Beyond the Chatbot: AI That Practices, Evaluates, and Builds Sales Readiness

I'll just jump right into it because it's the last session of the day, and I know everyone here has seen a lot about what AI can do across the sales process. Obviously, I want to focus on the training aspect before your reps get into a live training scenario. And there are many different conversation roleplay chatbot solutions out there, and I think it can feel like it's very commoditized. So what I want to focus on is just beyond the actual conversation itself. All the decisions are on there that's going to make employing conversation practice the most beneficial for you and for your teams. So as I said, I think it can feel quite commoditized. Here we go. There we go. But there are basically 3 levels to what makes something effective. The first is, can AI have a conversation? And the answer is yes. Yes.

Right?

That's widely available now. The conversation alone isn't where I think the biggest value lies. The second level is where I think most value is created, and that is, can you recreate the specific customer interaction that your salespeople actually need to prepare for? What level of control do you have over how the AI persona responds, the objections that they raise? This is where the level of realism needs to be very high. And the third, of course, is that once the conversation is over, what do you get from the sales enablement point of view? Do you actually have the ability to say what your salespeople actually demonstrated? So that's what I'm going to show you today, right? How quickly you can move from a sales conversation or a sales situation to creating a real practice scenario, the level of control you you have over that process and what readiness evidence you have as a result of that.

Okay.

If I can move ahead to the next slide here. I don't know why this is not working for me. Sorry, everybody. Give me one second. I don't know why I cannot bring up my, okay, my browser here.

Okay.

So. Yeah, just before we, we build a scenario, there are, this is the scenario we're going to use just as a, as an example. So in this situation, we have the example of perhaps we're trying to sell cybersecurity software program, and we have some information in advance, right? We have the customer profile, we have some information about the opportunity that we're providing to our sales rep, and we know a little bit about the relationship and the competition. So how do we go from having what this is, which is a realistic scenario, to actually creating, uh, an actual chatbot for our reps to practice. And I'm going to show you the first method. There are 2 ways, and I'll show you the first one briefly because sometimes you need to create practice scenarios very, very quickly. And one of our capabilities is using AI to support that process through auto-generate. So instead of starting with a blank screen and having to manually build every element of the practice scenario. You can give the system your existing materials, you can guide it through some basic objectives, and it's going to generate the foundation of the scenarios for you. And in this case, it's actually generating not just one scenario, but actually 9 scenarios with variations according to the information that we provided about the product, about our commercial playbook. It's going to be able to recognize what are the objections that your reps are likely going to face and even make suggestions about the types of personas they should be practicing against. And this is going very fast. We're going to slow down and take a look at all of these decisions in the next stage, but this demonstrates how quickly you can go from having your product information to creating realistic practice environments for your reps very quickly. So that's what this shows. We've now generated 9 different practice scenarios. So that's very fast, but we don't want to leave all of the decisions just to AI. So here we are in the configuration view where you can tailor the situation to match the conversation you really want your reps to practice for. So at the top here, we have the context that we started with, the Cyber Secure Awareness Program. The customer profile, the opportunity clues. This is exactly what we saw on the slide, but it's the information that you would want to give your rep before they practice or before they enter the meeting. We also want to define the goal of the conversation. Not every sales conversation is going to be about closing the deal. So in this case, we've defined it as securing agreement to introduce our solution to the CEO. Next, what's very important is that we define our buyer persona. We create personas that are realistic to the types of buyers that our sales reps are going to face. So here's our buyer. His name is Robert Hayes. It's just the name that I gave him, and we know that he's an IT security manager. He's the type of person we hope could become an internal champion for us. But we can go further than just giving him the job title. We can define how he communicates. So he might be open and conversational, but he might also be terse, skeptical, and maybe reluctant to share more information. I think The important part is that what we define here is actually going to impact how Robert communicates to our reps when they're practicing. And we can create different press— practice experiences around the same sales situation, depending on the type of buyer we want the rep to learn how to handle. And then we get into something that I think is equally important, which is the dialogue constraints. And these are effectively rules for how the customer, Robert should behave. So for example, we might tell him to raise an objection early in the conversation, or we might say that if your salespeople who's learning asks thoughtful questions, then Robert can reveal some more information. So we can create these kinds of if X then Y behaviors, as well as things that the buyer shouldn't reveal unless the salesperson has earned that information. So it's these decisions where the precision of the scenario design really starts to matter. We're not just telling the AI to pretend to be an IT security manager, we're defining how this particular buyer should react in this particular sales situation. And underneath these types of decisions, after we save the persona, is the hidden background. What this means is all of the information that the buyer might actually know, all of the context that we want to get at, but we don't have before we enter the conversation. So for example, who's involved in the decision? Who are the influencers? What has happened internally before this meeting? What concerns exist that the salesperson doesn't know about yet? So the AI knows this information, but your rep who's practicing doesn't. If they ask the right questions, they can uncover this type of information, but if they don't, some of it might never come out. And another thing that we can define in advance is the objections that Robert is likely to raise. So sales enablement teams like yourselves know the objections of your clients better than anybody. You know what they say. You understand what the underlying reasons behind their objections are, and you can define them in the scenario. So in this example, Robert might say, well, he, he's not sure that they need another formal program. They already send out reminders and security updates. And the salesperson might hear that objection, but they might not automatically see the underlying reasons behind that. And that becomes part of the context. And you can also define what a strong response should be. So the practice is about whether your reps respond in the way your organization expects. Finally, and I think perhaps the most important, is the evaluation criteria. How are you actually going to define what good performance looks like? In this scenario, I've chosen to weight objection handling more heavily. So we've defined the behaviors we want to see. What leads to good objection handling in this model is empathizing with the customer, understanding the root cause of the objection, resolving the objection, and then returning to the commitment. So when the AI evaluates the conversation, it's evaluating against those specific behaviors that we've defined. And, you know, every sales organization may define these things differently. You might have your own existing methodology, you might have different sales stages, or you might have very specific expectations for what your reps should do at each point of the conversation. You have control over all of that. So for example, in the case of message delivery, there may be required product messages that need to be communicated, maybe even for compliance reasons that your reps need to communicate to your buyers, or claims that a rep should never make. And those can also become part of the evaluation. So these choices become more than just giving a conversation to your reps. You control the situation, you control the buyer behavior, and you control the standard against which your reps are evaluated. So now let's look at what this looks like from the perspective of your learners. So here I'm going to use a recorded run so I can just move through the interaction quickly and show you the parts that are the most relevant. Distinction is really helpful. When those close calls happened, what was your bigger concern internally? Was it mostly the immediate security risk, or was there also concern about client trust or operational disruption?

Center. I am measured on incidents, so anything that gets past the first click worries me. At the same time, leadership jumped straight to client trust and reputation. The question was, what if the next one is a real hit and a client finds out? Operational disruption is a concern, but more as a downstream effect. The bigger internal noise was risk to clients and how we would explain it if something actually happened.

So here, this is where, because of— I did some discovery, then Robert is revealing some of the hidden information that we defined in the scenario. Later on, he can present some objections.

Realistic short practice is more interesting than another annual course, especially if it actually shows behavior trending in the right direction. My worry is still that we already send reminders and updates, so I need to be sure this is doing something different in practice, not just rebranding training.

Of course, I wouldn't suggest replacing what IT has already— So I continue the conversation, and it's also important that because we've defined Robert's scenario, if I ask to have a meeting with the CEO, he's not just going to be a pushover unless we've specifically, you know, designed him to be that way. So later on in the conversation, He'll respond in this way.

Before I take this to the CEO, I need to be very clear on something you have not really covered yet. What does this look like in terms of time and disruption for employees day to day? How long are these simulations or trainings? How often do they run? And how much pushback do you typically see?

So you get the sense of the conversation. What's important here is that the scenario that we've defined comes across in the actual responses. The questions that he raises, the objections that he raises are going to be realistic to your scenarios because you've defined them that way. So here, what does it look like when you get the reporting that's generated? So because you've defined what good performance looks like before the conversation, the evaluation can be tied directly back to those criteria. So at the top, I get an overall score and a summary of how the conversation went. And as the learner, I can immediately see what I did well And where I can improve. So in this example, one of the highlights is that I positioned a pilot approach pretty well to address his concerns, but the system also pointed out that I could have done a better job at the end of the conversation. I could have established a more specific next step. But I don't just get that overall score here. As we've seen, I weighted objection handling a little bit heavier here, and I get feedback specifically on how I handled 2 different objections that were raised. So in this scenario, the feedback recognized that I showed some empathy, so that's good, when the objection was raised, but it also identified some missed opportunities. I didn't ask enough follow-up questions to understand why Robert believed that his existing reminders were sufficient. So this is much more useful than, than generic feedback, and the reason why is because we have had control over how we design the feedback.

For—

again, for example, the same logic applies to things like message delivery. If there are required messages your reps need to communicate or claims they should never make, those are part of the evaluation as well. And this is obvious across any practice element, but repeated practice is very powerful. So if I'm a competitive sales rep, I look at the score 60, I'm not satisfied, I want to beat my peers, I want to practice again, and that repeated practice immediately after receiving relevant feedback is what's going to build that muscle memory and allow me to improve. But now let's change perspective a little bit. If you're running sales enablement, these individual scores are very valuable, but you don't just care about whether one salesperson improved. Now you can also have that evidence across an entire team. Once your learners have generated enough practice data, you can zoom out from the individual rep and look across the team. So as the administrator or the sales enablement leader, you can get the overall view of what's happening, qualitative insights across your learners, and recommendations about where the team may need additional attention. And underneath that you have supporting dashboards. You can see the engagement and the practice trends, how the different scenarios are being used, and how performance is developing across the team. And you can also drill down into specific practice scenarios to see where the performance lies for each individual scenario that you've created. So for example, here we can see with our limited amount of learners where the all learner initial averages, and then after some practice where that's expanded to. The important part is that now your enablement team has the evidence it can use to decide where more practice is needed, and that's what we strive to make easier for you. So now instead of just saying that, you know, 75 sales reps have completed your training program, you can answer those types of questions that I mentioned. Where's the team consistently weak? Which school skills are improving across multiple attempts? And one small caveat that I always raise is that this scoring doesn't replace manager judgment. It's much more consistent than human raters, but what it does is give managers, your sales enablement teams, more evidence than completion records or the data that you typically have now about whether your reps are actually ready to perform. Then the question is, does this work in practice? And here's one example of a deployment we did just last month with a major Japanese automaker and their sales teams in the US. And they had really just the classic challenge. They had more— they needed more opportunities for their reps to practice customer interactions, and they needed that to be more consistent. So working with us, they created 57 customer simulation scenarios. Those were customized across different personas and different customer situations. We did that in a matter of 1 week to finalize everything and test everything. And that breadth was important because the reps that were practicing weren't practicing just for 1 type of conversation. And their training program had traditionally run for 1 week every year. But using the roleplay, they were able to reduce that to 2 days. So they put more emphasis on the targeted practice element. So even as they reduced the training time, they instilled more opportunities for practice by using our roleplay chatbot. But the number that I, I find most interesting and actually the most satisfying is the one on the right, which is that more than 50% of the participants in the training continued to use the simulations even after the training program had ended. Ended and nobody was requiring to do them, requiring them to do that. Nobody was asking them to do that. We found that they continued to log in and practice on their own. And that to me matters because really any tool, any platform, any system only works if your reps see the value in doing it. And that was clearly the case in, in this deployment. So if you remember one thing from this session, I would just make it these— well, maybe these 3 things. I think the opportunity is to take sales situations that matter in your business, but if you are going to implement this type of solution, you need to know that you can deploy realistic conversation practice quickly, number one. Second, with enough control. to reflect the way your customers really behave, right? We give you control over every aspect of the scenario design. You can change it on your own at any time. You don't rely on us to update the scenarios. You have that complete control. And then third, do you have the ability to set the evaluation criteria to know that what your reps demonstrated is actually going to lead to better performance on the job? So that's the progression that we looked at today, and that's what we're trying to help organizations do with UMU Roleplay Chatbot. So if you're thinking about a particular sales motion, a particular training program, or a conversation practice where this could be useful, then I'm happy to talk afterward, or you can schedule some time with us and we can show you how you can turn just that situation and how we can help you turn that into a practice scenario. So that's it for the presentation. Thank you very much for your time today. Happy to chat or take any questions that come up.

Thanks, William. You know, we did have a question in the chat about the feedback and who's coming up with the feedback, where is it coming from, or in the assessment. And so, like, how is that developed? Can you provide any information on where it's coming from?

Right. So the type of feedback, it's not generic. It's not, did they Did they handle objections well? What we saw there, which was you're determining what the feedback is according to which criteria. So if, for example, you have a scenario and you really just want the feedback to be focused on how well did they open the conversation, did they handle the introduction aspect well, did they set the agenda well, that feedback can be specific to that element of the sales process, but it can be specific moreover to your specific sales methodology. So of course, we have the expertise to build the type of evaluation that we can recommend, but many clients come to us and they say, well, we have our sales methodology, we already do live role plays, we just want to import that into your system. And our answer is, oh, you can do that very easily. You set the criteria and then, you know, the AI is generating the feedback, but you know what actually happens in the background is the AI is determining Well, the behaviors that are demonstrated in the conversation, let's map that to the criteria, and that's how the feedback is generated. And that's what makes it more relevant to a salesperson rather than something generic.

Sure. Yeah.

Stephanie, you do a lot of role plays. What have you learned about how many times do salespeople need to practice one skill before they become proficient?

Well, it depends on the person, and it depends on if they're engaging in purposeful practice. But we did look at over a million role plays, and it was about 30. And so that takes into consideration the people that are, you know, doing a lot, but they're not really getting anywhere. They're maybe not focused on learning. They're just trying to get through it. And then you have people that can hit mastery in much less time if they're really focused and they're motivated and they're paying attention to the feedback and they're implementing it. And, you know, that they want to get better. But yeah, 30 is on average is what we saw from the data.

William, what's your experience?

Yeah, we have similar data client by client, but just picking up on Stephanie's point, it really is a matter of the design. And that's where we work with our clients to determine Right? Is this designed well? Is the practice mandated, or do they need to reach a certain threshold because— to determine whether the amount of practice opportunities are enough? So that can be determined really on an individual-by-individual basis. I think what Stephanie, I think, would agree with, and what everyone would agree with, is the more practice opportunities, the more people are motivated to practice, the more they are going to actually develop their skills. And that's what these types of solutions offer. You know, as I said at the beginning, there's many different types of solutions for role play. What we've seen though is that, as I shared with the one example, and we've seen across many examples, we see the learners going in to log in. If they see the value in it, that's what's going to increase the practice opportunities. That's what's going to increase the actual performance. So that's been, I think, the biggest reward for me is being able to see that. This is not a value just to the sales enablement team to say, hey, we implemented practice. It's the reps themselves that are saying, hey, give us more. And that's been, yeah, that's been really important, I think.

How important is tone of voice in roleplay?

I think it's an important element. I think this is where the different types of solutions are more commoditized.

Right.

The roleplay avatars are only going to get more and more realistic. So I think the marginal impact of, are they blinking a certain amount of times or are their interactions indistinguishable from a real person? I think that's valuable. You want people to feel like they're simulating a real conversation to a certain extent. But I would say the marginal gains of those Those developments are limited. So this is where, you know, fundamentally the scenario design is more important than the avatar. This is my perspective, but I think, you know, we're going to continue to see the technology develop and those things become a little bit more standardized over time.

Yeah, I think you're right. And, you know, it's really exciting to see the technology evolve And we look at managers who still feel like they don't have enough time training their people. And even if a manager is training their person, it's so difficult for them to be able to understand the customer and be like the customer for each of their people with enough time. And so this kind of technology getting in and allowing people to be able to practice is huge. It's huge for organizations.

So how do you use that in coaching?

Oh, so is that a question for me, Gerhard?

Yes.

Yeah. So we don't specify how our clients use this, but we give them the tools to do this. Some of them choose, depending on their training scenario, to have the frontline coaching really be handled by the AI in this case. But that's really the minority. Most of the times it's having the managers have easy visibility into their teams. So just in terms of the backend structure, a manager can have visibility to the scores of only the people that they're directly managing, and that gives them the visibility into who's struggling and where. So it's an element where we, we don't determine the coaching that's done. We help our clients on an individual basis say, Here are now the resources that you have on the basis of this practice to do coaching in the way that works in your organization.

And what is your best practice in terms of designing the right length of the roleplay for the new students and also for experienced salespeople?

Yeah, that's important. And I think it really varies on the type of scenario. All sales scenarios are different. Some people may, I think, Rather than answer your question directly, you know, we, we have some clients who are saying, you know, we're training our, our cold callers who just have a very brief, limited amount of time. So the scenario can be determined to be fixed at 2 minutes in length. They only have that amount of time, and that's realistic for that scenario. Whereas somebody who has a 30-minute meeting or longer with a potential client, needs to have more time. So I'd say it's hard to answer specifically, except that that's an element of control that's built into our system where, you know, your sales enablement teams know your sales situation best. You can define what that length of time is.

Um, and the other question, since you're talking about cold calls, what is your recommendation For opening a cold call successfully?

Oh, that's great. I think you just want to establish rapport right off the bat, but it's become, I think, harder and harder over the years. But that's typically breaking through that, just getting them not to hang up the phone, right? Can I have 1 minute of your time? Making the call very Easy and accessible. And I think even the best cold callers will know you're gonna get hangups about 50% or more of the time, but it's a numbers game, right? So it's maximizing once you have those people who are willing to have that 30-second, 1-minute conversation, how quickly you can make that pitch. I think there's no magic formula, but those are the things that I think still are going to be challenges no matter how much you practice.

Stephanie, do you have a recommendation?

Well, it's funny, you know, because we do this calling campaign at the university where we call on admitted students and encourage them to come in. And we call home, and it's usually a parent or maybe the student answers. And even though they have applied and they've been admitted to the university, they still are taken aback by the call to congratulate them. And it's like, why are you calling me? So I think it's a struggle. It's a struggle when you're doing that, but anything you can do to disarm that person that you're calling, because let's face it, who's usually getting a cold call when they want to get a cold call? You know, nobody. Rarely do you see that someone says, okay, well, go ahead, cold call me for the next hour. I'd love to take your call, right? So how do you walk into a situation where you know you're unwanted? And then, and turn that around into being wanted. Actually, I'm curious what you think, Gerhard. You've been around here giving advice for decades. When I first started teaching at Bryant, like 18 years ago, I had my students reading books from the bestsellers list on sales. And you had, I think you had 2 of them at the time.

Yeah. Well, my, The best answer I can give you is actually from Chris Beale, and I have to give him credit for that. He is the CEO of— he was the CEO of Connect and Sell, and he taught his salespeople, I know I'm an interruption, but may I have 7 seconds to tell you why I called?

So does anybody ever say, well, you just went through the 7 seconds?

It works 80% of the time where they say yes. And then he tested that, you know, may have 17 seconds or may have 27 seconds. An even number works better than an odd number, works better than an even number. So salespeople are playing with this and They get results. It takes a lot of confidence, and, uh, I think it takes some acting skills as well.

It's interesting that you said that because, um, there's a bunch of negotiations research out there, and it's looking at odd and even numbers. And an odd number is seen as the end of a negotiation, like you're not going to continue to ask for more, versus an even number, it seems like there's, there's a place to go. Or I wonder if that research aligns. I don't know.

Probably. What's your thought, William?

I haven't seen that. I'd love to read that research. Makes sense intuitively to me, but that's why I'm here at a conference like this, to, to learn about these, uh, all of these tips and tricks and new knowledge.

Well, I'm just mindful that we are 3 minutes over time, but William, that was a wonderful presentation, very enlightening, and congratulations on your progress, uh, with Helping elevate the profession of selling. And I encourage everyone to go to William's website, umu.com.

Very simple.

Yes.

Thank you, Gerhard. I appreciate the time today.