Good morning, and thank you for joining us. Today's session is focused on giving you a deeper look into our latest product strategy and innovation. As Marc emphasized yesterday, our top priority is helping our customers succeed with data and AI, and Informatica will be a powerful accelerator of that strategy. You'll hear more about this from our product leaders shortly. Some of our comments today may be forward-looking statements that are subject to risks, uncertainties, and assumptions, which could change. Should any of these risks materialize, or should our assumptions prove to be incorrect, actual company results or outcomes could differ materially from these forward-looking statements. A description of these risks and uncertainties, and assumptions, and other factors that could affect our financial results or outcomes is included in our SEC filings, including our most recent report on Forms 10-K, 10-Q, and any other SEC filings.
Except as required by law, we do not undertake any responsibility to update these forward-looking statements. With that, I'm going to hand the call over to Robin for a few quick remarks.
Thank you, Mike. Good morning and good afternoon. We thank you all for joining us this morning Pacific time. As I shared on yesterday's earnings call, my top priorities as CFO for the year include delivering customer success and accelerating AI adoption to drive growth, driving operational excellence to maximize shareholder value, and responsible capital allocation. We're really looking forward to getting your feedback on this forum. Up next, you're going to hear from four of our product leaders who will share more on our data and AI strategy. With that, Srini, I'll pass it off to you.
Thank you, Robin. Again, good morning, everybody. My name is Srini Tallapragada. I'm the Chief Engineering and Customer Success Officer. In that role, I wear multiple hats. One part of it is responsible for all the product and technology for what we call our platform, our infrastructure, our 360 apps. I also have an additional responsibility of customer success teams and support. I am also responsible for our professional services, which we use mostly to accelerate our customer success jointly with our partners. I am also responsible for the South Asia business, which includes India and surrounding countries, which we formed a new unit called South Asia. That's my role. I'm very excited to be here. We have some of our top technical and product leaders here with me.
I'll let them introduce myself, and I'll pass it to MK Murlizar Krishna Prasad, who we fondly call him as MK.
Thank you, Srini. Good morning, everyone. As Srini said, I go by MK for short. I lead the whole platform. I'm the President and CTO for Engineering. I lead all of the platform that includes the core platform, Agentforce, Data Cloud, MuleSoft, Tableau. Everything that you see in the diagram, you're going to see in the diagram from a platform perspective or running on it. I've been in Salesforce about six and a half years, love this company. Prior to that, I was at Microsoft for about 13 plus years, building the foundations of all of Azure services, Dynamics, and SQL. Before that, I started my career in the Valley with Oracle, building databases. That's my short history. I live in Seattle. Passing it off to Rahul.
Hi, folks. Good morning. Rahul Auradhkar, AVP and GM for Unified Data Services. Essentially, my job is the GM of Data Cloud and also forming the foundations for AI, predictive and generative AI. Prior to Salesforce, I was in a couple of startups, including one in the financial services world. I was a fintech startup in the belly of the beast. I was in Wall Street doing foreign exchange transactions optimizations using data and AI. Spent a long time at Microsoft in the data and virtualization space. Also grew up, started my career in Indian space research as a rocket scientist, as a matter of fact. That's where I started my career. That's me. I'm going to hand it off to Patrick.
All right. From rocket scientist to marketing guy, that's a heck of a transition. Good morning, everybody, or good afternoon if you're in New York like me. It's a pleasure to be here with you all and my colleagues here, especially Robin. You've been a member of the family for quite some time, but about a quarter in now as a member of the team. It's just been really awesome to have you here and the focus and the prioritization that you've brought. My name is Patrick Stokes. I've been with the company for, I think, about 14 years now or so. I joined as a 31-year-old. I'm now a 45-year-old. I've had a few different roles. Right now, I run our product marketing organization for Ariel Kelman, our CMO. I'm responsible for the messaging of all of our products, our whole portfolio, our platform.
I also head up our community and Trailblazer and Trailhead platforms for learning, which has been a big focus of ours in the last year. I also work on our internal sales enablement. We've got a lot of work to do on that as this new technology has really been changing the world. We have a lot of work to do to get everyone inside of Salesforce up to speed on how all of this works. That's me. What we were hoping to do here, myself and Srini and MK and Rahul, is just walk you through very, very quickly for about 10 or 15 minutes, and we'll take some Q&A on our product strategy and the overall platform that we have been putting together. You may have heard Mark refer to it probably quite a few times yesterday as the ADAM Framework.
We're hoping to kind of bring that to life here for you a little bit. Let's just jump to the next slide. I'm not going to tell you anything here that you don't already know. At Salesforce, our job has always been for 25 years to help our customers connect with their customers in new ways. Every once in a while, these technologies and the world that we live in, it changes around us, right? Suddenly, we have to start doing things differently. That's no different than what's happening now. This idea of agentic AI, it has flipped everything on its head. Every piece of software, every business process that we know, we can now think about doing differently.
At every single layer, the data layer, the logic layer, and the user interface layer, they're all now just a little bit different at the exact same time, which has really never happened. We've seen big changes like the cloud and social and mobile, which have changed kind of one or two of those layers. It is rare where they all change at the same time. We have a really big opportunity. If we go to the next slide, we like to think about our opportunity as this idea of digital labor, which is this idea that we think that we can help organizations build a workforce that is not entirely based on human labor, but is now based on intelligent labor, digital labor that augments that human labor that works side by side with it. This is a really profound idea.
We've scaled compute, we meaning the industry in general. We've scaled compute. We've scaled location with the internet. We've scaled compute and storage. The last thing we haven't really scaled yet is knowledge work. We think that this idea of digital labor has the ability to do that. Now, to do it, we would need a new architecture. If we go to the next slide, our customers and really everyone out there, they really need to rethink their IT architecture. They need to be AI ready. That puts this huge emphasis on their data and their metadata. I like to use this term. It's kind of like creating a context interface to your agents. You need to take all of this context you have in your business and put it together in a way where these agents can actually take advantage of it.
That's what we've been putting together. Obviously, we call this platform Agentforce. It starts with our platform itself, which for 25 years has been built on this idea of metadata. We put that together with Data Cloud, which is more than a data lake, as you know. It's this ability to kind of federate across all sorts of data lakes like Snowflake and Databricks and then activate that data up into, of course, our best-in-class applications like Sales Cloud and Service Cloud, but now, more interestingly, these agents. The key here is that we're not just building individual agents. We're not in an agent race. What we're in for creating is a platform where customers can come in, and they can think about Salesforce as more than just a CRM.
They can use these agents to attack any business process, whether it's an employee-facing process or a customer-facing experience. We will go to the next slide. It's been really awesome. You heard Marc talk about quite a few customers yesterday. It's just been really incredible to see the use cases that our customers are coming up with. I think it's important to note that these aren't like agents that we're running out and building and putting out there. This is customers that are coming in and using our platform. They're identifying their own use cases, and they're implementing them on top of Agentforce. Whether it's SharkNinja or Equinox recommending products and providing incredible customer service or Salesforce providing incredible customer service, that's right down the middle. We've all seen that.
There are so many other use cases as well, like the World Economic Forum, who put this right into their event application, just like we did actually at Dreamforce, or Priscina, who's using it for effectively sales enablement. I'm using it for that as well internally. It's been really interesting to see how that works. There is no shortage of use cases, no shortage of capabilities that this platform has put together. I'm going to turn it over to MK to walk us through a little bit about how it works.
All right. Thank you, Patrick. The way it works is fairly straightforward. You have somebody walking into your website or WhatsApp or SMS starting a conversation. It could be a marketing. It could be a service conversation. It does not matter. It could be a trigger. Maybe somebody sent an email. Maybe some predictive logic came in and said this customer is going to retreat. Regardless of whether a human asks a question, a background system asks a question, or some change in some data happened, we can trigger Agentforce. Part of the Agentforce is we have the reasoning engine, which uses what we call as an agentic loop, which is very simple, which takes your question or the thing that has happened and starts planning on what to do.
If the question is like, what's the status of my order, we know it has to go figure out, OK, who you are, look at the particular order that's related to you, and bring that up. It could be about triggering a smart care agent because based on the device status saying it's running hot. It takes the action and then basically generates the outcome. The outcome could be just simple a text that comes back in your chat window, or an outcome could be an action that it executes on Salesforce or non-Salesforce systems in your enterprise. That's kind of how Agentforce works end to end. It works either in a synchronous way when you are chatting with it or asynchronously behind the background. We call them autonomous agents doing things for you.
That is the power of Agentforce, also backed by our trust layer. Everything here that you are seeing is trusted, secure, and protected. Next slide. Rahul, go ahead. Are you muted, Rahul?
Yeah, how about now? Yeah. I want to dig into Data Cloud here. We all know that enterprises have numerous data silos, whether it's apps, or warehouses, or lakehouses. Those silos are here to stay, and they serve a good purpose. As a consequence of the silos, they have unreliable insights, poor actionability. It's like being data rich, insights poor, and actions poor. Now, if you think about the agentic area, this is of particular concern in the agentic era. What you do need in the agentic era is the right data, right context, produce reliable, accurate actions that are obviously grounded in customer records. We built Salesforce Data Cloud and deeply integrated that into the platform.
Our customers like Wyndham, for example, or FedEx can deliver actions and activations and intelligent AI on top of it, whether it is related to Reservation 360 or Customer Care 360. Then agents become that much more powerful. Also, Data Cloud is designed to complement systems that customers already have. Like I said, it bridges the silos across lakes and data lakes and warehouses and business applications. This is across both structured and unstructured data, and all of it being done with low-code and no-code tools. Now, using that, you generate predictive and other insights that can be used for automations and actions across the entire enterprise business apps.
We can see the logos out there, our partnerships with Amazon, Snowflake, Databricks, Google, IBM, and many more really helps our customers take advantage of their existing data investments and power these fluid activations, the notion of data fluidity. We have seen a significant amount of traction and customer success over the last three years across the CRM, where Data Cloud has played the role of a binding layer, if you may, for the Customer 360 type scenarios across sales, service, marketing, commerce, analytics, et cetera. We are seeing the same with AI as well. I want to dig a little bit into Informatica here. As we all know, Informatica has been a leader in enterprise data lifecycle management across data quality, governance, catalog integration, MDM.
Their Intelligent Data Management Cloud (IDMC) will be instrumental in accelerating and enhancing the value of what we have in Data Cloud, MuleSoft, Tableau, and our next-gen architecture, and how that drives our Customer 360 and Agentforce vision. If you go to the next slide, Moonli talked a little bit about how Agentforce works. Agentforce is differentiated with Data Cloud with harmonized and unified data. And I refer to structured and unstructured with the metadata. This is related to what Mark mentioned last night. The ADAM Framework includes data and metadata as well. For the unstructured content, we break it up. We refer to that as chunking, and then we index it. Essentially, what we're doing by breaking it up and chunking it is we're putting structure around the unstructured, if you may.
Now, when the unstructured data is indexed and all those indexes are available and ready, we need to be able to retrieve it appropriately and use it for grounding generative AI responses. That is where the RAG pipeline comes in. RAG, also known as retrieval, augmentation, and generation, means you can use the word that is used as embeddings. Those are essentially the index data that is inside of our vector database. You can use that for the right data for generative AI that is needed for Agentforce. Now that is how we make all our customers' enterprise data accessible to their agents. Now, that means that we are able to make the best use of the models customers have without having to spend money for having to retrain the models, if you may, for model training. Now, the key differentiator here with Data Cloud is that we can mix structured and unstructured.
You can query not just keywords and do semantic text query, but also semantic meaning-driven queries on top of it across languages, but also structured content like high-priority cases, as Moonli called out, or something that is related to where we are retreating a customer or there is a sales deal with a specific account and so on. This RAG, also known as retrieve, augment, and generate, is special with Salesforce and the Salesforce platform with Data Cloud and differentiated since we have a combination of structured and unstructured data that is harmonized and unified and with insights across the entire enterprise application and warehouse lakehouse data corpus that is available now for grounding the generative responses needed for Agentforce. With that, I'm handing it off to Srini. Yeah.
Thank you, Rahul. I think just I forgot to tell you that today is my 13th anniversary at Salesforce. I joined when we were just about $2 billion in revenue. Before that, I was working for almost 12 years at Oracle. All my life, I've been chasing this customer 360 dream. I think as the technology evolves, different sources are coming, data types are coming in. This has been my dream to chase. I just wanted to give you that context. I've also been mostly in the enterprise software business, working with a lot of big customers. The key thing you will find, as most of you already know, is the enterprise customers' landscape is very complicated. We assume, first of all, number one, that we are not the only vendor they are using. We have to exist in ecosystems.
One of the most important things as an architectural principle of our platform, it has to be open. Because why? Because we have to ensure, even though we want to be deeply unified, I would like to call it deeply unified but open, so loosely coupled but coherent. Because most of the time when we go there, a customer may say, you know what? I've already spent the time to create my own lake. What we really want to do is we don't want them to reinvest everything and bring it to us. We want to make it very easy for us to unlock the trap data. Because our entire goal is to ensure the customer gets fast value and really help their customers.
That's what we're trying to do, that customer success, which is why we invented this new technology based on open standards, which we have spirited, which allowed us to do this zero-copy data. That allows us to bring trillions of records without really copying. There'll be some customers where they say they don't have that. We want to bring in all this data in, as Rahul mentioned in the previous slide, directly through ingestion. We also have a lot of unstructured data, again, either as a zero-copy or to index it. You have the real-time events. Zero-copy is very important. At the same time, one of the things we are realizing with the LLMs is the LLMs are moving very fast. There's a lot of competition. Each of these LLMs are like maybe six months behind, forward, which in each other.
As things are happening, we want to give not only the platform. If our customers, for some reason, pick an LLM that they like for whatever reason, we want to be able to do that. At the same level, at the infrastructure level, again, we want to give choice. Most of the time, Salesforce historically was a first party, but we have been migrating to public cloud. We support multiple public clouds. Similarly, we have what we call an AppExchange. For example, a customer may say, hey, I have standardized on Copilot for my CI/CD systems. We would like you to use that. AppExchange has a lot of third-party partners which work on that. We let customers pick and choose. That is, again, building an open system with our partners.
Now, in the new world, I think we are creating this new agent exchange where all our partners who are building agents can interact, and our agents will work with their agents using standard protocols like A2A and MCP. If you really look at it, what the key part of it is, while we want it to be an integrated, deeply unified platform, we want to make it very open and give enterprises a choice. Because when we go into that segment, usually the complexity is there, and that's the key part of our architecture. It's open and not only open on standards. If you go to the next slide, at a very high level, the way you have to think about it is we want to build the world's number one digital labor platform enterprise-wide. It's not just CRM. It's an enterprise-wide platform.
We will definitely, out of the box, build out-of-the-box agents with clear definitions and roles in the CRM use cases. As you saw in Patrick's slide before, because the platform is agnostic and it is an enterprise-wide platform, it allows customers to build their own unique cases. The way you have to think about when we talk about this ADAM framework is first, at the top is the agents. This is the digital labor platform, as you saw. Customer 360 allows us to define the jobs to be done in specific roles. If you are a sales rep, if you are a service rep, if you are a marketing rep, if you are a commerce, we are very clear. Per industry, we have 13 industries. We are baking in a lot of out-of-the-box templates. It is very fast value. The last part, which is the data and metadata.
Again, I said, for an agent to work, sometimes most of the data is in CRM. Sometimes it's not in CRM. Maybe they have a custom app. Sometimes you have to go call that app. You have to do that action. That's what MuleSoft comes in. MuleSoft is like Switzerland. It works with Salesforce very easily, but it works with a lot of other vendors too. Same with Data Cloud, as we showed before. It's open. It brings in a lot of data from different ones. People may want insights. They want to know what the agent is doing, what you can do, ask questions as they're wanted. That's where Tableau comes in. Tableau is, again, an enterprise-wide platform, very well integrated with CRM use cases, but people use it for a lot of other use cases too.
The last one, which is Informatica, which I think what Informatica will do is if you have all this data all across enterprise, what you need is you need a system to be able to govern, to catalog the data, to know the source of the data, understand semantically what it means so the agent has the full context. I'm sure all of you have experienced this. If you go to ChatGPT or Gemini or something, the more context you give to the prompt, the better the result is. That is the metadata which we talk about. If you really look at it, this is a little bit of a simplified view of the architecture. Mentally, you have to think of Agentforce as a digital labor platform. It requires knowledge and actions and jobs to be done.
The jobs to be done are defined by the Customer 360 use cases with deep process and workflows and jobs to be done definitions per industry that we have built over 25 years. Then you've got this access to the knowledge, which is the data and metadata platform, either integration with actions through MuleSoft across enterprise, Informatica to look at, understand, govern, and catalog data, Data Cloud to unify and activate that data with zero-copy, and then Tableau, which drives insights and action. Hopefully, you understood now why we realized Informatica is going to accelerate our overall strategy. With that, I'll give it back to Mike for Q&A.
Yeah. Thanks, Srini. Thanks, team. We're going to move into the Q&A portion of the call. What we'll ask for folks on the webcast to do is raise your hand if you have a question, or you're welcome to put it in the chat window. Then we'll call on you to ask your question, or we're happy to read it aloud as well. While we wait for questions to come in, I'm going to kick it off here with a question that's come up a lot since the deal was announced on Tuesday. It's going to pick up on where Srini just left off, actually, on Informatica. Maybe I'll kick it to MK to start.
For the audience here who may not be as deep on the data architecture and all the terminology that we're used to in our day-to-day work here at Salesforce, MK, can you just double-click a little bit into what it means when we talk about MDM and governance vis-à-vis kind of how we think about Data Cloud today or what we're offering currently in Data Cloud today? What we've talked a lot about, obviously, is the synergy of the two together. If you can help just piece that together for the team on how Data Cloud and Informatica will work well together.
Yeah. Sounds great. Thank you, Mike, for the question. As Srini just mentioned, Informatica does a whole bunch of things. One, it gives us the large-scale connectivity, 10,000-plus different connectors that we can go integrate data sources into, ability to kind of run some of that transforms and everything in the local clouds, including five of the major hyperscale clouds. Now, in addition to that, it also provides us a few more things. One is the MDM capability, which gives us the ability to kind of create these unified profiles of not just customers. We are really Data Cloud is really good at creating a customer 360, account 360, and so on, what Informatica sort of expands that into the B2B scenarios in terms of product 360 and other kind of 360s as well.
In addition, MDM becomes a source of record as well, which can be like we are a source of truth. They're kind of a source of record for that. We integrate, as before, Data Cloud is an extensible platform. We already integrate with MDM. This will be a natural sort of integration point for us to offer to our customers, joint customers. The second thing is governance. Governance is a lot of different things. First, if you really start with from the data side of the house, you also want clean data first. That's part of the governance. Because if you want to have agents answering your questions, as an example, you want to make sure it's doing the right thing. Informatica sort of expands our data horizon to add a lot of the cleansing capabilities today.
That will be a great complementary thing to us so that it's not just about data connectivity. It's also making sure you get the clean data. Second, you want to make sure that you put the right governance rules in terms of who can access it, who can see it, and so on and so forth. Now, Data Cloud has a very good sophisticated governance model on top of all the Salesforce data and other data you're bringing in. Informatica takes it to the enterprise. You have one data catalog and metadata catalog across your entire enterprise. We have the opportunity to bring in data and APIs and actions and everything into that one framework and push it down and make sure you have a common, consistent governance across your entire enterprise landscape.
You also have this thing that Informatica adds to our arsenal, which is the CLAIRE agentic metadata agents, which basically allows you to then start querying and understanding your metadata better so that you can go create better agents. You can go create better business processes in Salesforce for servicing your customer. The way we look at it is a great complementary solution to what Data Cloud already offers but brings in the power of cleansing, enterprise-level governance, catalog, and metadata.
Great. Thank you, MK. Just to pull the thread one more layer, one of the other key synergies from a deal standpoint was really the combination of MuleSoft plus Informatica as well in the parallel construct. Maybe I'll give it to Srini, and he can talk a little bit about that.
The way I think you should think of it is integration. There are a lot of these integration patterns. Basically, like I said, just as a meta level, remember this. An agent requires jobs to be done. What is asking to do these roles? It needs access to knowledge, just like a human. It needs to be able to act on those. What Informatica helps us is there are multiple sources of data. Informatica really helps us to access the data. We have zero-copy integrations if somebody has done, but there are thousands of connectors. There is still, by the way, 70% of the last study I read of data is still in on-premise in some mainframe. Informatica has this built, like MK mentioned, thousands of hundreds of thousands, 10,000 connectors or something, which you will be able to access.
Sometimes you need to be able to act. What MuleSoft really gives us is an app-to-app integration, an application integration. You need both data integration and application integration along with a gateway. That is how they are complementary. You have to think of these as not a one-size-fits-all. Enterprises are complex. Sometimes you want to just call an API and get the job done. Sometimes you really get the data inside. Sometimes you want to do zero-copy. Sometimes you want to stream in those data. Sometimes this is unstructured data. That is the complexity, and that is what really makes the agents work.
Great. Thank you. Now that we've got everyone warmed up, we're starting to get some questions in here. This next one, I'll throw to Patrick. Patrick, we've talked a lot about, obviously, the customer adoption cycle on Agentforce and what we're seeing within the customers. The question really that's come in is about kind of what are the use cases that we're seeing most commonly? I'm paraphrasing a little bit of the question, but what are we seeing most commonly with Agentforce use cases? Are we seeing any unique use cases coming in, or what's been the reception from customers on the various use cases?
Yeah. I mean, I think the customer service use case, obviously, is right down the middle. I think what most customers are attempting to do first to kind of really learn the technology. Because remember, this is a new technology. Everything, the way you build with it, is very different. What most customers are starting with is just this idea of getting an agent that can answer a question. I'm going to expand on that a little bit because the difference between answering a question inside of something like ChatGPT or Gemini versus answering a question about your business is quite different. If you go to ChatGPT or Gemini today, you can ask it a question about your business, but it's only going to know what's publicly available. That's not what our customers want.
Our customers want to be able to ask a question that has the latest customer data connected to it, the latest inventory data connected to it, the latest web engagement data connected to it, whatever that is. Using Data Cloud to first get access to that data and then using our RAG techniques, which Rahul talked about, that are connected into Agentforce to make it actually answer that question, it sounds maybe small, but it's usually a massive unlock for our customers because it kind of runs them through everything that it kind of takes up to one last piece, which is, OK, now how do we get it to perform an action? Sometimes, MK showed you the example a few moments ago. In order to go query your inventory system, you might not have that data on hand.
You might have to hit an API to go get it. Or maybe you want to change an order, and you have an API for that. That's usually what comes next. In terms of the kind of things that have been a little bit surprising that we've seen is, one, it's really, really good at kind of matching. If you have any type of marketplace type of scenario where you have someone that needs something, someone that has something, and you try to match them together, it's really good at that. That's something that humans are really slow at in that matching, that analysis. It's very fast at that.
The other thing that's been interesting, I think, and this is a little bit new, is we're seeing this need for a hybrid environment where you have some what we'll call determinism, which means strict business logic mixed with what's called non-deterministic or stochastic, which means random logic from the LLM. You ask an LLM a question, you get an answer. You ask it the same question 10 minutes later, you might get a slightly different answer. In a business context, that's not always what you want. You might want some real determinism mixed in. It has been interesting to watch customers think about how to mix those two modalities in a way that doesn't eliminate the value of the LLM in the first place. You don't want to bring so much determinism in that suddenly now you've just re-architected something to achieve the same thing.
You want to bring that value in. That's been really interesting to watch our customers spend some time with.
I think, Patrick, I'm going to keep it on you. I'm going to go to the next question here in just a second. I want to pull a thread of what you just talked about. One of the other things that we talked a lot about internally, but I don't know that we've really brought it up with investors before, is just the ideation that occurs with customers as they get the deployments going from an agent standpoint. One of the use cases we're actually starting to see now is from an internal versus external-facing agent. Can you just talk a little bit about that dynamic?
Yeah. I mean, most customers want to start with an internal agent first just because it's easier to control. If you kind of mess up internally, depending on your use case, it's maybe not the end of the world. If you mess up in a customer experience, it can be very, very costly for the reputation of the business. Most people want to test there. At least that's what we're seeing.
Got it.
Can I add one more to that? I think, particularly with internal agents, we are seeing the explosion of Slack agents like crazy within our enterprise because Slack is such a natural interface for employees. Everything from sales agents to engineering agents to service swarm agents and so many more. It is like literally we have hundreds and hundreds of agents running internally on the same platform but integrated with Slack.
Let me add one more thing as well. I'm sorry to come over the top again. One lesson that I've really learned interacting with customers that are new to this, and this is, I think, frankly, I'm getting a little inside baseball here, but I think that's the point. It's a little bit of a difference in our sales motion of the last 10 years. What I mean by that is we've built this incredible sales machine that kind of starts with a business outcome and then goes to our customer and says, look, Salesforce can give you this business outcome.
If you do that right now at this moment in time with Agentforce, if you build the customer their agent, if you identify a use case, and then you go off and you build them the proof of concept, or you bring in a partner and have them build it, and you put it into their organization, they might get some value from it, but you're really, really limiting the long-term applicability of this. What we really want to do is we don't want to build that proof of concept for the customer. We want to bring the, we want to send someone out to the customer. You've heard Mark refer to this term forward-deployed engineer. We want to join their scrum team.
We want to work with them to build the use case on top of the Salesforce platform or on top of Agentforce precisely so that they see that it is a platform, that this is not just a series of applications or pre-built agents, that this is a platform which then they can take the first agent and build their second and their third and their fourth. I think that is really fundamentally different from both the small startups that you see that are building domain-specific agents but lack a platform. It is also fundamentally different from the Googles and the Microsofts and the Anthropics and the OpenAIs, which do have a little bit of a platform, but it is largely optimized for the consumer side. Like you see Gemini, it is really slick how it is embedded into Google Docs and spreadsheets and things like that.
Is it really going to be your enterprise platform? I think so. Helping our customers or exposing that platform to our customers, I think, is a really important part of our sales process, our sales motion at the moment. Frankly, we still have some acceleration to do and some room to build on that.
Great. Thank you. The next question, and we're going to stay on AI here for a moment or on Agentforce for a moment and talk about differentiation. There's this concept around being the platform and having the Atlas Reasoning Engine as a key differentiator when you have the LLMs and whatnot that are operating from the same context but doing it from a different angle, if you will. From Jack, how would you talk about the Atlas Reasoning Engine as a differentiator when the LLMs are providing similar reasoning capabilities and becoming more prominent?
MK, you want to take that, or I should take that?
Yeah, let me take that. Yeah, I can take that. Actually, as Patrick briefly mentioned, if you're just doing public content searching on ChatGPT or Google, et cetera, the reasoning engine is good enough there because it has access to all the public data. When you come into enterprise, you do not have access to the data. In fact, what Mike can see, probably I can see. He can see all the financials. Maybe I can see. There is security as well on that data. It is not just the data. You need better than RAG and other techniques. What we have done with our Atlas Reasoning Engine, which is different than just what you see, sort of the Gemini's, ChatGPT of the world, is the ability to then mix the two.
You want the power of the big reasoning models out there so that you can actually have a conversation, et cetera.
It's good to clarify that we use the latest reasoning models as the base models.
That's right.
We're not replacing them, but I think that's not enough in our experience.
That's not enough. That's why those models are awesome at reasoning, conversing, and talking to you, et cetera. We then need to blend it with the secure data that you may have access to, what Srini has access to, Patrick has access to, along with the app APIs. That's where the Atlas Reasoning Engine comes in. It takes that query or question or whatever that may be. It figures out using that LLM to see which data makes sense to answer the question, which APIs make sense to take the action. It goes into a loop. It's not just a one-note, go look up the data, call the LLM, answer. It makes sure that the answer is even right. If it's not right, it'll go try again. Maybe refine the question a bit more.
Ask the user saying, hey, did you really mean this? And so on. That loop capability is what makes our Atlas engine so differentiated. We are seeing the results already. Like if you look at our own help.salesforce, we have now really reduced escalation to humans by more than 50%. That's because our sort of answers are so good that people do not ever have to then talk to a human. At least 50% of them do not ever have to escalate to a human as an example.
Great. Thank you, MK. If we could unmute Garvan. Garvan, if you want to come on and ask your question. Garvan, are you there? OK. I'll go ahead and ask Garvan's question. What are the key blockers for us? Garvan, are you there?
Yeah. Just Garvan enabled. Thanks for doing this session. I just wanted to understand what are the key blockers for the scaled adoption of the Agentforce trials? Then how important is Informatica in ensuring that happens? To the extent Informatica closes until next year, how does that limit CRM's capability to scale up the use cases? Thank you.
OK, let me take that. Garvan, thanks a lot for the question. I think, as I mentioned previously in my expanded role, I also run customer success. We created, first thing we realized is because the technology is changing so fast. What we did is we created what we call a forward-deployed engineer, where we go and work with these customers, initial customers. We are trying to mature the product, really understand the use cases, pattern match it, and bring it to the broader platform. That is the motion we are in. As part of that, one of the things we are seeing is there are three types of friction. One is customers have a lot of pilots. The reason they are not able to do the pilots to production, and this is what the DIY they have been trying for two years.
Everybody has a good demo. They don't have a live. One of it is they don't have an infrastructure to test it because V1 is very easy. What happens once the agent is in production? Who updates the knowledge? What happens if the reasoning model changes underneath that? Who is testing it? This is a different motion. One of the friction points we see is people not understanding that it's a different type of system. How do you give the prompts? They don't have the knowledge. It's one thing to just build a temporary demo to a production consistent environment. That's one friction. Other friction we see is, again, we see different class of customers. Some customers have invested a lot in having their enterprise data estates right. Usually, there is a central team which owns that entire data estate.
If you are the head of customer service, you don't know. They have to ask those persons sometimes. Some customers have also similarly invested in APIs and have an enterprise API gateway strategy. Depending on the maturity, it's at different levels. One of the friction points we see is getting all of the data. In most of the cases, when we are going in, we are partnering. By the way, even today, we are not only an investor, but we are a partner. We have a lot of knowledge about Informatica. A small difference, so it is not going to stop us to implement Agentforce success. What we really want to do is usually there's an Informatica person in a company, us, sometimes there's DBT, some other person. A good example is MuleSoft. Let's say if you have MuleSoft, you standardize on MuleSoft.
All these MuleSoft APIs are automatically available in Agentforce as actions. You do not need to do anything. We really compress that. If we are using EPAGY, will we not work? No, we will work with that. This is very important. We want to be open. We are not going to force the customer to say, now you have to do a huge migration project. There is a little bit of hand-holding. There are multiple vendors. Part of us is our goal is to really solve the customer success problem in a fast, efficient way for the customer, leverage their investments, and accelerate it. In that sense, Informatica, while we will continue to partner, it is critical. It will not be limited to Informatica.
It will be limited to multiple open vendors, including with us as part of our stack, will make that part of friction even faster. That is how we see it.
Great. Thank you, Srini. If we can unmute.
Yeah, please, Rahul.
A couple of things on top of Srini, what Srini said. We're seeing pretty rapid, as you saw the numbers we talked about last quarter and this quarter as well, the thousands of customers that we have closed with Agentforce. We're seeing rapid deployment. Now, these very few enterprise companies really are doing business use cases at the pace and rate at which we are doing with agents in Agentforce. Just to put it in perspective, we really launched Agentforce in October, November. Since then, we have seen this rapid success. There was another piece of the question that Garvan had was, what does Informatica help us with? In the early days, we are seeing a significant amount of traction in the space that we are seeing value in customer service use cases that, for example, Patrick mentioned, the matching service, the product recommendation service.
Those are some of the things that we see, for example, a member concierge type service, for example, that our folks in Heathrow and Equinox are using. Those are all so closely related to CRM, C360 type use cases and the adjacencies associated with it. Now, imagine a world in which Informatica, like Murlizar said, gives us access to all the metadata across the entire enterprise, all the business apps, the catalog across the entire enterprise, the governance across the entire enterprise. Just increases the aperture through which we look at agents in the enterprise business use cases, not just in the adjacency of CRM and CRM use cases, which are right for us to drive agents with, which we're doing fast and quick. Just start when the deal closes, this changes the aperture.
It changes the entire landscape through which we can look at agents through the entire business landscape, business apps landscape.
Great. Thank you, Rahul. Parth was asking actually a similar question. I am going to pull the thread and we will keep it with you, Rahul. Can you conversely then, because it has been a common question on partnering versus buying Informatica, just pull the thread a little bit more on what is the incremental benefit of having Informatica in the family as opposed to just partnering with them? What is the limitation when we are just partnering with them?
Yeah, I think some thoughts that come to mind with partnering is that we have had quite a bit of success with our partnering strategy. As you've seen, I refer to this as the tail that was wagging the dog, as in we brought all the big data players to the mix here, the Databricks of the world and the Snowflake of the world with a standards-based approach. It takes time. It takes a long time to bring them all to the same protocol, the same sharing standards, same ability for us to share data across different planes. Also, in turn, every enterprise, every partner would have some watermark through which they believe that it's commodity and they don't compete about that. All partners come to that level and say, at that watermark, we don't care whether it is query-based abstraction or file-based sharing, et cetera.
Once you start looking at other things, as in saying, hey, I want to do things that are magic sauce on top of it, it's harder to partner on those levels. Now, with deep integration with what they have with Catalog, for example, what deep integration that they have with MDM, Murlizar and Patrick mentioned both the fact that, for example, Informatica does Partner 360, Vendor 360. They do all the hierarchies underneath that. The hierarchies are updated every day, every hour, if you may. All of that is hard to accommodate within a partner framework that is just standards-based. We can only go that far with a standards-based partner network. Now, with this deep integration, it changes the dynamic pretty dramatically.
It is also a single sign-on, security flows seamlessly. There are so many other benefits that you get when you integrate these systems together.
Great. Thank you. The next question really is around differentiation of our agentic AI. When you look at what we are offering compared to the marketplace, and obviously one of the debates within the investor community is about us being crowned a winner in AI versus a loser in AI. A lot of the questions tend to revolve around differentiation. I will look to Srini on who he wants to hand this to. How do you talk about or think about our agentic AI solutions vis-à-vis the competition out there, whether it is copilots or other AI offerings from some of the agentic competitors?
Let me frame maybe the landscape and then we'll get you to one level deeper. Today, what we see is when we are on these accounts, like I think we announced we got more than 8,000 customers, 4,000 that are paying, 1,000 are in production. What we see is a range. There are two types of competition we see. One is what we call point vendors. At last known, there are 1,000, more than 1,000 vendors. Everybody is starting an AI company, small point vendors. They are going for the niche way. They go with a line of business or a specific use case, and they're trying to get something live. At the same time, we've got these hyperscalers, the big vendors, where they have their own, they have what we call do-it-yourself, they have a broad ecosystem. They have a lot of block services.
Customers have teams which understand this, whether you take AWS or GCP or Microsoft. That is another side. Meanwhile, we also have the third vector, which is a company like Anthropic or OpenAI trying to get into the enterprise segment by trying to create these forward-deployment engineers and things like that. This is what we see. Most of the time when we are working at these customers, sometimes they're having multiple vendors trial out. Maybe in one department, they try Salesforce. In another department, they're trying a vendor A. Meanwhile, their IT team is trying to build their own DIY on the hyperscaler. This is what it is. This is very early.
The key for us is to really ensure, that's why I'm obsessed with customer success and really ensuring not only the prototype stage is fine, but we also get them live and consuming. Within this V3, we see three types of customers. The customers are like, they tried it. They tried it with different vendors. They're realizing that it's not, day one is easy. What about day two? Who monitors it? Who updates it? Things like that. Those are the customers who built an agent on Agentforce, the first agent. They're coming back to us. That's the 30% net new renewal we are seeing from existing customers because they're now going to scale. They realize that I need more enterprise-grade data. I need data residency, auditing, compliance, testing center. That is what is the requirements. There, I think that's one use case we are seeing.
The second is like they're trying to do A/B test. They'll give 50% of the traffic to us, 50% of traffic to some other vendor, and they're really testing because they also do not want to commit to one vendor because everything thinks it's too hard. Where the biggest friction point we saw from them were in a little bit in the older model, they would sign huge contracts. Now they're saying, hey, I'm not going to sign a huge contract. And that is the feedback we got, which is why we changed our pricing model because there are two use cases. The customers are saying, hey, the conversation model is very useful when I'm trying to do my customer. I don't want to know how many interactions he's doing. But in some other cases, I'm not so sure. So I want an action base.
If the agent does more, I'm willing to pay more. We really simplified our pricing model based on their feedback. We have two options. This is what they're realizing with the hyperscalers. They learned that it's easy to build, but I have to call 30 APIs in the core services. Each API is metered differently. I need to figure this out. This is very complex. Usually, like anything else, like just I'm running it internally on hyperscalers too. What Robin wants me to tell her is I'll always meet my she doesn't want me to increase my spend. She wants productivity and she wants predictability in my spend on the infrastructure. This is what they are facing. We have a competition, and this is how we are doing.
The third case is the halo effect sort of vendors, I call it. The model vendors like Anthropic or OpenAI, where they're saying that, hey, I own the model. I'm going to build it. Their problem really is we work with them. They don't even understand SOC, ISO, FedRAMP high, IL6, IL7. This is all like a compliance and audit. They don't have what we call the DNA, enterprise DNA. Can they do it? Maybe. Right now, no enterprise really. It's one thing to do for their internal thing. They really do not want it as a platform. We are differentiated. We are doing an AND motion. We are trying to compete with the smaller vendors by having a platform and really going head to head with those use cases.
We're also trying to go the hyperscaler route and simplifying, saying that, hey, do you want the Lego blocks or do you want a much more platform that you can move fast? I would like to say that it's very early. We have as much, which is why this forward-deployed engineer customer success, that's what we are laser-focused on. I think initial indications of 30% of people coming to us for additional things, just like in Data Cloud is, I think, more than 45% or something. That shows the traction and the flywheel. That's how we are focused. I don't know if MK or Rahul, if you want to add anything else.
No, I think you covered it, Srini. I think the one thing I would say, I met like 10 different CIOs the last few weeks. Almost every one of them had tried, I think, Copilot as an example, spent a year. The good part is they actually know all the challenges. Now when we are going with the Agentforce, it's actually very easy to explain to them saying, you know, you tried it, it did not work. Here's why. It's not just about the reasoning engine or others. We also have a full-fledged lifecycle around it. People are thrilled with what we have shipped recently with our tooling, like Interaction Explorer, which actually tells you what's working with your agent, what's not working with the agent. Our entire analytics works on the same stack because all the data, everything is aligned.
You can actually start doing analytics on what's happening with your agents. With Service Cloud, we're releasing this Omni Supervisor, which would actually tell you how your humans and agents are behaving in one screen. You can actually escalate. Even if it's an agent talking to a customer who's not doing the job, you can actually intervene as an example. It's not just about the agents. It's about the entire platform. It's also about the tooling. It's about the entire experience. I think because people have tried other things, now it's much, much easier for us to go in and say why our platform can actually make them better.
Yeah, one thing I'd add there is Murlizar mentioned reasoning engine. It's not just about the reasoning engine. Reasoning engine is the LLM. And we all know LLMs are infrastructure. In fact, Microsoft CTO said it as well. Just with infrastructure, you don't make business context-sensitive agents work. You don't make business context-sensitive apps work. That's where the entire set of the Salesforce Unified Platform that we have comes into the picture, whether it's automation or whether it's data or whether it's metadata or whether it is context, the unified context, all of that makes a difference. LLM is just infrastructure. That is a powerful infrastructure, but it's infrastructure.
Great. Thank you, team. We only have a couple of minutes left. I think this is a good question to end on that's come in. Patrick, maybe we can start with you on this one. One of the things that we talk a lot about, obviously, is the fan that you showed where we've got the layers. You've got the data layer. You have Agentforce on top. You have Customer 360, obviously, as the center of gravity in all of it. I think for the audience here, can you help us understand a little bit around what innovation is happening and how we think about the interaction of Agentforce and our AI strategy, data and AI strategy, feeding our applications and how that helps power our applications for our end users?
Yeah, I mean, I started this morning by saying that everything that we know about software is changing. The data layer is changing. The logic layer is changing. The UI layer is changing. To me, what that means is it's important for Salesforce to think about that even with our own applications. We have an incredible sales application, an incredible service application, marketing, commerce, et cetera, all of our industry applications. We can't just sit around and let everyone else figure out a new way to do CRM and do service. We have to do that. That's our job to do that.
When we think about these applications, I like to use this silly analogy at the UI layer when I'm talking about that, which is that if you're a digital artist in the last 15, 20 years or so or a photographer, maybe 40% of what makes you good at your job is being a creative, artistic person. But 60% of it is just knowing how to use fricking Photoshop. That's a really hard application to use. It's very dense. You have to know what buttons to click and how to click them and in what order to get the drop shadow that you want just right. Salesforce is not much different than that, if we're being honest.
It's a high-fidelity, dense user interface that you need to know quite a bit about, whether you're on the implementation side, setting up in IT Salesforce, or if you're a salesperson who just came out of a meeting and your job is to enter details about that meeting. We need to be thinking about those interfaces completely differently. I think this is where Slack really comes in. We think that all of these agents, all of these conversations, all of the way that we're going to do business, we think that these high-fidelity user interfaces full of buttons and drop-down menus that have been built and optimized over 20 years for humans to click are going to slowly start to melt away. Some of those actions will happen autonomously. The ones that do require human interaction will happen through Slack just in conversation.
We're really rethinking every single application a month from now. No, two weeks from now out in Chicago, we have our Connections Conference. It's our biggest marketing or digital conference that we have. We'll introduce Marketing Cloud Next. I don't want to blow the lead too much on it. Imagine a world where your campaigns are self-optimizing. Imagine a world where when you send a marketing email or a text, instead of it coming from no reply and if you try to reply to it, no one will answer, imagine if now all of those marketing communications had an agent on the other side and you could reply to a marketing email. Imagine if your campaigns not just optimized themself, but built themself, all the creative. This is the world that we can go into.
It all starts just from ideation and being able to articulate your ideas to some sort of agent that can go out and act. I think you're going to see that. You're already seeing it start to filter into all of our applications. I like to use this phrase like, we're going to disrupt ourselves. That's what we're going to do here. We're going to go disrupt all of our own applications. If anybody's going to reinvent CRM and service, it's going to be us. We're in it to win it.
Just to add a metric there. Just as a, I'll make this 30 seconds. We have talked about lead scoring and matching service product recommendation, et cetera. Salesforce on Salesforce, we refer to that as Customer Zero, for example. We have disrupted the way we do lead scoring, for example. We are now able to score leads. Once you have a web agent that you're looking at asking for questions, we can create and score a lead within 30 seconds. Others tell us that they take hours, if not 20 minutes. We are so used to take 20 minutes. Using all of that, a combination of marketing and sales with privacy-safe first-party data that's harmonized, unified, now agents are able to go drive ROI associated with paid media campaigns. We're getting 5x improvement in paid media just in the first 12 months that we've been starting using it.
That's just an indication of how we are disrupting not just all the applications, user interfaces, as well as the business outcomes that we're seeing.
Perfect. I want to thank Srini, Rahul, MK, Patrick for being here. We appreciate all the color and insight. For everyone that joined us, thank you for joining us today. As we mentioned at the start, this is a new forum for us that we are experimenting with. We would really appreciate and like your feedback on it. Please reach out, provide us ideas if there are other topics you want to hear about, but also feedback on the forum and the structure of it. With that, we want to thank everyone for joining us and hope everyone has a good day.