Welcome back, everyone, to theCUBE here, live in San Francisco for Dreamforce 2025. I'm John Furrier, host of theCUBE with George Gilbert, head of our research for the data platforms of our CUBE Research team. Breaking down all the action.

The theme is agentic, and all the branding, all the messaging is pointing to the future that AI infrastructure is enabling a wave of agentic systems, of course. The person here to break it down, Ariel Kelman, president and chief marketing officer of Salesforce. Ariel, great to see you again.

Yeah. Thanks for coming on.

Thank you for having me here, guys. Welcome to Dreamforce. You've seen many waves of innovation, and the agentic AI everywhere story, which is one of the themes here, is real and it's happening. We're in the show-me moment of the hype cycle, where the proof is in the pudding, the meat is on the bone. If you don't have it... It's happening fast, so, if you can't show it.

It's amazing. It's only a year ago that we introduced Agentforce. We had five beta customers that were sharing their stories. We were using it internally for powering Dreamforce, so you could ask questions. But in just a year, the whole industry has grown up with this capability. And now if you go down to our campground, there's 17 of our customers. We've actually built a whole city where there are companies, like Pandora, and Equinix, and Lululemon, and Pepsi, Lennar Homes, Heathrow, are showing how they're using Agentforce live in production in their business. And I think you've seen over the past year, there's a lot of focus on, "Look at our flashy demo. Isn't this cool? Look what we can do." But we've been methodically iterating with our customers to work through the difference between something that can demo as an agent, and then how do you roll out agentic AI in the enterprise with all the complexity of actually making it work? Because what's kind of funny is it's not that hard to make an agent work 95% of the time, but that's not good enough if you're an actual business who's trusting this agent with your customers. So, if you look at all the functionality we rolled out this week, we can talk about that, a lot of it's really focused on making agents real in production.

It's interesting. You guys did something that you don't see very often at all, which is you released stuff before the conference. Most people release stuff at the conference and then now people come with working examples.

Yeah, we did it last year too, and it's not the conventional marketing strategy, but unsurprisingly, we're optimizing on customer success because we got 45,000 customers coming here. And as they want to plan their week, it's kind of useful for us to tell them what's all the new stuff so we don't have to hide the sessions. I mean, this gets a little bit into the inside baseball, but if you're not going to announce something until day two, you can't have sessions on it before, you can't have workshops and you've hidden it so people can't see it. That's all just kind of crazy.

We've known each other for over a decade, Ariel, and following your career, theCUBE and our CUBE Research has been really digging in past three years, watching the transformation. It's actually unreal to look at the future. It's a different company now and I think it's misunderstood. And that's a good thing, I guess if people don't understand it, but you guys are powering away. As you guys look at this platform that's now out there, it seems like a new category. We were talking on theCUBE yesterday in our closing segment that Salesforce has evolved so much. It's a true agent platform. A ton of data, that's a competitive advantage. What's your views on this? You're a marketer positioning, you got the brand and you got the demand gen. You've seen the movies before. It's disruptive. It's new.

Yep. I think where we're heading is there's going to be a category. You'll see things like Magic Quadrants of an agentic AI platform. As companies increase, they're going to look on standardizing on one or maybe two platforms to be able to create data-driven agents that drive effective accurate AI drive useful results and also do a combination of both elevating the capabilities of their human employees, but then also augmenting their workforce with digital employees. And this category's changed. It actually doesn't exist. If you look at the analyst reports right now, because you've had a lot of focus on the two ends of small niche startups that are going to do AI agents for one particular use case or people using lower level developer tools, cloud service providers to DIY your own agents from very granular developer services. And I think where the industry's heading is platforms for building agents that abstract away a lot of the undifferentiated heavy lifting, a lot of the really hard parts, like how do you make agents secure? You can't just take all your enterprise data, throw it in an LLM and expect it to respect all your security rules, your governance, your workflow. And even something simple like, how do you make sure your agent never tells your salesperson to sell product 47 to the state of Idaho because it's illegal to sell there? If that works 99.8% of the time, that's not good enough. So, that's why we've built in deterministic logic into our platform and that hard aspects of really making it production ready is what we've been focused on.

So, I want to follow on John's question and the way you frame this agentic platform. There is a lot of confusion out there. Some companies have these awesome agent development tools, some companies have data platforms, some companies have two of those things, some companies have applications. What is it about putting the pieces together that's more than the sum of the parts?

Well, it's always been the philosophy at Salesforce, to provide people a deeply-unified platform, so things work well together. So, if you go add a new database object in our Data Cloud, having it show up in a screen or having it be available in a prompt-building template for your agents, it's all hooked together automatically. But at the same time, I think it's also important to say that it's really important for our platform to be open. So, for example, you have companies that have standardized and built data lakes with Databricks, with AWS's technology, with Snowflake. And so, our Data Cloud supports zero-copy technology, so you can have federated data access, but Data Cloud's still providing that intelligence, that activation layer that makes your data AI-ready.

So, you've got the Customer 360 application surfaces and then a value-add layer on top of your existing data platform. And when I say value-add on top of, I mean because not replacing Databricks or Snowflake, but for the customer relating it.

We're activating the data they have.

Yes, and adding context to the customer-related data. Just to be clear, and it's that richer context about the customer engagement journey that you couldn't do by hand in a.

Yeah, because if you're going to have customer-facing agents or agents that are employee-facing related to sales marketing support, you're going to get worthless outputs if they're not grounded and very detailed data about the customer, what they've bought before, do you have an active sales cycle? Do they have a red customer service issue? If they do, don't go promote the new products. There's so many ways that you really can risk the whole brand reputation of a company. Sort of like the surface area for brand compliance, if I speak as a marketer, has magnified and you have to make sure that these agents are governed in the right way and we take care all that hard part with our agentic platform, with Agentforce.

We're activating the data they have.

Yes, and adding context to the customer-related data. Just to be clear, and it's that richer context about the customer engagement journey that you couldn't do by hand in a.

Okay, you added another layer to what I was talking about with the contextual data and the actions to operationalize an agent decision, but then it's uniform governance of the data and the actions as well. So, that in other words, if a customer is trying to pilot. Because they've got a directive from their CEO to, "Let's do something with agents," but if they start on your platform and they have a customer-related business outcome they want to change, then they're going to get there a lot faster. Maybe tell us some of the customers where they were 12 months ago and as the product has matured, how much faster it's getting them closer to production?

In general, we've had some analyst studies comparing building agents on our platform versus do-it-yourself, and we found it to be 15 times faster. But the experience of a lot of our customers, and you'll hear some of the stories in the keynotes of people building these in a couple months. One of the things I just did in the event yesterday with 45 CMOs, and we were showing them the website agent that we built, so that people can go ask questions and so we can essentially have a digital salesperson on our website. So, that overall has led us to a 15% increase in the pipeline that we create for our salespeople from the website. And what I was telling everyone who joined the event is the tools have become so robust now that, you go build that in a couple of weeks, especially with this new context intelligence feature where we are making the unstructured data aspects of this a lot easier. Because before, you had spent a lot of time getting the data and having it chunked correctly, making sure it's stored in the vector database in the right way, we've automated all that. So, you just load PDFs in and it understands the context, even the very complicated documents. So, getting a website agent that really puts your best salesperson on your website, you can do it in a couple of weeks now. It's really amazing.

Yeah. Ariel, one of the things that we've seen with AI clearly in the first couple of years was coding and marketing and sales materials are the two low-hanging use cases. I asked Rob this on the chief product officer side of Slack, the same question. How has AI and you guys transformed your marketing? Because you guys do a lot of customer-zero, you eat your own dog food, drink your own champagne. How are you using the AI across the Salesforce estate to be better?

Yeah, we're using it in a lot of ways. I mean, starting with my team, I said we have the website agent. We also have an SDR agent that follows up on leads. So, think about this of having infinite capacity of your best SDR to go after leads. So, before we had this, did a little math and over the past 10 years we didn't follow up on 20 million leads.

That's a lot of leads.

And that's super normal because you score them and route the highest ones, but now we follow up on every lead. And in just three months, it generated $60 million of incremental annualized pipeline and we're just barely rolling it out.

That's a great stat.

Our customer support agent has really been effective, over almost 2 million conversations with our customers and very high-resolution rates. We're doing, this may sound like a simple one, but an employee-facing sales agent that helps prep our salespeople for calls. So, we access this through Slack and I say, "I'm going to go meet with FedEx. Give me an account briefing." And it's going to go take all the data from Salesforce and from third-party data sources, from news articles and go and give me a briefing that normally you'd spend a lot of time on.

George and I were talking yesterday around how in the old ways of disruption, the new guard replaces the old guard. And what's interesting is you guys are at scale, 20 million leads, unbelievable success at scale, but you're creating a category. It's like you don't see this very often. So.

I think the best companies, they disrupt themselves.

Yeah, share what's happened, the formula, because the branding now is changing too. So, I want to get your thoughts on as the company evolves, the company's been transforming in real time at scale.

Agentforce has completely transformed our company and it's completely transformed our products. So, if you saw in the keynote, we've renamed and relaunched our core products as Agentforce Sales, Agentforce Marketing, Agentforce Commerce, because we built agentic experiences deeply into both the employee-facing user interfaces. And then, all of our products now have direct customer-facing interfaces where the agent's talking directly to the customer. It completely changes how you use our products in a really great way.

Just to be clear, did you rename the Customer 360 apps or. Oh, that I didn't catch.

So, if you go to our website now, it's not Sales Cloud, it's Agentforce Sales.

Oh, I missed that.

You're rebranding everything Agentforce?

Yeah.

Okay. So, maybe how.

We do like changing the names of our product to shake things up a little.

The titles have changed too.

All right. Well, I got to wrap my head around that one. But how do you want your customers to perceive you differently now that. I mean the agentic platform is something that I think IT would make sense of.

Well, it's interesting because in some ways we want our customers to think about us in the same way. We're about helping our customers connect with their customers in new ways and to be effective at working with their customers, whether it's selling, marketing, customer support. So, if you're in one of those business functions, it's about trusting Salesforce to give you the latest technologies to do the best job in the world at selling marketing supports, selling on your website with commerce. But for IT departments, I think the use cases for Salesforce are really expanding rapidly because you have all these IT departments that have been experimenting with AI agents. An MIT report said recently 95% of AI projects have failed. No surprise, I'm sure we've all had our own projects too. And so, IT organizations are moving from that experimentation phase to actual, "Hey, I got to solve this list of problems for my business users of all these agent use cases." And they're finding out that Salesforce is really an amazing solution here because Agentforce is giving them the fastest way to make highly-accurate agents that they can trust. I just came out of the Agentforce keynote. We had Samir, who's the CTO at Williams-Sonoma, talk about how his team is using Agentforce to completely reinvent their website experience to really bring the magic of going to a store at Williams-Sonoma directly to their website with their Agent Olive.

So, following up on that, this is what I wanted to unravel a little more, which is it's helping your customers connect to their customers. But it seems like when you wanted to add agents, you needed a richer data foundation, and with the richer data foundation you had deeper integration between the apps. And so, it's almost like.

Which, by the way, you always needed for good CRM too.

Yes, but it got deeper. And so, now it's easier to provide customers a deeper end-to-end experience, not just within what was once a departmental function.

Because with agents, the departments blend away. So, think about this, and we've learned this ourselves. Everyone's learning here, this agent thing's new. So, at first, we developed our customer support agent on help.salesforce.com and my team was going and developing our website agent. And then, someone asked this simple question, it sounds obvious, "Well, if someone is having a support conversation and they're asking about what a product does, shouldn't it give the best explanation?" And then, if someone's on the website and they say, "Well, how do I do an Apex code post-trigger?" It shouldn't be dumb about that, it should give you a good answer. And so, we've unified the knowledge base that both of them have access to, so you get smart answers about different topics.

I saw this happen 30 years ago when SAP was taking off in the US, where they put in finance or sales or distribution, and then they saw that there was so much value to add all these adjacent modules. And what I'm wondering is now that there's so much richer integration, just that scenario you mentioned, you're answering support questions, but those turn into product questions. Does that not now make the whole platform more valuable and there's.

Yeah, I mean it just really highlights how having well-built-out, well-structured and clean customer data, what some of us call our customer data platform, like what a lot of our customers use our Data 360 product, including ourselves, having that rich data, it really is the fundamental foundation to do agentic AI, but also to really manage customer relationships at scale. Just this example of if you're having marketing agents giving people offers, you don't want to give these marketing offers to customers that are having really bad customer service issues.

And reinforcing bad behavior.

Yeah, just like you'd want to tell a sales rep not to go and do that in the pre-agentic world.

That's the old ad tech.

So, it's like the stuff we're good at and the stuff our customers have trusted us for and they've used our platform for, they're familiar with, but now it's more important than ever.

Does the IT service management product start to give you more street cred within IT, so that if they're tasked with these ill-defined agent development pilots, but if you have IT service management, you have more and more street cred on the IT side of the house. And then, you already have the operational applications, so it's easier to bridge the two and get these end-to-end outcomes?

I mean we've had IT customers running our projects in a lot of companies for a long time. It's actually split. Sometimes it's sales ops or service ops, sometimes it's IT. So, we've been deeply embedded in IT for a while. And one thing they've been asking literally for 20 years is, "When are you going to have an IT service product?" And so, I can't tell you how many CIOs we've gotten text message from, they're like, "Holy shit. You guys finally did it. This is awesome. I can't wait to use it." And I think this is be.

This is the transformation.

... one of our next hip products?

Yeah, this is the transformation that I love about the company. And I want to ask you to end the segment. Salesforce always had its own little ecosystem of trailblazers. And it was in the app and it had a good ecosystem, people were building on it, huge success. Now, as you come more with a platform, you have an ecosystem developing. So, you're seeing that clearly with IT buyers, it's more integrated in the system. What is the vision on the ecosystem of Salesforce as you guys continue to grow outside of Salesforce? What's the plan.

There's a lot of ecosystem opportunities. I think for our ISV partners, there's never been more excitement because you can think about our partners building agents that run on our platform, just like they used to put apps on the AppExchange and they can even build out skills. So, let's say you want to be the expert in sales tax in Latin America and you've gone and coded all that complexity and add API connections, you can build that as a skill and then all of our developers can go use that in their applications and you can have a great business.

Yeah, what people don't understand is that your customers have the data and that's the moat. Ariel, thanks for coming on. Congratulations.

Absolutely.

A great show. Final question, what's your takeaway? Are you happy? I mean, everything looks beautiful. I mean, you shut down Howard Street.

I've been mostly checking on if the customers are happy and people seem pretty pleased. We got a lot of energy. Our keynote was standing room only. Our sessions are full. It's amazing, over 1,500 sessions.

It's packed. Great branding. Congratulations.

Yeah. Thank you.

Thanks for having us. Appreciate it. I'm John Furrier with George Gilbert here at Dreamforce, streaming live. Thanks for watching. We'll be right back.

Get Our Mobile App

Ariel Kelman, Salesforce | Dreamforce 2025

Transcription