In this Salesforce‑presented session, Greg and Ali walk through the company’s experience as “customer zero” for an agentic‑enterprise platform called AgentForce, which is designed to scale seller enablement and accelerate new‑hire onboarding. They first outline the growing complexity of modern sales roles and the need for adaptive, personalized enablement that can deliver the right knowledge, practice, and feedback at scale. The presenters describe how AgentForce integrates Salesforce’s knowledge base, Data Cloud, and the Atlas Reasoning Engine to provide AI‑driven next‑best‑action recommendations, cross‑sell insights, and real‑time coaching within the flow of work (Slack, CRM, etc.). A live demo shows an account executive rehearsing a renewal pitch for Cengage Learning, receiving AI‑generated feedback, and being prompted with actionable suggestions such as attaching a Data Cloud course. The demo also contrasts early, generic machine‑learning messages with the current, context‑aware recommendations. Greg emphasizes the hidden human expertise that fuels the system—curated product and sales knowledge that is ingested into the platform, enabling agents to surface institutional truth across multiple channels. The session concludes by summarizing the four‑stage loop (knowledge → practice → application → feedback) and highlighting how this loop continuously refines both the AI models and the seller’s performance.
Chapters
All right, so let's just jump right on in. So we're going to talk really about our journey as customer zero, focusing on agentic enterprise and bringing agents into distribution. In this case, onboarding for AES. I don't think it's any surprise, as you guys look around and see all the booths here, that the world has become extremely complicated. So as part of that, the demand on sellers is extremely high. On day one, you don't just jump into a here's your product and go approach. We have a lot of different needs that we need to teach a seller from onboarding. What's your product? Who's your customer? How do you do things the Salesforce way? All of these have to come together. And to really illustrate that point, this is basically just Salesforce's version of our product today. Now, I probably should put an asterisk on this slide because quite frankly, this probably changes every month because we keep finding new values, especially with Agentforce cruising along so well. But this is the reality of a seller's role today. As you join on day one and you say, hey, we have all these clouds we need you to focus on. We need you to figure out the best way to do this discovery with your customers. And then we have to get you. This is the goal of the field readiness team, is to enable you so you can do the best version of that job. So today we have a solution which is probably anyone that works in enablement is very familiar where we do have a path to deliver this. We want to make sure that we hit the needs of the business, the needs of the customers, the needs of the product, and we develop amazing content, we bring it to groups and we have this at scale, delivered to enterprise level wide. I mean, Salesforce itself, we do over 40,000 learners all the time, big pieces of enablement, getting people the things that they need. So this is great. Now to do this though. On the previous slide, this is where agents come in. Is this where scale starts to hit? The complexity of those three areas start to really balloon the personalization and focus on how can I get the most impactful enablement at that time. So what we're shifting to is more of an adaptive enablement. So what this means now is how can we use agents throughout our enablement process so that we can essentially go and take our key source of truth, our knowledge, and scale that across everything. The main goal, though is we want to first show you a little bit how we bring this to life via sales coach. So what we've done is we've gone through, deployed this in some pilot groups, some onboarding. And what's our result here? Right. We basically now can bring our expertise to a person where they are. They can either have a coach that learns with them, they can have a mentor, they can have a peer, they can have a buddy. So this is what's super important. It's no longer a one size fits all approach. It is now. How do you need to learn the best possible way and that allows agents to bring that to you. So with that, I'm going to hand it over here to see it live in action on the demo.
Perfect. Thank you so much, Greg. Now let's jump to the demo. So I want to take you to a day to day life of an account executive at Salesforce. And it's like checking with customers, checking emails, checking Slack messages and then boom, they receive a message and let's look at it. The message is that there is a renewal opportunity coming up with Cengage Learning one of our customers on Tableau Next, one of our products and it says that you have a renewal coming up for Tableau next with Cengage Learning on November 15. The renewal amount is $150,000. Before calling the customer, practice the pitch with Agent Force and these are the next best action that our AI model predicted here. So launch AgentForce sales coach and rehearse your pitch, refine your pitch and make sure that you incorporate the feedback you get from the coach and then a schedule, a call with the customer. And the last person, person of record, the last person of contact was Mr. Pete Griffiths. Keep that in mind and let's look at what happens when I click on practice with agent. I'll come back to this a little bit later. So now I click on it, it takes me to this page. Then I have my opportunity page. I can see this specific opportunity associated with this customer and now I'm going to click start, I'm going to click the scenario and now I'm going to go through the pitch. Okay. Most teams are data rich and insights poor. Dashboards are everywhere but decisions wait because data lives in silos and AI isn't grounded in the business. What leaders want is action in the flow of work. Tabletnext is analytics that acts built cloud for zero copy access to snowflake databricks bigqueries and applies a semantic layer so pipeline health and churn risk means same thing everywhere and brings insight right to a slack and CRM with agent Force skills. Now concierge answers business terms, inspectors proactively alerts and opens the exact slice that you need. And datapro accelerates governed logic for analysis. It results in trusted contextual answers in seconds. Sellers act faster, managers coach on the outcome, not on dashboards. So let's see what happens when I end practice here. It's going to take like 30 to 40 seconds usually for me to see the feedback. Meanwhile, I just wanted to show you one quick thing first to make sure to just give you assurance that this is not just a demo, it's a real thing. Let's do this. These are all the Slack messages that we have sent to our account executives over the past two years. So let's look at the very first one that we sent out. So this is just a conventional machine learning model. No Atlas reasoning engine, no sales coach. And it's two years ago, I guess it says it's July 2024. Yeah, more than two years ago. So the message that we sent was that your top growth factor is pipeline generation. And here is a link to a webpage that can help you. And here is a video, a two minute video that you can watch. And now let's compare it with the very last message that we sent out. Just go all the way down. First it's not wake the previous message that hey, your problem is pipe gin. I mean good luck, here's a two minute video, go and fix it now. It's like hey, pick up the phone and call the customer. It's not about abstract conceptual stuff now anymore. It's like take action, pick up the phone and we have a recommendation system in place that looks at the probability of conversion and recommends the best next action based on pipegen, openpipe, acv, all the metrics they have. And then the action that we recommend is not watching a course or video or it's just taking action. Practice with an agent, see how it goes. So now let's look at what agent Force told me. There we go. Here's a summary of the feedback. I nailed it folks. I just didn't want to spoil it for you just because before getting to the end I nailed it because I practice 10,000 times. On top of that I'm getting another next best action. Another beautiful, beautiful insight nugget here that while you're scheduling that tableau call with Cengage Learning, consider attaching a data cloud course. Because our data cloud AI models thinks that there's a high chance of cross sell opportunity here for data cloud basically to build something like this. What we did, we brought all the analytics, conventional machine learning outcomes, many things that many of you already have in your companies. I mean we have propensity to buy, you may call it leader scoring system, we have cross sell, you may call it, I don't know, something else. But these are really a standard conventional machine learning analytics models that almost any enterprise level companies have. We brought everything into one place to Data Cloud and there we go. And then that's where we basically leverage Atlas Reasoning Engine. We brought our institutional knowledge into Atlas Reasoning Engine. We trained Atlas Reasoning Engine that hey, in Q1 we care about Pipe Gen, in Q4 we care about closing. It has all the context and using that it has the ability to make the recommendation based on all the outcome, all the analytics that we give it to it. And also you can build your own model for whatever reason you don't like Atlas. It's a fantastic product by the way guys. I mean I highly recommend it, but for whatever reason you don't like. I built this, my custom deep learning model. I want to bring it in. I built it on aws, Azure, gcp. Bring it in, you can bring it any model that you want. And from here all these insights go back to data cloud again. It stores there and then we have processes to trigger that through a slack sales coach CRM and leverage it to
different places back to this kind of agentic enterprise and the way that the world is changing. The most important thing that Ali's demo didn't highlight is all the people behind the scenes needed to make this work. Not so much from a tech perspective. Agent Forest does a lot of the heavy lifting but it's the expertise on the skills, the products, the techniques, all of that getting funneled into data cloud. So coach, so slack, so whatever can use it. This is one of the biggest takeaways out here is that it is multiplying the expertise of people, it is getting it agent friendly and it is getting it ready to be surfaced wherever we need to. So this source of truth is one of the biggest parts of what we've been building here at Salesforce. So what this can now do is we have this now a way to express this source of truth in many ways. So what Ali's going through is showed a couple of these different ones together. But really this is the way we like to think about it now. Starting with knowledge practice, going into your application and then your feedback overall. I'm going to go into some quick details on this. So when I say knowledge, a lot of the agent work that we've seen helps you answer questions the best way possible. So this is really what knowledge is, again, it takes that expertise that we have available and it lets you have that conversational chat with an agent at 2 in the morning when you're a new hire, when you're a pro, whatever you want. Hey, what does it mean? He mentioned cross selling. What is cross selling? And it comes back in that conversational safe space nature. It's a great way to help communicate. It's very human driven, as I'm the one reaching out for this knowledge and now I'm the one getting that knowledge. The next step is practice. So as we start to learn a new skill, it's too hard to throw you into everything in the world. We showed the complexity earlier, so this is now what we've been doing, where we have these more product certifications or skill certifications, where we control the environment. We don't want all the noise and the lights and the complexity of your customers to take away from the skill or the product you're trying to learn. So in this case, we still take that same source of truth, but we tailor it even further. So back to what Ali was showing here with the opportunities that we can create, those can be created in essentially an environment that allow you to communicate the message and the evaluation specific to that skill, specific to that product. Then all of this has to be applied. This is where, again, coach, what you saw at the end for Ali with the next best action, this is where it comes together even more. So you've done your knowledge, you have some practice now. We want to help make sales, we want to close deals, we want to get you upskilled and actually use it. And to do this, we actually have to pull in the data so we understand who the customer is, what's the value you're matching up to your product. So in the case when Ali, if you were to go one step further to the next version of that coach, where he can then practice on that next best opportunity, it won't give him the feedback saying like, hey, you did a great pitch there. It's going to say for your customer, here's where your pitch needs to go a little bit differently. Here's where you want to angle this a little bit more. Hey, based on their history, you didn't mention this value of the product cloud that you could be offering. And it tailors it to that specific customer, that specific ae. And what's critical about that is if we were to start there, that's too overwhelming. They're not learning the skills on the fly. They need to adjust. That's what that practice is now the last part of this is the feedback. So what's super important about the feedback is that agents and the data evolves. So as we go through, we want to understand what worked well, what didn't work well. There's a couple ways to hit this. One is from the data that exists in data cloud, from essentially a seller's patch, the way that they analyze their work, the results of winning deals or losing deals, we can look at that and say, hey, this knowledge needs a tweak, it needs a change. We also through Slack can directly channel that information in to say, hey, this didn't quite land with the customer. Could you help me answer this? And that feedback just comes back and ties it all back together. So with that, we wanted to say thank you. We appreciate you guys taking the time today. Have a great day. Thanks for listening.
