not only that, when you look at that cohort, and if you further peel the onion and look at the customers that have already moved from RPA to both, we are seeing that DBNRR to be even much higher at 120%.
So what it is doing for us, it is doing actually three things. One, it is improving the overall durability. Two, it is further providing us an opportunity to land much larger sizes with this customer. And three, it is pulling deterministic. You heard a couple of our customers, including Srini from PPL. This is exactly what is resonating with our customers, is every time we move them to a platform where they are able to go and agentify an entire process, it is providing significant value. For us, this is also helping us from a stronger durability point of view. In addition to this, Ashim highlighted something really interesting, is we are really focused on our land and expand, and we are going after new logos.
Ashim also gave a bullet point there saying, not only that we are landing new logos, especially when you go to our customers with the entire platform, it is allowing us to expand the ASP by 4x . This is what we've continued to see is when now, not only we are focused for new logos, but we are also going after them with a platform so they can start deriving much more higher value. Here's an interesting example. This one resonates with me extremely well because in my previous role, when I was an auditor, I had an opportunity to go and audit chargebacks process. This is one of our Fortune 500 customers who came to us initially as an RPA customer. They landed with deterministic automation.
They wanted to solve their accounts payable problem, and then they further expanded with IDP to solve the invoice matching within the chargebacks process, and then they clearly realized with the power of the platform, they can go and transform the entire chargebacks process. It's a very complicated process within the pharmaceutical industry. Not only that, we were able to help the customer solve their problem with the help of our FDEs. It allowed us to expand our ARR base by 24x . As Ashim highlighted, what is more interesting here is, it's actually pulling the deterministic automation forwards along with the ability to agentify the entire process for the customers.
Not only that, the beautiful part about this is we are going and solving an entire process, which becomes a use case within the industry that we can go and bring to other customers within the same industry. So while we are focused on transforming our top line and we talked about how we have multiple vectors of growth, we are at the same time focused on transforming our bottom line. What we've done really over the last three years is improve our operating margin by more than 1,500 basis points, 3 x. We have done this while doing two things. We've continued to invest in our platform, which is very visible based on what you see. At the same time, we are investing in our go-to-market, with our direct sellers.
This all starts with best-in-class gross margin, and one of the key levers there is our hybrid platform. Our ability to deploy our technology, both on-prem and cloud, provides us an opportunity to drive higher margins, and which further helps us continue to be more efficient from an operating expense standpoint. If you look at our expense profile for R&D, we've continued to invest in our platform expansion. We have leveraged the power of AI to continue to transform ourselves. If you look at our sales and marketing, continuing to invest in sales and marketing globally while remaining very efficient. Not only that, we are strong believers of UiPath on UiPath, us as customer zero. Daniel has tasked the entire company to make sure that we are transforming ourselves on the back of automation. Fortunately, at UiPath, automation mindset is within our DNA. It's from the get-go.
Now, with our ability to go and transform the entire process, we are further expanding on that and doubling down, which is providing us incremental opportunities to continue to improve our margins. Some of the additional levers that we have on how we can continue to further transform our global footprint. We have our sellers in each of the geographies where our customers are. Our customer centricity is very important to us. At the same time, we have operational efficiencies based on where our operations are. We are really focused on taking advantage of that to continue to further bring incremental efficiencies within our operating expense profile. I mentioned earlier about customer zero. As I mentioned, we have hundreds of automations already in production. Not only that, beginning with last year, we also ran an internal agentification program.
We have almost 80+ agents in production right now across the company, and it is not just in finance and accounting. It is across the board. When you look at sales and marketing, R&D, G&A. This actually allows us to become more efficient. At the same time, it provides us insights and great assets that we can bring and share with our customers as well. While we were focused on non-GAAP operating margins, we are equally focused on GAAP profitability. We made a structural change. We became more disciplined when it came to our stock-based compensation. When you look at our stock-based compensation expense ratio as a percentage of revenue, we brought it down from 58% at the end of fiscal 2022 to 12% for the first half of FY 2027. Again, this is done with a structural conscious discipline and structural change that we made within the company.
One of the core factors that has helped us contribute to our GAAP profitability. We have demonstrated this over the last four quarters that we are GAAP profitable. We have guided to GAAP profitability for the rest of the fiscal year. One of the core reasons is because of our disciplined execution around stock-based compensation. We are doing this while still continuing to invest in talent and retaining the talent. Our operating margin and stock-based compensation behavior has also helped us continue to remain focused on free cash flow margins. Over the last four years, for instance, FY 2024, we have continued to improve our free cash flow margin. This has allowed us to make sure that we continue to invest, at the same time, helped us with our cash position in the balance sheet. The interesting part is, we have done this while executing on our buyback strategy.
We have already completed $1.1 billion in buyback. We still have another more than $400 million in authorized buyback from our board, and we continue to execute on a buyback strategy, at the same time, making the investment in platform expansion and our go-to-market, as I mentioned earlier. While doing all these things, we are also focused on keeping our dilution to less than 0%, so it becomes accretive to our EPS. While this is all there, of course, we want to make sure that we are focused on our long-term model. As you see, we have significant activity in the business, and we feel very positive about the guidance that I provided during our Q2 earnings call. While that is there, we are really focused on Investors Day is all about long-term margins, so we want to make sure a long-term guidance.
We want to make sure that we are spending time there. If you look at where the company was at the time of IPO, we had negative 4% in operating margin. We were committed to transform that. We wanted to make sure that we are converting ourselves into a positive operating margin company, and we were really focused on GAAP profitability. Based on our results that I shared earlier, we have already achieved GAAP profitability, and that momentum is going to continue. All the cost efficiencies and the drivers that I highlighted earlier will help us continue to make sure that we are focused on efficiencies, at the same time, continue to work on our targeted operating margin of 30 plus percentage.
At the same time, we are continuing to focus on our stock-based compensation and make sure that we are going to bring that within the range of 8%-10% of our revenue. While doing all these things, we are really focused on capital allocation strategy. Our framework is relatively simple. Our number one priority is to make organic investment within the business. When it comes to platform expansion, that is number one priority. Of course, we are doing it in a disciplined fashion with using the power of AI. At the same time, we are focused on customer acquisition and expansion, both. While doing it, we remain opportunists in terms of our M&A strategy. You have seen WorkFusion in Works. You have seen Peak.
We continue to remain opportunists in terms of acquisitions that allow us to continue to expand our platform, at the same time, focused on bringing in the right talent that can help us expand that agenda forward. In addition to this, we also remain opportunists to continue to keep an eye on the dilution percentage in terms of buyback and returning the capital back to our investors in form of buybacks. With that, as I mentioned earlier, we have definitely transformed the company over the last five years, expanded the customer base. It is a really large, durable, recurring customer base. We have also done it at the same time, improving our non-GAAP operating margins and really converting the company to be a GAAP profitable company.
I also focused on some of the multiple growth vectors, including what Ashim highlighted, which provides a significant opportunity to continue to drive the company forward. In addition to this, we are focused on our operating leverage and also strong cash generation. All right. With that, we will turn it over to Allise. We are going for Q&As now? All right. Thank you. Allise, do you want everyone from the leadership?
Yes.
All right. Just give us a minute. Okay. All right.
Hi, Sanjit Singh from Morgan Stanley. Daniel and team, I thought this was the most well-articulated, well-targeted, most focused presentation about the opportunity since you guys talked about going on an act two with Agentic automation. I've been trying to figure out the ways to ask this question, but essentially, it's like, if you're successful with the BOAT strategy and all the different motions that you guys are executing on, essentially, do we just have accelerating growth over the next couple of years? What I'm implying here is that, is there a part of the business, do we have to go through an RPA transition for the business to accelerate? Or just given the customer testimonials that you're seeing that as BOAT adoption increases, so does the deterministic automation adoption also increases, so we really don't have a headwind to overcome.
I'd just love your thoughts on if you guys are successful, and if you guys execute, does that just mean faster ARR growth over the next couple of years?
Yeah, I think this is a fair expectation, and this is why we are all here, in all fairness. We want to build a growth company. We all believe we have a tremendous opportunity in front of us. Regarding your question about RPA moving into BOAT, I think this is a trend that shows, first of all, that RPA is an important technology in your automation toolbox, and it's here to stay. The numbers that we put on the screen shows that it creates BOATs, including the RPA. It's a more stickier technology. This is a technology that is strategic into the enterprise tech. Some people can say that RPA alone being more of a task-based quick fix until you transform the systems. But this is actually about transformation. Many of our clients and partners are using basically BOATs foundation to transform their processes.
That makes it stickier and durable.
I think just one thing I would add is, I think people think about RPA as a negative. I think what we hopefully articulated throughout the presentation is this is a positive. When you see customers adopting our platform, you see them accelerating with deterministic automation as well. Sanjit and I think we've talked about saying RPA is a durable category. Our only piece is saying, when you add BOAT around it is durable and has a ton of market in front of us. It is both.
I can add one more thing. I think I spoke in my presentation about 100-plus customers who have Maestro use cases in production. Majority of them have tasks implemented as RPA. The reason for that is, complex business processes run a mixed state of applications, some built years ago, some more modern. So our ability to be able to have tasks that use RPA as ways of interrogating and gaining information and being that system of action layer alongside agents is a very powerful capability that we offer alongside the orchestration layer. Do you need RPA to build a UiPath? No. But the reality in the enterprise is there is mixed state. Some applications do tend to be old and dated, and for that, RPA still plays a significant role.
It's Michael Turrin, Wells Fargo Securities. Thank you for hosting. Appreciate all the content. I'm looking forward to the event this week. You've been admittedly talking about orchestration for longer, but we're hearing a number of different software vendors start to talk more about the importance of orchestration and their position within the harness layer. I'm just curious, your perspective on the points you're hoping land with customers this week, because they're also kind of getting a lot of this new press release, type of announcement. What are the key points you'd emphasize?
I know there's a core piece of the RPA foundational technology, maybe your agnostic position. I'd be curious to hear more on that as well as do you expect this is an either/or discussion for customers, meaning there's one orchestration platform that they choose for all of their agentic use cases, or how you think this evolves from here?
Yeah, let me start. To me, I think the main point that we want to make this week is that, in order to deploy AI in an enterprise, in the context of enterprise processes, you need to put in place this map and rails that we call. You need to have a Map of Work, because otherwise AI cannot act, cannot really understand your enterprise, and you need to have the rails that creates boundaries for AI, because you cannot let it run astray. You need to create what we call controlled agencies. That's one thing. Second thing that I want to make a clear point is coding agents introduce a very interesting asymmetry into how you automate processes. Building AI agents into the context of enterprise processes, it's still difficult. AI has to graduate to trust it.
It has to gather a lot of evidence in production in order to give them increased capabilities. But in the same time, with coding agents, you can discover processes, you can print automations much faster, and you can maintain them. This cycle, it's much faster. It makes sense for every enterprise to automate as much as they can, since the entire implementation is basically cheaper. Another point that we try to make here is we learn with vertical solutions, and that makes sense to many of our customers because they address their clear needs. They don't have the symptom of the blank page that the platform gives them. But it's very powerful to have a solution that is built on a horizontal platform because it really scales to the next use cases.
Once you are versed into printing one solution or configure one solution for one particular case, basically, you can apply the same expertise to a slew of processes or sub-processes.
If I may add, in terms of the key announcements that we made to strengthen this point. As Ashim was describing in his presentation, as was Hitesh, we're very much about selling vertical use case to our clients. That's a far easier sell for us in selling that outcome as opposed to set of components. But the products that we're going to launch tomorrow are going to directly assist with verticalized use case selling. First is the Cartographer, which maps out the context, the process context that we talked about. Second is UiPath for Coding Agents that we just announced general availability for. And finally, the orchestration layer that orchestrates and runs these capabilities, and then continuous improvement, where the Map of Work stays up to date. So when we sell these use cases, we sell the whole life cycle of how these processes are managed.
These are the key announcements that we're going to make tomorrow, which we believe are massive differentiators over state-of-the-art that exists in the market today.
Thanks.
Hi. Thank you. Bryan Bergin, TD Cowen. I want to follow up on what you were just getting into there. So I wanted to understand on that Map of Work, how does the economic model change as clients are leveraging more of that pre-built motion? Is it direct or is it more so monetizing the components to operate that workflow, like an accelerant to the implementation?
I think we address both questions. Creating a Map of Work is valuable in itself because it gives you visibility into your processes. I think many of our customers pay expensive business consultants to help them map the processes. So there is clearly a need. Even when you train a new employee, having a Map of Work, it's much easier to get them through the training processes. But clearly, it's essential in order to print the orchestration and the automation. I think our biggest play is not necessarily as a standalone product, even we consider it that. I think it has good potential as a standalone product.
For us, Map of Work, the Cartographer and Map of Work, and the Map of Work that is alive inside the platform and is kept alive by all the exceptions that human decide on, eventually go back into the Map of Work. To me, this is the overall value that we are bringing.
Hi. Scott Berg with Needham & Company. Thanks again, Dan. It was great today. So thanks for the AI disclosure in terms of ARR. It is probably the number one, two, three, four, five and six questions I have had for the last couple of quarters from different investors.
We wait especially for you.
It is funny, I had it several times this morning too. By my math, that net new ARR coming from your AI functionality over the last trailing 12 months represents about 40% of your net new ARR, at least organic. My number is not yours, plus or minus. How do we think about that mix going forward as customers are bringing more and more AI functionality in, but at the same time, obviously buying more of the deterministic and other platform features that you obviously offer? Just trying to help understand what the expectation is. Does the AI functionality, I do not know, does it get to 60%, 70%, 80% of what your net new ARR should become over a period of time? Or should this continue to be a balanced mix? Does that change at all with the vertical-focused sales?
Yeah. As I mentioned, I think we are really focused on moving the customer up the base, right? Our customers are right now. The primary purpose and intention was how can we take our customer who was RPA-only customer and make them a complete business orchestration and automation platform customer? As you heard from some of our customers, they are really trying to solve their process problems. They do not come to us with what technology is going to help them get there. Our main intention here is to help them solve those problems, and the technology just becomes a part of it.
Of course, when I mention the cohort of the customers, especially when you look at customers more than $100,000, what we have seen clearly is as these customers move from RPA to orchestration and automation, which includes our AI products as well, they are really able to solve the problem, and we have a greater opportunity and from a durability standpoint. In terms of how you expect, of course, our intention is to continue to focus on that customer base, move them up the platform, and that should drive the overall durability of our ARR base.
Hi. [Shelk McMain] from Barclays on behalf of Raimo. Thanks for taking the question. I wanted to ask more on the Map of Work and that context component, I think is very compelling. These AI agents and models are extremely intelligent but do not know much about your business and need that. However, you also have other platforms and providers out there that are also trying to build that context. I think about the data platforms like a Snowflake and a Databricks, and I know you have a tech partnership there as well. Maybe, how do you see this playing out? Because I would imagine these large customers do not want to build out these context maps in multiple places across their organization. So, maybe is there an opportunity to join forces with a data platform to I know it is slightly adjacent, but to work together there?
I guess if it is kind of a winner take all approach, then what gives you the right to win there? Thank you.
I think that's a great question, and it helps us a little bit to explain better the concept of the Map of Work. I will start to give you a bit of a high level. What are the layers into the Map of Work? Then I would like for Raghu maybe to go deeper. If you look at our exception of the Map of Work, basically, it is comprised of fundamentally, the bottom layer is what we call business ontology. Business ontology is business entities, their relationships between them, which is one thing. This is where Snowflake and Databricks kind of play. Many enterprises are modeling their data. We are not a data provider company. We always integrated. Even our data fabric provides virtual entities that can reside spatially in Databricks or in system of record. So we virtualize this data layer.
To answer directly, we do not compete on this data layer. We basically integrate with their already defined ontology. It is there. But then there is an interesting point. You have actions on entities. This is more difficult to capture into a data layer alone. This is where we start to shine, because we can combine data layer with actions. Who modifies? What are the rules? Who are the owners of an entity? How an entity? It is because it is already, you need to move an entity into the context of a business process. But even further, if you look up the stack, is you have workflows that touch all of this, and it is multiple workflows across different departments across, and they touches multiple entities. This is basically what an automation company is doing, not a data company. We combine this in the Map of Work.
Finally, you have the process orchestration that you will need this information. This is what I said until now, that's one main component of the map, which is the structured knowledge. Everything here is totally structured, well-defined in rules. But then on the top of it, you have the operating knowledge, which basically is the informal knowledge of how you run your company. Policies, exceptions, examples that are maybe in people's heads, words scattered across the enterprise. When you add these two things, you create a Map of Work. Basically, the data layer and business ontology is one part that is required.
Yeah. I think, by the way, it is a great question, one that we spent a whole bunch of time thinking about. As you saw in the other presentation, there are issues and machines. Our intention is to go from the task layer to the process layer. As Daniel and I, we think about the Map of Work, we very much think about it as the process context, not enterprise context. So when you bring up a Datadog or Databricks or Snowflake, you are very much in the enterprise context. You have a warehouse with everything where you can interrogate it to answer all kinds of questions. The problem we are solving is a layer on top of it. It is process context. It is the scraps of information that is living all over the place that humans are using to read and retrieve, to take a process along.
Because of our elevation from task to process, we also want to elevate the context layer to a process context layer, not a data context layer necessarily. This is where we play hand in hand with the Snowflakes of the world, where we add the complementary process context layer, which goes hand in hand, and with the process transformation capabilities that we want to add to our system. That's our play, is the process context, not enterprise context.
Hi, this is Brian Schwartz from Oppenheimer. Thank you very much for all the content. It was really great. I just wanted to follow up on your AI business and just talk about sustaining pricing power over time. You have a multi-model approach. You're agnostic, which is absolutely the right approach, seeing that number. But what happens in the future if suddenly we get a commoditization in model pricing and at the same time, your consumption, your outcome-based, and token revenue is ramping very high. How do you sustain the pricing power? Is it more use cases? Is it selling more products? Will you be able to raise prices on renewals? Just wanted to get at the ability of sustaining pricing power in a world that model pricing could commoditize very quickly in the future. Thanks.
Well, I think we need to untangle a bit our revenue. How much of our revenue comes from what we add on the top of tokens? I don't think it's really material. This is our source of revenue. This is an interesting question about the gross margin. Many of our customers would like to bring their own models. Basically, program cost is totally irrelevant for our business. For me, I'm a big believer that we will see much bigger commoditization on the models, and that works only to our advantage because you can use Cartographer is based. Cartographer consumes a lot of token. Cartographer agent has to analyze hours and hours of recordings. That's an expensive proposition. And our coding agents use a lot of tokens. The more intelligence gets commoditized, I think people can consume more of our platform.
Right. Specifically, tokens are not a primary component of our AI monetization. We are monetizing our AI capabilities surrounding the models. From that standpoint, if you have a use case, like we talked about one of the large automotive manufacturers, they will have their own tokens, their own models that they can plug in and integrate with us. What they're buying is our solution, our orchestration, our governance, and other aspects to make that use case come alive. So long as the value holds of the return, it gives us pricing protection long term, and we've actually seen that upon renewal for many customers.
As we position outcomes, I don't see a scenario where as models get better and cheaper, it's not effectively a net tailwind for us. If you're selling a use case, you're selling an outcome. It doesn't matter how much AI, how much deterministic is used. And if there's a cheaper model that's compliant, governed, and it meets the organization's, whatever thresholds, that is more power to us and more power to the customer, I think. So we benefit when AI models get cheaper, faster, and better.
Hi, Pat McIlwee with William Blair. Thanks for doing this today. I just wanted to ask one on Test. As we think about that as a growth lever, I think it's really interesting and notable that it puts you in front of a different buyer, with the CIO or testing buying center. But as we think about that, can Test evolve over time into more of an independent land motion where you then go in and pull through Maestro or automation workflows, or is the larger opportunity still just cross-selling it into the existing user base?
It's a great question. It's actually both. One is we should define different buyers. There's a different user, like a QA team, et cetera. But the CIO, if we're going, we also work and say, "Hey, you're moving from ECC to S/4HANA," as an example. In that migration, you should also be designing automation directly into your blueprint. Right? In many cases, there's a synergistic customer area, not a synergistic user at the end. But we are seeing also a motion that is beginning to land Test directly. And that is an area that we would continue to invest in. So it gives us a bifurcated, like a dual land lane, both a synergistic play as well as a net new buyer and a net new motion that we can capitalize. And that's what's super exciting about it and why we've specifically labeled it as a growth engine.
Yeah. I want to point you to a very interesting fact about the connection between testing and automation and agents. Testing, as we showed here, is one of the areas where you can run agents completely autonomously, and it's actually indicated. But when an agent basically tests and explores an application, it's a lot of learnings that we get from this that we plug in when an agent will automate the same application. And these learnings are not lost during the testing. So we can incorporate them basically in the Map of Work. So people start already when they use an application, because that's the cycle. You as an enterprise are building a custom application. You build it with coding agents. Human review and testing is really a big bottleneck. You use the Dark Testing Factory that tests it automatically.
You capture, you create basically a summary of all the interactions, what is successful, what not, how do you use an application? Then you can close the loop, take this summary, and put it basically when you use agents to use your application.
It is YC from Citi. Thanks for all the slides on the customer example. It is great to see the multiple from 16- 25, especially the one in Hitesh's slide. I think by my estimation, the AI piece kind of jumped 4x at least, from the FDE. I am trying to guess one around FDE and pricing. To what extent outcome-based pricing has an impact in that jump in the AI consumption piece, and then what would that kind of justify your ramp in FDE spending? Thanks.
I think there is two flavors of outcome, right? Motion two for us is you can sell an outcome of automating procure-to-pay, right? And generating value for that. From that standpoint of an outcome, we definitely get more value. We have less discounting pressure because we are articulating that value. But in terms of an outcome pressure related to FDEs going and solving a specific problem, and pricing is tied to the full outcome, like a three-as-basis, we are actually just at the early stages of those types of transactions. It is an exciting part of where we could go, especially as the platform moves to the C-level. Daniel, I cannot count the amount of C-level meetings, chief digital officer for one of the largest hotel chains within Europe or hospitality chains within Europe and globally.
Those are areas that we are able to now selectively go after it, and we want to be selective in terms of how we are attacking that area. It is actually very little of a full outcome-oriented area that is driving it. It is more just the incremental value that we are providing on our solutions.
Radi Sultan , UBS. I wanted to ask, a lot of your customers have large automation backlogs that they are trying to get through. I remember at this conference last year, every slide was, "Hey, we have 300 automations in our backlog, 200 automations in our backlog." It seems like UiPath for Coding Agents is a really meaningful accelerant of converting those automations into production. I guess my question is, can you see a core acceleration just from that quicker backlog conversion? I guess maybe there is any data points around that, converting existing automation backlogs into production faster, how that could impact the business?
Yeah. I would say that we are seeing the early signs that the productivity of the developers involved in the automation program might increase up to 50%- 60%. It is clear expectation from us that will propagate the scale. This is a thing that we just release in GA. Customers need to update. They need to get into our last latest versions in order to use this capability. This is something that we continue to roll and focus over the next quarters. Clearly, we put a lot of focus and effort, and we continue to perfect our coding agents to become better and better at the code generation.
While it is in the early stages, the other piece, it is not just a time to clear the backlog. The ROI for AI is one of the biggest topics amongst C-levels. Where am I getting my ROI? The coding agents also reduces the total cost of ownership. If you had a group now that is able to produce 20 automations a month, 50 automations a month, the average cost per automation is going way down. So we feel it is both an economic differentiator for us as well as an accelerant in terms of clearing it.
Give me a second. All right. We are at time for today. As a reminder, we have investor reception next door. Thanks, everyone, for joining us, and we hope that this was a productive session for you.
Thanks, everybody.
Great. Thank you.