What Selling Looks Like When Your AI Agents Have the Context Your Sellers Do

It's always a pleasure to be here and it's been very exciting to see the evolution of AI in sales through this conference and its evolution, you know, year over year as we kind of see how it's changing and it's exciting as we now start to talk about glasses and new form factors and all sorts of new things. I was also really excited to hear from Chris and what's, what, you know, what they're doing with, with, with Clogforce these days and Anthropic and Salesforce. So I, I wanted to talk a little bit about context, and I thought I would start with a very short 30-second video to introduce what I mean. Let's see if we can get it going here. So, you know. Just a little silly one, but I think we're trying to make the point that, you know, as you think about— if you think about context, it can mean a lot of things, but it's very important, particularly when we start to talk about AI. And we're going to talk today about how context is not only important in AI, It's important in just sales. And one of the things that we've observed is that over the last couple of years, and some of you may be aware of this, as you think about AI, one of the things that we do is we go buy another tool. So we buy, you know, a— we start with a chatbot and then we get a demo and we buy another tool. And we have many clients who are buying acquiring tools, deploying them out to market, running pilots. And what they're finding is that it's never, it's never quite the right fit, right? And it's either not the right fit in their org or against their sales team. And at the end of the day, it's not providing the value that they're looking for. And there's always this hope that maybe the next one will work. And so if we think, let's, let's talk about why. The short answer is enterprise sales is quite complicated, right? I mean, when we think of it very simply, you know, we start to prospect and qualify and discover, and everybody on this webinar knows this very well at this conference, right? But, and the instinct when we talk about AI is to think about the tool that we're going to give the seller, or maybe even the workflow and reimagining it as the most important thing we have to solve for. And I think what we've started to observe as we've seen scaled AI deployments is the most important thing to solve for is context. And building context and getting the right context to your AI is actually the hardest part of an AI transformation. And it shows up in a couple of places as you start to actually implement it. And we'll sort of, we'll talk about what those are. And so let's talk about what happens today, right? Like, what do we even mean by context? So just Think through this for a second with me. You've got a deal, and we go through this lifecycle where you've got to start with prospecting, right? So maybe you're an AE, maybe you're a BDR, maybe you're a field sales or key account manager, but there's some common things in here. You got to understand your account that you're chasing or that you're trying to expand in, and your account history is living in many places. You might have notes, maybe CRM notes if you're very good at actually populating them. Or if you've, you know, had some kind of activity capture system, your emails are very valuable, your Slack messages are likely very valuable, and the most valuable is probably the memory of the sales rep, right? And if you think about that process, what you're now doing as a sales rep is you're trying to build that context. You're looking at your notes, you're looking at your data, you're looking at what you remember, and you're going, okay, now who's my buying group? Who's involved in the sale of this process? And maybe you have a good tool, but that's in another system where you have to go search, or maybe you're on LinkedIn, or maybe you're doing both, but you're trying to map out who am I selling to. And then as you go to explore how to propose and how, you know, a solution as you walk through the lifecycle and sets of meetings, you're also probably looking at, well, what did they buy from me in the past? What did they buy from me potentially through a channel partner if I sell through a channel that's not direct? What quotes did I send them? What did they— what pricing did we give? What contracts do we have? What renewals might be coming up? Again, all context, and in many cases sitting in different systems around your organization. You might have install base, you might have order management, you might have sticky notes, and you might have, you know, all kinds of other places that they sit. As you go through this process, The other thing that happens is you make decisions, right? And all of these decisions are often not recorded. So for example, what, why did we give the discounts that we gave? What was the reasoning and logic behind that? And how does that impact our future pricing and discounting? Or what was the reason it was scoped in the way it was, or the term decisions were made the way they were? If the seller has changed, often that context is lost, Right. And when we look at deal reviews after the fact, we're often listening to a seller's narrative of what happened, often missing the underlying context of the voice of the actual customer in this.

Okay.

Finally, as you walk through this cycle and you get from, you know, qualifying the lead and discovering their needs and building a solution to let's close a deal. Now you're gonna negotiate and you're gonna You know, build a contract and you're going to get it signed. You're navigating, you know, many of you are navigating legal or cyber procurement operations credit as aspects of something that needs to get done before you can close this deal. And as you're doing this, every single time you're rebuilding your context and understanding of those components. You're often rebuilding the contract from scratch. You're often re, you know, going through a legal process over and over again, not only with the customer, with your own team. And there are many stakeholders in your buyer's organization that are involved in this process that you're often not even interacting with or you're not aware of. You know, we've seen that in typical B2B, you're talking about 6 to 10 people involved in a typical buying process. And because of your sellers need to navigate this complicated process, they're still spending often 30 to 40% of the time actually selling, whereas 60, 70% of the time is being spent, you know, on administration, navigating these, the people and processes associated with this. And it has real implications. It has capacity implications, you know, 500 deal days as an example lost. It has pipeline that becomes at risk as we are trying to build this context. Now you go apply AI to it. Now we say, okay, well, I want my AI agents to do this work that humans have to navigate. And the AI is supposed to build and navigate this context. And honestly, it's like this warehouse. You're kind of saying good luck. to your, to your humans, but especially, and especially to your AI, go find the right information and make sure you can do the right reasoning, go take the right action, and, you know, create value that's measurable. It's not particularly easy to do. So when we talk about context and you go build an AI and agentic system, there's really 2 things that have to happen. One is the heartbeat of your organization and how you do work. What do your sellers do? What do you sell? Who are your customers? How do they actually win in the market? Understanding what makes you unique is the context you need to build into an agent. So the idea that I can take any standard agent that does account prep and call prep and, you know, account briefs, and it's gonna work for me, it'll work a lot better for you if it has the context of your organization. Which, which comes in how we build the agent. It also comes in the learning loop that's required in actually building the agent. The second thing that's really important is the context of the organization, meaning the knowledge and insights of the organization. And that's data that is sitting in systems that matter a lot. Systems of record are not going away, and we'll talk about that. But it's also the unstructured knowledge. It's the knowledge of your rep. It's in notes, it's in emails, it's in Slack, it's in Teams, it's in other places. We need to capture and leverage that and do it in a way that doesn't become all of these boxes but becomes organized. So let's— I want to show you what it looks like when we get it right, and then, you know, why, how does it work? All right, so this is a demo. We've stitched together this video. It was recorded last night, and for speed, a couple of the prompts and responses are sped up just so that we can get through it quickly. But this is a live system making real calls against the data. So what you're looking at, first of all, is a seller's dashboard. So this person is doing business development, you know, top of the funnel, and you're looking at a headless CRM experience. So this is a composable front end. It's quite simple. It's an AI-native experience. You can see the chatbot off to the, to the right, and the content on the left is all AI-generated as well, and it can change depending on the context of the conversation we're having. And underneath it is an AI orchestration layer that is connecting all of your systems together and is running your agents. And that allows you to connect not only this but other environments, and we'll show you that in a second as we go through this. So I'm a business development person. I've come into my home screen here, and the first thing I've identified at the top is that it's identified not the 100 things I need to do, but the top 3 things I need to do today. And as I look at what I do, I'm going to start with a quick prompt. I'm going to say, okay, I can see that Lisa's come in and she came into the website and I'm going to want to prepare for my meeting. So I see that I've got a call with Lisa coming up. What, what do I need to do with Lisa? Tell me about it. And so it's going to get us prepared. So a couple of things it does. It's pulled what we know about Lisa. Lisa came to our website. Lisa filled out, you know, talked to our agent on the website and asked for this call and got it scheduled. She's the VP of infrastructure for this hotel group, and she's got a problem because she's got a Wi-Fi and a connectivity issue that is impacting her customers. And what we're going to do now is help this business development rep prep for this call. So you can see it's— First of all, it's telling us this is our 4th conversation we're going to have, and here's how we might prioritize. Let's understand their needs. Let's identify their pain points. Let's understand who's involved in the process. Let's understand their timeline. Make sure we listen for value anchors, like what creates value for Lisa, so that we can put the right proposal together. And it kind of tells us our goal is to understand her priorities so that we can figure out what the right next step is. So I'm going to continue and say, okay, well, let's draft the agenda for the call. And again, the response we sped up just a little to make it, make it easier. So it's prepared a response for us. So it said, okay, well, look, you're gonna— we're gonna have, you know, you're gonna want to introduce yourself and make sure you understand who Lisa is. And then understand Lisa, like what, what, tell us about you, Lisa, like what problem are you having? Tell me about your properties. What impact does it have on your business to not be able to have your guests have connectivity? And then what are you trying to accomplish and when? What, you know, what's the, what's the, what's your budgeting cycle around this? What's your implementation? What else do you have going on? And then being able to close with, okay, how can we help? You know, what, making sure that we share what capabilities we have that answers those questions on the call. providing proof points as necessary and addressing, you know, concerns that they have. It also kind of reminds us how to approach the call itself. And, you know, once we, once we go do that call, and I'm going to speed through, one of the important things that we can do on the call itself is provide live feedback. And I'm happy if anybody has a question to see what that looks like in terms of live coaching. during the call, but it requires 2-way audio and I'm gonna skip that here. But after the call, one of the things that we've done is we've captured Lisa's information. So now I come back after the call and I ask, okay, can you help summarize the Lisa meeting? Like, what do I need to go do next? And so now my agent's gonna be able to pull this unstructured data from the conversation. And so it's going to pull the transcript of the conversation itself in the context of what we learned about this company and help us plan out what we need to do. And now we've understood that Lisa's got, you know, 14 properties. She wants to start with 10 of them. They've got to replace access points and switches. I'm a hardware networking company, as you can see. And we've learned that there's budget associated with this. Lisa has the authority to approve. There's a timeline associated with it. And we, and the agent has gone and found that we have a product that best meets that need. And so it's suggesting us, to us that, A, we need to confirm, can we actually meet the, you know, meet the delivery deadline because it's coming up. And then it's recommending that we convert this lead to an opportunity because it's communicating with Salesforce. And it's saying, let's go build a proposal together. And in some cases, clients actually ask for the lead conversion to be automated. Like, we don't need to actually have a rep say, let's move to an opportunity. We can do that because we have the signal. In this case, the company wants the seller in the loop. So what I'm gonna do now as a seller is understand, okay, like, I've understood what, you know, my opportunity. Let's go see if I've got the inventory so that I know how to approach the next discussion with Lisa. And so I'm going to check, you know, inventory availability and lead time. And it's gone into my ERP, into my supply chain environment, into my warehousing, and identified that, look, we can meet this deadline that's coming up, and we can deliver it to Lisa in Springfield on time with some to spare. It's also recommending, let's go ahead and move this to an opportunity. And I'll say, okay, yeah, let's, let's go do that. And now it'll create the action in Salesforce to do that. Directly. So the other thing it did at the end of that is it identified that, hey, we may want to collaborate with some of our colleagues now because I'm going to need sales support to build a proposal and the quote out. I'm going to make sure an account record is created for them, or if it is, to make sure we, we have credit available on their account. There are maybe 3 folks that I'm going to need to interact with. And so the agent has automatically gone and generated a Slack channel that I can work with my organization on. So now I go and I go into my Slack and a couple things you can see here is one, it's got, it's got all the same context. So it has my leads, it's pulled it from that orchestration layer, which has not only Salesforce but other things underneath it. So it's my window into my CRM and it's generated this Slack channel. that I could work in. And so I go to Slackbot. And for those of you who don't know, Slackbot is Salesforce's Slack AI capability that is its window into the underlying data. And I can essentially run the same agents in Slackbot. So I talked about, hey, you know, I had met with Lisa. I know I converted this lead to an opportunity earlier. And What, what, what should I go do next on this? Right? So I, I've got, I got Lisa being worked on. What am I gonna focus on next? And, and perhaps this is the end of the day or the second half of the day. I'm thinking about what I'm gonna do tomorrow. So what Slackbot's done is said, okay, well, that wasn't the only lead. We've got a bunch of other things we should go work on. And I can now look at, okay, well, you're, you're saying to talk to, to Daniel, but why is Daniel a hot lead? So now it's looking at the underlying data that we have available from Daniel, not only from the interaction that Daniel perhaps had with our website or with a, you know, with some interaction with us, but what do we know about Daniel and Daniel's company Meridian Logistics from our ZoomInfo, you know, or other data provider as well. So this we can see came out, came as an inbound lead. We have some work to do here. There's some timeline pressure. So let's, let's we should work on that one tomorrow. We can also see that Whitney is on the list and I want to know, well, what's Whitney's pain point? And so, well, Whitney's got unreliable Wi-Fi in service bays and she, you know, and there's a quote there of some of the things that she shared through an email exchange. And so now at the end of the day, I said, okay, you know, can you send me the top 3 actions for tomorrow morning? And it will generate reminders for me. It will prepare Outreach, it will prepare account briefs. I can essentially call the same underlying agents. And so what's, what's important in this experience is that, A, my— you can see we're not in Salesforce, but we're interacting with our CRM data. But we're also interacting with all kinds of other data throughout the, throughout the environment and your ecosystem. And that's when an AI tool goes from being, you know, a chatbot that we all have access to And where you can connect to things to something that's actually doing work. And the difference between building context that is reliable and building context that's available but not reliable is the magic. And maybe I'll talk briefly about how to do that. So there's really 3 things we do when we think about how to do this with clients. Number one is we look at how to re-engineer the process. So The work we just did that we showed you, that's different than the work they would have done 2 years ago. It's a different way of working, different place of working, different steps. You need slightly different skills associated with it. And so we want to re-engineer the process and rethink the role of the human, the skills they need, and how they work. And as you go from this process into your enterprise AE, your technical seller, your sales engineer, a field rep, it changes and gets even more significant. 2, we build the agents and we give them the tools so that they can take action. So you, you have to define not only the agent, and everybody here is probably familiar with a, with a Claude skill or similar. So it's not only defining what the agent does, but it's defining what it should do deterministically. Not everything should be an AI solution. Some things you need to pull the right data for and validate it's the right data. And, and it needs to be able to take action in your systems across your org. We had one client who built a similar composable front end for their sellers, but it wasn't able to write back to their CRM or to other systems in their org, and they weren't actually able to take a lot of— there wasn't a lot of value there for sellers, so they still had to go and swivel chair in other systems to update them. The third is we connect and build the agents to the right context. And so that's understanding the culture of your organization, the knowledge and the data. That's structured data, but it's all the unstructured data as well. So that's the ability to generate what you— MCP servers, or another way to say that is the semantic layer that allows an agent to access it. And in order to get it right, what becomes really important is that you stitch all of this together. And if you think about the analogy I'd give you here is when you hire a human and you want them to be productive, you give them the right tools in terms of a laptop and access to the systems, and you give them context about your organization, like What, what is the job? What does success mean? What do you measure it on? Where, who do you ask for help? Which, when should you not do something? Do you, you need to give your agents, your AI agents, that same information to be successful. And so when we think about a digital worker at scale, we think about them against those dimensions because it's not just the agent and the skill, it's the context you bring to them, it's the re-engineered process. And it's the ability to retain the context through a long set of interactions, right? Because, you know, whether it's a seller virtually selling independently to a customer or a forecast that's constantly updated with all the signals of your organization, these are not point-in-time single conversations with an AI. They are multi-step, multi-system, and its ability to persist context becomes particularly important. And the way we do that is to have really think about this in terms of 3 key decisions here. And the first one is where, what are the surfaces or experiences that you want to give to your sellers for where they work? And you've heard for many years about a single pane of glass, but in reality, I would argue there's no single pane of glass. Almost every organization was going to have productivity tools, either the Microsoft or Google Suite. They're going to— you're going to have email in there. You're going to have some collaboration tool, Slack or Teams. You're going to have some AI workspace. Most of our clients now have a Gemini, Claude, or ChatGPT, or their own front-end version of that that uses that model in the background. And then you're going to have some sort of structured workflows of CRM because there's still a place for structured workflows. Not everything that should be done should be done through a chatbot. If you want to drive consistency of process in your organization, you still need common dashboards or common workflows or deterministic actions. And so there are still going to be multiple places to work. Deciding where your employees work, and I showed you 2 examples, is important decision one. What we, what we say is bring your AI to those environments. So you do that by building an agentic orchestration layer. And so our Converge Sales OS layer is exactly that. It's a, it's a coordination layer that builds the semantic layer to your organization, connects it to all your systems, builds and runs the digital workers that we talked about, and then can surface it in any of those areas, including the composable custom front end that I showed you, or into Slack, or in Salesforce directly. And lastly, it's your data foundation. And systems of record are not going anywhere. Governed systems of record are increasingly important. Being able to take unstructured data and structure some of them into fields is important. And if you've ever chatted with your— an AI agent and you gave it information and documents, but it forgot something or dropped something, then you know the risk of not having the right context available to that agent or losing context. And so building semantic layers and data engineering processes that actually persist context for your, for your sellers is really important because there's money at the end of this, right, that's impacted. So where do you start? Well, all this, I'll close there. There's some no-regret moves. The number one is build the adoption muscle in your org. At this point, everybody should be doing this. I think number one is Checkbox for probably all our listeners today. Make sure that you're starting to build skills, you know, build common ones across your organization, build them specific to your specific workflows, get people using them. Number 2, build that orchestration and data context. And you do that with number 3, which is your— what are your applications in your organization? Are they API and MCP ready? MCP is Model Context Protocol that allows an agent to go in, you know, English language, natural language, pull the data it needs to go run the process it needs to do. And that becomes very important. So build your orchestration layer, build your connections to the rest of your organization and pull the data in because otherwise any agents you surface to your sellers won't come up much value. And then go work in small 12-week increments. Where's your biggest pain and where's your data the cleanest? That's usually the best place to start. Start with a problem where you have data, and then you can go execute and get it to users and test it. Run an A/B test, see if your hypotheses are working, see if your technology is working, see if it's reliable AI, it's trustworthy. And we've typically seen people start either in pipeline gen, like top of the funnel, or kind of middle of the funnel, like deal progression, understanding the deal, how to move it along. And where clients are really moving now is much more around how do I do my quoting and transacting and renewals processes as well. So, you know, if you do it well, you're, you know, it's not just going to be, you know, a single pilot anymore. It's not going to be a single use case that you deploy. It's not going to be just one system that operates on its own somewhere or an isolated new software rollout because you bought a new tool. It's going to have to be something that ends that doesn't just end when your implementation does. It is a much more coordinated experience across all of those. And so that, that's where we see the market heading, Gerhard.

Well, I'm really curious how about your internal shop. How did Deloitte make the transition to Agentic AI?

We're still making it like everybody else because we have the same problems of context that's sitting everywhere and people that do all kinds of work. But I would say we've leaned in hard on the adoption muscle. So we have, you know, A, all the AI tools you can imagine. So they're— and we use them regularly in our work. B, we learn from that a lot. So there's a lot of learning loops built into our skills and to our agents that refresh it. And, and, and see, we're using it. We share a lot. So one of the, you know, one of the most valuable calls that I have of the week, I have 2 different ones, which are basically community calls where we share interesting things we're doing. And it could be in our personal life, it could be at work, it could be with a client. But we learn from each other, like, what's actually working and why. And that's one of my, my favorite meetings. If your organization doesn't have that, I highly recommend it. It gets lots of engagement very quickly. You learn a lot.

When you have those meetings, who is there? Is it— are there people there from similar tasks? Are you bringing anybody from the organization that can come in? How does that work?

It's sets of communities, but just because if you invite the 300,000 people at Deloitte, it gets unwieldy. But so my customer team in the US meets. We have an AI-focused team. They meet weekly on this topic, and then my team, which is our sales transformation team, also meets. And so we, we basically go to our teams and Slack channels and say, hey, well, you're welcome to join. And each week we identify, you know, 3 people that would like to share for 10, 15 minutes. And we've got a laundry list of people sharing, and we learn a lot from each other, both about what to do, you know, on projects, things that are working for our personal productivity. But we learn, we learn a lot that way.

And now, are you taking that and sharing it back to a larger group in the organization so you can kind of see what people are doing that maybe isn't related to you at all, but you might be able to find those opportunities to save time, be more productive?

Yeah, it does. I mean, certainly there's, you know, it gets recorded and goes back to these communities so people can read it. It also goes back into formal trainings. But my, my personal bias is that The, the informal learnings from each other has been the most valuable way versus all of the formal. But by the time you get a formal thing out there, it's already moved on. So this is the rapid pace is pretty important here.

And you do it every single week?

I have 2 of these meetings every single week.

2. And then when do you host them? Is it the morning, the end of the day?

What time of the week? Mine are Friday. One is happening now in this hour and one is a little bit later in the day.

Oh, so both on Friday?

Yeah, 2 different communities of people, but they're both on Friday.

Yeah.

Great.

Yeah.

Stephanie, I want to address the same question. How do you lead that transformation at Bryant University?

Well, so I've actually taught some classes to the faculty and the staff, and I teach to the students as well. But we also have meetings. We don't have them as frequently. We have them every other week and sometimes monthly, depending on when it is. In the school year where we come and we share what we're working on. And there's different communities, kind of like what you mentioned as well. So.

That's great.

It's a never-ending job, always learning.

Well, I love the idea of Friday. You know, we have our meetings on Wednesday, so I'm wondering if maybe Friday is a better time.

People are loose. They're like, you know, excited for the weekend. So you get a lot of good vibes on those calls.

How, how long are the meetings? 20 minutes?

One is an hour and one is 30 minutes. And we usually get 50 to 150 people on each of them. And you have to understand, again, 300,000-person organization. This is 2 of many ways to train people, but it's my favorite.

I think it's helpful. I, we also host like research retreats and, and business, College of Business retreats, all these different things to help support teaching retreats. But to your point, sometimes it fades. Like, if you're planning on something 6 months ahead, whatever you thought you wanted to share is no longer really that interesting anymore. And so those little impromptu sessions are, you know, it helps. But I like the idea that you're identifying a couple people each week to do that.

Yeah, gives it a little bit of structure.

So how long do they have? Is it like they decide on Monday who's going to do it, or?

Yeah, we decide, you know, at least a week in advance who's going to present, and it's usually 10 to 15 minutes per person. And we try to do a mix of either something with a client or something that somebody's doing for their personal productivity, or so it's a little mix of— we try to keep it interesting that way.

Great.

And also levels, like, we— this is— there are brand new analysts coming out, you know, interns, you know, versus like our AI leaders. Like, in fact, I'd rather hear from the interns half the time. So It's, you know, it's okay. It's, we're all equal on that.

That's great. I, I love hearing this, like, this learning organization that y'all are creating. That's very valuable.

Yeah. And it brings back the metaphor that you introduced in the beginning, Abe, which is that of an orchestra, you know, creating a pleasing sound and then getting confronted with the reality that kids want to scream.

Right.

And I think it's a very good metaphor. But what comes to my mind is that you are the orchestra leader for transformation in, at Deloitte. And Stephanie, you're the orchestra leader at Bryan University. What is sort of your guiding vision, where you get your inspiration, to move this forward in spite of the fact that there are a lot of kids screaming around you?

You first, Stephanie.

No, go ahead, please. You first.

For me, it's about the, it's about the people and the humans in our organization. There's the, the world is changing around us. And there's, there's 2 ways that this plays out in my mind. You know, organizations get smaller and that's a quiet story, or we look back years from now and we remember that, that, you know, we were enabled, you know, for the future and there's a future out there for everybody. And so I want to make sure that everyone feels, you know, that there's a place for them and they're a part of creating it. And so that's, that's my guiding principle, not only for me, but for our clients as well.

Very noble.

I love, I love that take. And I mean, you know, it's— there's so much fear around it, and being able to show that everyone kind of has that place and we're building it to support each other rather than to displace each other, um, it's a nice perspective. Gerhard, what about you? What do you think?

Well, I think that the— I, I saw you trademarked something, orchestration. Is that Is that a trademark that you got?

The, uh, Converge, uh, is our, is our product essentially, but it's our, it's a convergence of all of this context and all of the systems and how we bring it together. And so we've, we've built a solution that allows you to basically take that, what, what we saw live very quickly within, within the, uh, weeks to months.

I, you know, what comes to mind, Stephanie, is that we are human beings and we go through cycles. And we go, we have moments of high energy and we have moments where we rest and sleep. And I think an organization is going through similar cycles where there's high energy, there's high sales, there's high growth. And then all of a sudden something happens and it collapses and then you need to rebuild and need to refresh. So I think my metaphor is reconstructions, renewal, where we constantly create new ideas that we bring to life and with the goal to serve people and create something that's useful for everybody.

Awesome.

I love it.

Thanks for attending.

Abe, thank you so much. You always bring clarity and insight, uh, to our conferences, and I appreciate your loyalty. And, uh, I hope you join us next time again.

Always. Thank you both.

That was excellent, Abe. Thank you.