How Salesforce Is Using AI to Power the Enterprise

This episode contains sponsored content in partnership with Salesforce. At Dreamforce 2025, Every CEO Dan Shipper sat down with Silvio Savarese, chief AI scientist at Salesforce, to discuss how one of the world’s largest software companies is shaping the future of AI for the enterprise. Together, Dan and Savarese explore how his team at Salesforce develops AI solutions that now power more than 13,000 businesses—including OpenAI, Dell, and FedEx—helping them become truly Agentic Enterprises that operate with greater scale, speed, and precision. Examples include a large language model built for Salesforce developers years before ChatGPT’s release, and Agentforce, the company’s agentic layer that enables a hybrid future of work where humans and AI agents collaborate to achieve more than either could alone. They also discuss how Agentforce gives enterprises a deeply unified AI platform that connects their data with agent functionality—making it both powerful and practical. The conversation touches on how Salesforce builds trust with enterprise customers amid the jagged frontier of AI by ensuring consistency in results, while continuing to push the boundaries of what agents can do autonomously. Savarese shares how enterprise-grade simulation environments help them strike that balance, and reflects on how AI agents will ultimately transform how businesses and individuals alike get things done. @Salesforce #SalesforcePartner #DF25 Want even more? Sign up for Every to unlock our ultimate guide to prompting ChatGPT here: https://every.ck.page/ultimate-guide-to-prompting-chatgpt. It’s usually only for paying subscribers, but you can get it here for free. To hear more from Dan Shipper: Subscribe to Every: https://every.to/subscribe Follow him on X: https://twitter.com/danshipper Timestamps: 00:00 – Start 01:16 – Inside Salesforce’s early AI innovations 02:50 – How Agentforce works and what it can do 07:03 – The real challenges of deploying AI at scale 08:57 – Why Salesforce builds simulation environments for AI 12:35 – The future of agents and enterprise AI

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:15Speaker 00

  Silvio, bienvenido al programa. Gracias. Así que para las personas que no saben, eres el principal científico en AI en Salesforce. ¿Qué significa eso exactamente? Sí, el principal científico es un rol que empezamos cuando Ariesch fue fundado hace alrededor de 10 años. Y el objetivo es que yo supervise y mire cómo está evolucionando el AI en una empresa, cómo puede hacer AI para crear algunas transformaciones, innovaciones para el tipo de negocio, el tipo de productos que Salesforce está produciendo.

• 0:52Speaker 00

  Así que realmente nos enfocamos en la innovación central, la investigación central que eventualmente se puede implementar en algunos de los productos clave que Salesforce está produciendo. ¿Qué sería un ejemplo de eso? ¿Están haciendo investigación en el modelo de base? ¿Están haciendo fine tuning? ¿Qué tipo de cosas están haciendo en el frente de investigación? Sí, algunos de los conceptos e ideas de investigación que desarrollamos fueron realmente en la base de algunos de los productos clave. Y uno de los primeros ejemplos fueron los modelos de lenguaje grandes que construimos para los desarrolladores. Así que era un lenguaje para código y fue antes de que GPT se presentara.

• 1:37Speaker 00

  Así que fue al principio, casi como hace cuatro años, empezamos a construir este tipo de modelos completamente en casa. Y se usó para mejorar la eficiencia de nuestros desarrolladores y la oportunidad de construir código a escalas más rápidas con mayor eficiencia. Y eventualmente esto se convirtió en un producto de G A hace unos años. Entonces, otro ejemplo es los modelos que construimos para el trazo. Así que Agent Force tiene un componente que permite...

• 2:10Speaker 00

  To verify that the output of those of AI is safe to use for customers. And some of the components that are the foundation for the trust layer were built in-house from from our organization. Um. A third example is the Atlas Reasoning Engine, which was introduced last year. So when Agent Force was announced to the public, it was equipped with a reasoning engine, which was incubated within my organization. And it was done in collaboration with customers through a number of partnerships we had over the years. Got it. And so for people who are listening, who maybe haven't interacted that much with the Salesforce ecosystem,

• 2:51Speaker 00

  Agent Force is essentially like your chatbot builder, no code or low code application builder for your enterprise customer. Yeah, I would say Agent Force is more like a platform for agentic systems. So it's it's where our customers can use, deploy agents for their business. And it's it's almost like thinking about portfolio of agents, each specializing in different tasks. And those can be used for scaling up and for making their own business more efficient, more effective, more productive. Got it. How do you position that against like, I don't know, if someone's thinking about, I want to make a custom GPT, should I make a custom GPT or should I build something in in Agent Force?

• 3:30Speaker 00

  Like, when should people use OpenAI directly or Anthropic directly versus going with a Salesforce model? Yeah, I would say that, let me actually back up a little bit with this question. A great question. In a way, we have to think that agents is not just an L L M. Agents is much more, it's a complex systems which has many various components. As a memory, so which allows you to retrieve information and data, documents, fuse conversations. And typically, the techniques are you, they're relevant, are rag, the retrieval augmented generations or other type of appendix that are used for for enabling, empowering this memory. The other component is the brain. So it's a reasoning engine that allows you to

• 4:13Speaker 00

  um turn a certain query, a certain task in a series of steps. Uh, and that's that's the the. The innovation that I mentioned earlier, so the ability to to reason about a certain um certain task that has to be performed, the agents. A third component is is the actuators. So how to turn those tasks into actions? So essentially, our API calls, function calls that can be uh invoked by the agents. It's similar, it's a bit like a toolkit of functions. Let's say, send email or, you know, open this website or, you know, make these reservations. So these are actions. And then finally, the interface, how the agents communicate with the external world through voice, through text, through different modalities. So these four components creates an agent.

• 4:57Speaker 00

  And and um and as I mentioned, it's not just an L M. So in a way, it need to really provide the right uh context, the right uh and uh, you know, connect better these different components to build uh a full solution can be helpful for enterprise. So uh back to your questions, you cannot just use L M for notifications. You have to build all the infrastructure around and provide the right context for for making those uh uh agents function uh functional.

• 5:25Dan Shipper?

  Got it. And and so uh thinking about that, like, if I have the option, for example, to uh, you know, chat G P T has like an has an agent builder or uh, so if I have the option to use both uh and presumably I can like pull in Salesforce ecosystem stuff into chat G P T. How would I make the decision to like use a Salesforce specific agent builder versus the other ones?

• 5:49Speaker 00

  Yeah, so I think uh agent builder is fundamentally useful if you want to um connect the dots between the data that customers have and the type of functionality that can be implemented by the by the agents. Uh, you know, an agent uh without data, without the context, it's kind of useless. So then if you just have an agent without plugged into the rest of the infrastructure, it's gonna be very difficult to implement. That's why uh the the the agent builder, the agent builder, an agent force allows you to give all the ingredients for make those agents uh successful in the in the this kind of context.

• 6:25Dan Shipper?

  It's been really interesting to be here. We're recording at Dreamforce, which is Salesforce's biggest conference. And it's been really interesting to be here because I spend most of my time at uh, you know, OpenAI or Anthropic or uh or with small startups. And watching the keynote today, for example, there was like the, you know, people high up executives at Williams and Sonoma and Pepsi and you're starting, I'm seeing the way that AI is talked about and being deployed at big brands that people have heard of, which is just a bit different from the, you know, startup or foundation model company type stuff that I hear about. What are the, what are the unique challenges that Salesforce has to contend

• 7:07Dan Shipper?

  with to help these kinds of companies do AI deployments? Like what is actually happening inside of these big companies as they try to deploy AI?

• 7:14Speaker 00

  Yeah. So I think first of all, it's going to connect to what we were seeing earlier. So connect them to the right data and be able to provide the right harmonizations across data to ensure that different touchpoints are connected because often customers have siloed data or siloed functions.

• 7:32Dan Shipper?

  What's a specific example of that?

• 7:34Speaker 00

  So for instance, you know, you would have a division that sales, a sales division that has a way, a certain way of managing data. Then you have the marketing division, which are in different way. So connecting dots is very important. If you don't connect the dots, if you don't have the right level of harmonizations across data, AI cannot operate well. That's one aspect. Second aspect, which is very important is about trust. So customers are very concerned about the, the quality, accuracy of AI. So the kind of if we are safe to use from the, from the trust perspective. And in a way, we've seen that many of those L M are sometimes are producing what you call jagged intelligence. So they produce performances which goes to very, you know, amazing peaks. They can prove theorems, they can become

• 8:18Speaker 00

  gold medalists, but then they can also fail in some tasks which are very, very to common sense.

• 8:23Dan Shipper?

  Yeah.

• 8:23Speaker 00

  And unfortunately, in many situations, actually, these common sense tasks are relevant for business. Imagine that you're asking an agent to get a loan and you're get an answer from a loan agent and the loan agency is coming up with wrong computations or something that is, you know, doesn't make sense. It's going to be a big, big bummer, you know, for the customer and also for the bank or for the financial institution. So ensuring consistency and at the same time also push the Boundaries of those agents can do is very, very important. And that's where, you know, we uh it's useful really trying to uh ensure that the the the quality is matched with the capabilities.

• 8:57Dan Shipper?

  I know in order to do that, you're building simulation environments. Can you tell me about why that is and what kinds of simulation environments you're you're making?

• 9:04Speaker 00

  Yeah, so simulation environments uh for us is a key for uh uh enabling this kind of trust and this kind of quality that I just mentioned earlier. Um unfortunately, you know, we are getting a situation in the scanning law of learning. So uh those uh even G P T five is not, we have not seen a really breakthrough that uh, you know, some people uh it was imagining at the beginning. Um and we still see this kind of jaggedness, this kind of inconsistency in some of the results. So uh we have to move into new paradigm of learning based on experience, based on uh, you know, see how these agents uh behave in the real world. And the best way to do that is plugging them into simulation environments. Simulation environments essentially allows us to stress test the agents and understand when they fail.

• 9:46Speaker 00

  Based on uh on this kind of awareness, you can provide feedback to uh improve the agent performances. And these are enterprise uh environment, it's not like uh game environment, it's not like uh uh formal one or other raising environments. They are are actually simulation of C R M or enterprise environments with their own fields, their own objects. Uh and we simulate, for instance, for agent force, we have been start simulating a voice, voice agents uh by um uh testing them with a number of uh different type of voices. You know, voices with noise, background noise, uh with different accent, like sometimes with my Italian accent. Uh how how the agent will react to this kind of situations or maybe conflicting uh request or incomplete request.

• 10:29Speaker 00

  These are all uh corner cases and uh that that makes agents tend to fail and that's why we want to make them super robust.

• 10:36Dan Shipper?

  How do you make an environment for that there? Like, what does that actually look like?

• 10:39Speaker 00

  It's a, it's a, yeah, it's it's an environment where you have to simulate um organizations, B to B, B to C. And we start feeding them with the data, with synthetic data. And this data has to be also ma- mimic real world data or can be validated by experts. And the challenge here is to uh to do uh to fill the so-called reality gap, the gap between the simulations and the reality. That's actually uh part of the work in progress.

• 11:04Dan Shipper?

  Got it. And so like.

• 11:07Dan Shipper?

  Is is the is the simulation environment that you're talking about, is it like uh uh a computer that the that the L M can use with a bunch of simulated data about what what kind of company it's working inside of, like, give me a little bit more, yeah?

• 11:23Speaker 00

  So these are essentially a simulation organization, let's say, you know, in um you're simulating a service um uh customer service where you're uh simulating the way how the agents will uh talk to uh a human, in this case, you simulate the human as well, uh and then the the environment will have different um uh components which which allows you to pull out data uh for answering this kind of customer service questions, let's say about return, about, you know, product availability, about policies, this is all simulated or, you know, built based on the real information that customers are giving us, and this allows us to kind of be more almost like um gameplay where

• 12:05Speaker 00

  uh the agents plays with another of, you know, fake human, and then they they they measure potentiality of of failure.

• 12:12Dan Shipper?

  Interesting. Uh so we're almost out of time, we have time for one more question. Um, one thing I'm curious about is you've been in the field for a long time. Um, obviously, the rate of progress has been pretty high for the last like couple of years. Where do you think we'll be in the next in the next couple of years, uh in AI in general and um in AI as it's being used by businesses?

• 12:35Speaker 00

  Yeah, I think the main innovation will come when we're gonna see agents deploy to scale. Uh soon, we're gonna have um, we're gonna be having our own personal agents, and personal agents will be will perform a task on our behalf. Uh so if I want to, let's say, back to the example of the loan, I want to get a loan, I won't call the bank myself, I can ask my agents to call the bank, uh and then there's gonna be interaction between my agents, my agents and agents banks. So we're gonna be, we have to imagine a future where agents to talk to each other, we need to define protocol of communications, uh what are the guidelines, you know, and uh it's gonna be almost like thinking about uh the internet before the internet, right? Uh so, but but the world society will completely transform, and it will be very exciting.

• 13:16Dan Shipper?

  Awesome, sounds exciting. Sylvia, thank you.

• 13:18Speaker 00

  Yeah, thank you, thank you for your question.

• 13:28Speaker 02

  Oh my gosh, folks. You absolutely, positively have to smash that like button and subscribe to AI and I. Why? Because this show is the epitome of awesomeness. It's like finding a treasure chest in your backyard. But instead of gold, it's filled with pure, unadulterated knowledge bombs about ChatGPT. Every episode is a roller coaster of emotions, insights, and laughter that will leave you on the edge of your seat, craving for more. It's not just a show. It's a journey into the future with Dan Shipper as the captain of the spaceship. So do yourself a favor. Hit like, smash subscribe, and strap in for the ride of your life. And now, without any further ado, let me just say, Dan, I'm absolutely, hopelessly in love with you.