In this Dreamforce interview, Rahul Auradkar, executive vice president and general manager at Salesforce, joins theCUBE’s John Furrier and George Gilbert to unpack how Salesforce is operationalizing the agentic enterprise. Auradkar explains why “the AI revolution is a data revolution,” detailing Data 360 as the foundation that unifies context across structured and unstructured data – spanning Data Cloud, Tableau, MuleSoft’s Agent Fabric and (pending regulatory approval) Informatica. He shares how “intelligent context,” semantic understanding of documents, NL2SQL and policy-based data governance (tagging and masking) enable agents to act with precision. The discussion explores agent-to-agent governance via MuleSoft, governed enterprise search across sources such as Slack, Google and email, and the interoperability path using A2A and MCP so heterogeneous agents can work together under consistent controls.
The conversation dives into real-world progress and lessons. Auradkar cites Salesforce’s own help agent answering ~1.8M customer questions in under eight months and a small-footprint Data 360 deployment at MIMIT Healthcare saving over $1.5M annually. He addresses “agent exhaust” (telemetry and traces) feeding back into Data Cloud for continuous improvement, why context can trump sheer data volume (“needle in the haystack” and data fluidity), and how open standards plus an ecosystem – including Snowflake – accelerate time-to-value. Looking ahead, Auradkar outlines the expected benefits if the Informatica acquisition closes (catalog/metadata, MDM and governance) saying customers should see some immediate value alignment and deeper product integration on roughly a one-year horizon, all in service of measurable customer outcomes.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Dreamforce 2025. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for Dreamforce 2025
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for Dreamforce 2025.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Dreamforce 2025. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to Dreamforce 2025
Please sign in with LinkedIn to continue to Dreamforce 2025. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Rahul Auradkar, Salesforce
In this Dreamforce interview, Rahul Auradkar, executive vice president and general manager at Salesforce, joins theCUBE’s John Furrier and George Gilbert to unpack how Salesforce is operationalizing the agentic enterprise. Auradkar explains why “the AI revolution is a data revolution,” detailing Data 360 as the foundation that unifies context across structured and unstructured data – spanning Data Cloud, Tableau, MuleSoft’s Agent Fabric and (pending regulatory approval) Informatica. He shares how “intelligent context,” semantic understanding of documents, NL2SQL and policy-based data governance (tagging and masking) enable agents to act with precision. The discussion explores agent-to-agent governance via MuleSoft, governed enterprise search across sources such as Slack, Google and email, and the interoperability path using A2A and MCP so heterogeneous agents can work together under consistent controls.
The conversation dives into real-world progress and lessons. Auradkar cites Salesforce’s own help agent answering ~1.8M customer questions in under eight months and a small-footprint Data 360 deployment at MIMIT Healthcare saving over $1.5M annually. He addresses “agent exhaust” (telemetry and traces) feeding back into Data Cloud for continuous improvement, why context can trump sheer data volume (“needle in the haystack” and data fluidity), and how open standards plus an ecosystem – including Snowflake – accelerate time-to-value. Looking ahead, Auradkar outlines the expected benefits if the Informatica acquisition closes (catalog/metadata, MDM and governance) saying customers should see some immediate value alignment and deeper product integration on roughly a one-year horizon, all in service of measurable customer outcomes.
In this Dreamforce interview, Rahul Auradkar, executive vice president and general manager at Salesforce, joins theCUBE’s John Furrier and George Gilbert to unpack how Salesforce is operationalizing the agentic enterprise. Auradkar explains why “the AI revolution is a data revolution,” detailing Data 360 as the foundation that unifies context across structured and unstructured data – spanning Data Cloud, Tableau, MuleSoft’s Agent Fabric and (pending regulatory approval) Informatica. He shares how “intelligent context,” semantic understanding of documents, NL2...Read more
exploreKeep Exploring
What event is being discussed in the text and who are some of the key participants?add
What assets are being utilized to support the AI foundation and the concept of Data 360?add
What was the evolution of human interactions with customers in terms of data maturity, particularly distinguishing between structured and unstructured data?add
What new features or capabilities have been released related to governance in MuleSoft and Data Cloud?add
What is the role of foundational products and frameworks in the context of the AI revolution and data management?add
>> Welcome back, everyone. I'm John Furrier, host of theCUBE, here live in San Francisco for Dreamforce 2025. Here, all the experts are gathering to discuss the future of agents, agentic infrastructure and how that can turn into productivity, breaking down those silos, turning the data into action. And ultimately, creating more productivity. Of course, George Gilbert, our analyst at theCUBE Research, has been doing a ton of work in the area and we've got the leaders at Salesforce. We've got an EVP CUBE alumni, Rahul's back, general manager of Data 360, which is new, and AI Foundations with Salesforce. Rahul, great to see you back on theCUBE.
Rahul Auradkar
>> Absolutely delighted to be here. Thanks for the opportunity again.>> I've loved your podcast with George. You've done multiple deep dives on theCUBE Research, so thank you for doing that. It's been very helpful for our audience and people making decisions. To your title change, so 360 is out there. You got Data 360, a new name. What does it mean? What was it named for or what does it encapsulate? What does it mean? And how is it different from the conversations you've been having?
Rahul Auradkar
>> If you think about what we've been telling our customers and we are telling the industries that the AI revolution is a data revolution. Now, what does it actually mean? We have had a whole bunch of assets that we are bringing to bear to deliver on that AI foundation. Things like Data Cloud, as you know, has been an organic innovation that's grown significantly. And then, you have Tableau and then you have MuleSoft that is driving iPaaS, also activations with Agent Fabric, et cetera. Then, we have this new acquisition that is still not closed. We have an agreement to acquire Informatica, that post-regulatory approval, we're going to close. If you think about all of those assets that forms the foundation for what we refer to as Data 360. What was in the past, we called ourselves as Data Cloud. As we were getting started, we were looking at digital automation, bringing all the signals for marketing and personalized experiences in real time, and then we bound this 360, which means getting consistent experiences across sales, service, marketing, et cetera. And now, in the world of the agentic enterprise, that bar has gone up pretty significantly. Finding 360 information, the context, if you may, across all different sources, they all have to come to life in real time. So, that's what is Data 360.>> So, in your conversation with George, you've talked about system of intelligence, which I love. We've been talking about that for a long... But you guys drilled in on this, what it means today. If you look at the AI market, the road to superintelligence is very much an acceleration to machines and humans working together, that's a data problem, data opportunity. So, how does systems intelligence move from these silos, which you guys were working on, to a path to superintelligence, which means more reasoning, more reinforced learning. Take us through here the dots connect because what Agentforce points to is a fleet of agents. You're seeing all these things. What is that path from here to there? Assuming we forget the data and superintelligence, that's a debate over dinner, but it shows acceleration.
Rahul Auradkar
>> So, here's the paradox, right? We have all of these insights and data across different systems that enterprises are built over time. You want to know more about a customer, you want to know more about... I always use this example. My daughter and I were buying a car and after we researched the car, I keep getting ads and I keep getting information about the car I already bought. Why would I buy another car within a month of having bought one? So, that you have insights about what that car was... They know me, that's good news, but some other system knows I bought the car, but there are AI agents telling me that I should buy the car again. That doesn't make sense. So, what is intelligence in the context? You need to unify, you need to harmonize that context. That is intelligent context. We did talk about intelligent context in Marc's keynote. I talked about it in my keynote. So, that's one big change that we are doing, as in providing context across structured and unstructured data across the 360, that allows agents to be that much smarter and more intelligent. That transition is what the industry is going through and we are right there providing it for our customers.
George Gilbert
>> So, let me ask about this. In the age of AI, data programs AI. So, the applications are just surfaces to activate or to operationalize a decision that an agent makes, but customers have to get that unifying data foundation in place. Help us walk through a maturity model for where groups of customers are in integrating and harmonizing their data and not just the structured data, but increasingly, the unstructured data that gives context even around that.
Rahul Auradkar
>> Let's break it down into what you called out, which is... You called up three concepts there. One is the activation and the second one is the type of data, and the third one is the maturity model, right? Let's break it down. From a activation standpoint, our agents really require that much more data and context to drive activation. Now, just imagine a world in which there were no agents, humans required that context as well. How were humans interacting with their customers? From a maturity model, then let's break it down into structured and unstructured. From a structured standpoint, we were getting reams and reams of structured data that people, like CRM data, along with data that is sitting in Snowflake or Databricks. In that maturity model, we are fairly mature. That's what Data Cloud was in the past, we were the best CDP in the market. Now, you add unstructured data. Within unstructured data, you want semantic understanding of that unstructured data, that's what intelligent context is. I want to know more about a really complex document. If somebody is eligible for a diabetes trial, for example. A lot of charts and a lot of flows, a lot of tables, you can't chunk that in a syntactical way. You have to chunk it in a semantic way, right? Now, the next maturity model on top of it is layer on top of it, personalization. I have the structured data. I know the patient's name. I know the past five visits the patient has had. What conditions a person's been through. Now, you combine those two. Now, you provide personalized experience through agents across structured and unstructured data, but what's most important thing here, George, is it's all governed and secure. That's the area where we are putting a lot of investment in, governance and security.
George Gilbert
>> The governance and security integrates both the data and the actions data?
Rahul Auradkar
>> Exactly, and also the agent. So, we just released Agent Fabric on MuleSoft, that drives agent-to-agent governance. Similar to API governance that we had in MuleSoft, now you can do agentic governance. Then, we have data governance inside of Data Cloud that allows you to tag data, that allows you to mask data inside of Data 360 and allows you to then have policy-based governance on top of it. That, when combined with agentic governance that you get from MuleSoft, then we have got the holistic experience from a maturity curve standpoint. Now, we can start going into industries that are very sensitive to this as well.>> Rahul, on that, I was saying this context is interesting. I'm sorry, George. I didn't mean to interrupt. You mentioned context. It sounds like data is super important, but it's not necessarily big is better because small context might trump large data quality old-school data thinking is getting more data. The answer is this. Just having the right amount of data, you can be statistically accurate but operationally wrong on data. So, take us through how AI will work there because what you laid out is that in these transitions between these decisions, the right data might not be the biggest data. So, talk about that dynamic. It's very nuanced, but everyone thinks large language model, which is great. They train it, but it gets bigger and there's a lot of context in there somewhere. How should people think about data quality? Because context might be small data or the right data because you're basically saying get the right data right there.
Rahul Auradkar
>> Yeah, it's an interesting thing. It's a needle in the haystack problem there, right? So, we believe in this idea known as data fluidity or fluidity, which data and context fluidity the ability for the context to be available for the agent in the moment for the right use case for the right outcomes they want from a customer. We have consistently done that through. So, one thing that we have done is we have built Data 360 from ground up on standards. We did the same thing and then we are an open and extensible platform. What it does do for us, it opens up an ecosystem. So, if you want to provide data fluidity, what you want is a vibrant ecosystem that's playing with you. I had Snowflake CPO, Christian Kleinerman, on my keynote stage about an hour back and he's our biggest partner. I mean there is no reason why his Data Cloud doesn't work with our Data 360. We always get this confused as in saying we are another warehouse, we are not. So, that sort of a ecosystem vibrancy is needed for data fluidity that allows us to provide intelligent context in the moment. But there's a lot of detail under the covers, as in how do you have the right metadata model? How do you tag all the metadata, so you have governance? How do you unify the metadata, so you have the right context? But to your point, it's not about unifying all of the data, it's the relevant data, so you know what the use case is and that is getting defined by agents. And we also have the significant investment we have made on NL2SQL, which was what George was bringing up earlier as well. Natural language to SQL because all of that comes down to some conversion into code.>> You're saying context is greater than quality?
Rahul Auradkar
>> Context is greater than quality. Quality helps with context, it gets better. They kind of play off each other.>> Okay, thank you. Sorry to interrupt. Go ahead.
George Gilbert
>> So, building on that, this is the core data foundation, but we were talking earlier about Data Cloud as a foundation for Agentforce is expanding not just unstructured data, but what I would call Datadog for agents. The information that you need to test the agents before you roll them out. All the traces and data that you would collect, essentially, the exhaust not of clickstreams from websites 15, 20 years ago, but now the exhaust from the agents. Explain how a data platform and its analytics changes in the age of agents to support this new programming artifact or programmable artifact.
Rahul Auradkar
>> Yeah, so in the early days we saw... We are still in early days when it's agentic transformation, agentic enterprise. What we saw was, as you saw from the figures from MIT, 95% of gen AI projects are failing. There's a lot of reasons behind it. First of all, we talked about what John just mentioned, which is the right context and right place. The other one is also the feedback, as in are we doing the right thing at the right time for the customer? How do you get feedback? This is your point about exhaust. What we are doing is we are bringing all of that exhaust and we are making that as part of the operational aspects of the agent and it requires for that exhaust also to have the right context. Think about it from a context that you're providing to the agent. Was that the right response? Was that the right action? Did the governance go right? Are we locking out people who were supposed to get access to the data? Are we locking out some actions that were supposed to be performed through governance? We keep getting that feedback. As a matter of fact, you saw William Sonoma, we released the sous chef agent for William Sonoma. Our teams in the background have been continuously monitoring it to learn from it. And now, you automate that learning by bringing that operational data into Data Cloud.>> Rahul, real quick, what's the governance equation outside of the Salesforce world? You mentioned ecosystem, Snowflake. How are you handling governance outside of Salesforce?
Rahul Auradkar
>> So, we respect governance of data coming in from external systems and we are able to respect that, but there was an announcement that we made for agentic enterprise search, for example. So, when you are sitting in an enterprise and you do an ask me anything. I want to know more about Nike account if a sales rep is looking at it. They want to know everything. They want to know about Google data. They want to know Slack data, the Google emails, et cetera. You can't search on every corpus. You have to search on the governed corpus. We are respecting that governance. We have brought that governance into our system. So, we not only respect that governance, we can overlay that governance using our governance framework on top of it as well. So, if anything, we are going slow to go fast, meaning we slowed ourselves down so that governance becomes an underpinning.>> You control what you can control. Control coming in, that way it's in the data there, that's-
Rahul Auradkar
>> That is Correct. You control what you can control and then we can override those controls or you can add more within when you're inside of our boundaries, if you may.
George Gilbert
>> On the topic of control and adding in the data engineer as a role. Now, that Informatica has close to closing, how will that affect the functionality, the power and simplicity of Data Cloud? Because right now it's very accessible with configuration-driven interface for the whole ingest to harmonization to unification activation. But when you have a product that powerful that knows that much about your data that you can do all the transformation on demand and master data. What's Data Cloud or Data 360 going to look like 6, 12 months out?
Rahul Auradkar
>> That's a great question, George. We are still going through the regulatory approvals post-close. We started the post-close planning, assuming we get regulatory approvals. We look at Informatica as being across three dimensions. One is data transparency, the other one is data understanding, and the third one is data governance. So, think about the end-to-end scenario where a customer is looking for, hey, it's not about the data that you have to activate your CRM apps or what you define as CRM, what you define as business apps. It's data that we have across the entire enterprise, which is your point as in they have a great catalog, they have fantastic metadata that our customers are telling us that that needs to be brought to bear. They have a governance, the catalog, they have fantastic data integration. We can have data integration and app integration now learn from each other for our customers. MDM, their MDM does stewardship, it does account mastering, account hierarchy with account and B2B and then person ID, unified ID that we do, combine all of that. Then, you get significant value for our customers across understanding, transparency, and also with the governance.
George Gilbert
>> So, how long will that take to start integrating it into Data Cloud?
Rahul Auradkar
>> So, we are just starting our post-integration planning. You should expect, realistically, we'll start seeing some integrated value right away from our two products, but closer to a year from now, you'll start seeing product integration. Maybe at next Dreamforce, I'll be having a different discussion with you.>> Rahul, talk about interoperability. I like this governance strategy you got. You said open standards. One of the big surprises this year that I'm super pumped for is the MCP organic growth. Obviously, it's early days, you got A2A out there. How are you looking at that piece of it? Agents have to talk to each other. We see the fleet of agents. What's the role of interoperability with, say, MCP? Is that part of the unification strategy? Take us through how you're thinking about that because that's a great path.
Rahul Auradkar
>> Yeah. So, our open and extensible strategy now got significantly better because of MCP. So, John, I think I said earlier that it helps to start from ground up and start late. We started late with Data 360, meaning we started five years back. When you start with standards up, anything that you do with extensible, anything that you do with data fluidity comes naturally to us and because that's what our customers are asking us. Now, with MCP, you can put that on steroids. I mean, you can do so much more that we are required to go find the right APIs to query into. If every source of data has an MCP head on it, now that makes it so much easier for us from our ability to bring in the context that you are referring too.>> All right. So, I have to connect the dots and ask you to define what... I should know this, but I admit I don't. What is AI Federation? What is that about? What is that role you're doing? Because trust is huge with MCP. APIs are great. You hit an API, there's no state, okay, REST APIs are great. When you start getting into state, what is the AI role, your new title, what does that mean?
Rahul Auradkar
>> Oh, what I have in my new title is AI Foundation.>> I'm sorry, I said Federation.
Rahul Auradkar
>> So, what it is when I started at the outset, I said that AI revolution is a data revolution. So, our point is that everything that we do, whether you take the traditional CDP, I call it traditional, even though it's just a few years old, with CDP and marketing automation and what CDP was doing for service sales, et cetera. All of those foundations are needed for agents to become smarter. We have those foundations. That's what we are investing in.>> Is it building blocks or this a framework or is it product?
Rahul Auradkar
>> It's actual product on which whatever we do with Data 360? Whatever we do with... For example, you have Tableau Semantics layer, right? The insights that you get with Tableau Next, those insights now are actionable using a concierge service that allows you to ask questions off of those insights. That's coming from the underlying Data 360 foundation.>> Okay, got it.
Rahul Auradkar
>> That's what I mean.>> All right. Thanks for clarifying that. Sorry, George. Go ahead. Continue.
George Gilbert
>> So, I'm curious about when you go into customer accounts, and a lot of people are talking about agents, but they're approaching it from many different directions. I think, from what I gather, most are saying, "I've got the ultimate set of tools." I kind of joke that's the Sean Penn and Fast Times at Ridgemont High scene when he cracks up the car. There's the tool-centric view, then there's the guys who are like, "Well, I've modeled my data in a data lake, I'm set to go," and they don't realize that the data model is entity snapshots and you've got a process. And then, you've got the data space and the action space all integrated into the tools. My question is, sort of roundabout question, what are the categories of customers and the relative sizes, buckets of how they're approaching their agent pilots? How many even go in thinking, "This is an outcome I need to achieve. What do I need to support it? Where am I going to get that support?"
Rahul Auradkar
>> That's a great question. Let's take two ends of the spectrum on this, since you refer to the types, right? One end of this spectrum is, I can talk about this eloquently because Salesforce on Salesforce is our customer zero. That's been our first and best customer on Data 360 and Agentforce. Michael Andrew just did a keynote here with me. So, if you take a look at the help agent that we have right now, which is answering 1.8 million customer queries, how did this start? We didn't start with, "Hey, we have got the right set of tools. We have the right set of agents, we can build with the tools." It started with the business use case. It started with the outcomes that desired. We wanted to get the right questions answered for questions of our customers, so they don't open cases. They get the responses fast enough, they get actions fast enough. We have answered 1.8 million questions and it's been less than eight months we have that live, right? Now, underneath the covers, we started realizing that we need unstructured data close to, what, I think I lost count of it, 600,000 documents that have been modeled. And then, you have all the structured data that's inside of our customer-zero instance. Now, why is that important? It's important because we know the questions our customers are asking. We know the customer sat service that we have. That's defining what we do there, it's not the tools underneath. Let me take the other end of the spectrum. Other end of the spectrum is yesterday I was talking to Romi Chopra. He's the CEO of a company called MIMIT Healthcare. It's a small physician's office and they have a small footprint of Data 360. They got the foundations right. They're saving over $1.5 million a year just with a small footprint. Why are they saving the money? Because they now know how to reach the patients consistently more proactively. They don't have to wait. They can work with their ecosystem partners. That foundation became now so much easier. Now, he's about to deploy Agentforce agents on top of it, that's the other end of the spectrum. It's not about the tools, it's not about the platform. All of those are great because we have the best of them in the industry. It's about the business outcomes, the business use cases, the business stakeholders. That's where we are starting.
George Gilbert
>> Yeah. But what portion of customers are still thinking, "I've got the ultimate set of tools, I can fix it," and they're going in with one of 1,000 different standalone agent development tools? In other words, how are you starting to reach them?
Rahul Auradkar
>> Yeah. Look, I think, George, one thing is that we know there's heterogeneity out there. There's no doubt about it. We have no illusions about the fact that our customers are not building agents in other platforms. It exists out there. We are trying to build the best agents on top of Agentforce. And then, we are driving data and agent governance using Data 360 and Agent Fabric. When you have agents that may not have been built on Agentforce that need to coexist and need to talk to Agentforce agents. So, we are making our agents the best that can be out there, and then we can always interoperate and to what John was saying, A2A and MCP, with A2A, we can work with other agents. We have shown that it can work. We have memory, for example agentic memory or conversational memory. We can move back and forth between agents and we can do it in a governed may with Agent Fabric.>> Yeah, and that helps with the agent trajectory of having a concept that turns into transactions. To me, I could book a hotel and hit a payment rail. That's the holy grail. We got to stop there. You guys could pick this up on the next podcast with George, which by the way, congratulations. Those are really one of our top-viewed podcasts. People love that content. Rahul, thank you for your time. I guess my final question for you to wrap up is what's on your to-do? What are you optimizing for? What are you looking to accomplish over the next 6 to 12 months? What's on the agenda?
Rahul Auradkar
>> John, great question. So, I keep saying this, I repeat it until I get hoarse, is customer outcomes, customer success is the best currency we have. We should continue building on that. We are in the early days of what we have got here. For us, volume and data gravity is not an issue. It is not what we are going after. We are going after data fluidity that drives customer outcomes. So, what's on top of my mind is customer outcomes, customer success, customer consumption, and what ROI they're getting, and I track that on a regular basis with our customers. You are much better off having 10 customers who are successful than 100 who are still struggling, that's because we are very early in this game.>> Rahul, the outcome of this interview is a lot of insights. Thanks to you, the fluid conversation, data fluidity. We're fluiding out to you here on theCUBE, doing our part, sharing the data with you here at Dreamforce '25. I'm John Furrier, George Gilbert, breaking down the future as this early days sets the table for the foundations and the building blocks that get developers productive, putting data into action. As agents come on, you're going to need to have context, high-quality data to make it all work. Stay with us more after this short break.
George Gilbert