#201 – What it’s like to train an AI on The Sales Mindsets

 

 

Full episode transcript: 

​Please note that transcription is done by AI and may contain errors.

​

 

Jesus: Hello. Hi, Phil.

Phil: Hi, Jesus. Nice to see you.

Jesus: Yeah, nice to see you. Yeah. So, um, great. We have, uh, the chance to have a few minutes today with you to discuss about some our new baby that is gonna be launched soon.

Phil: Yes.

Jesus: Um, just for those that don't know, um, I mean we, we, we announced in the DST and we've done some announcement on, on the social media, but, um, we, we are launching.

Something that is gonna be completely new for Consalia. So it's an AI software as a service product? Yes, it is. And let's, let's go into it. Let's go into it. So it's what we call sales Mindsets Intelligence Studio. So it's powered by the Aviso AI that, uh, Aviso is, uh, is a revenue operations platform, uh, based on AI working for the last 10 years with most of the most prominent, uh, companies.

And, uh, we just partner with them to create, uh, something in combination with our sales mindsets, right? So not just, uh, our academic research that Phil has been living, uh, but also, um. Um, a lot of, um, you know, information outcomes, results from our consulting projects as well. So, um, this is gonna be officially launched by the end of May.

Very soon, we are gonna be, uh, having a, a landing page available in our site. Um, some of you guys that you're looking into the podcast now, maybe you'll receive an email if not. Don't worry. Just in the, in the, in the Consalia landing page, you will be able to just, uh, register. Uh, we'll do with a small event in London, but they will follow another events in Madrid and, and some other places, right?

Because we have customers from all over the world.

So Phil, I mean, um, everyone will be thinking about what is the sales mindsets intelligence studio, right? We, we, mm-hmm. We just started to connect, uh, all the intelligence we have through this 20 years, more than 5,000 assets in, in various topics. And, um.

We train an AI for the first time, so we move from humans to the ai. So, um, why do you think we, we, we, we, I mean, of course it's something that we work together, but in your view, what is, what, uh, why Consalia decided to move into this space?

Phil: I think for me it, it started, um, beginning of last year with, uh, the huge amount of focus on AI and what it could and couldn't do.

And so we began thinking that, uh, some, a topic that Consalia really needed to. To kind of focus on as a topic, and it kind of made sense for us to go back to our core, uh, research and ideology around, around the sales mindsets. And we had this, uh, concept emerging, which we called the Consalia brain, for want of a better word, and looked at developing our, um, our own sort of proprietary AI agent, if you like.

Um. And we then began to realize the complexity of, of, of developing agen AI agents. And we had the good fortune to actually, uh, link up with, um, Aviso and Ofer Zilberman, who is their chief product officer. Who had already known about the research and who had said, why didn't we partner with them? They've got the AI technologists, we've got the knowledge of the, the sales mindsets and all the research that's gone into that.

So kind of that's. How it started. And, uh, since then we've been, or I've been doing quite a lot of work, uh, as you say, training an AI agent. And, um, it's been a really interesting sort of exercise that we've had, and it's amazing to see, um, how it's progressed and also what its potential is looking into the future.

So we're, we're really excited by it.

Jesus: Yeah. Yeah. I mean, uh, yeah. We'll, we'll, we'll, uh. We'll talk about the process in, in a, in a minute, but, you know, uh, uh, so I think just to, to, to put up another 2 cents on this, I think, I mean, we were lucky that Ofer Zilberman became this Chief Product Officer of aviso, but we were lucky as well that Aviso is one of the top, you know, uh, KA technologies on AI for sales.

That is looking into an end to end, right? Because, uh, we see that many technologies are just calling certain parts of the sales process, but you know, they are really working end to end. So, Phil, uh, I mean, of course we've been together, uh, during this journey, but, uh, I'm pretty sure that many people will love to hear from you.

How, how do you perceive the development of, it's a, it's a development, right? It's a, we started with the training of this brain.

Phil: Yeah.

Jesus: Right. Uh, how have you, so how was the process of training the, the, the tool for you? I mean, uh, I think we started just, uh, sharing some material we had. But then, I mean, how, how do you feel about that?

How, how was the process?

Phil: Well, I think the process is really interesting. And just, just to, um, just to set some context perhaps before we go into that, is that essentially, you know, 15 years ago it started with the doctoral research and to cut a very long story short, identified certain sales mindsets that, um.

Customers were looking for from salespeople, but rarely saw. Um, these were around authenticity, client centricity, proactive creativity, and ful. Audacity. We also saw a lot of, um, uh, mindsets which clients saw that they didn't like, which was around manipulation, supply centricity, complacency, and. And arrogant.

Overt, arrogant. So, uh, since then we've been working with control groups and we saw a direct correlation between living the positive mindsets and sales performance. Yeah. Mm-hmm. And so, um, and we've been working with many sales teams over the years. Typically in face-to-face kind of, uh, situations, coaching and training around these four differentiating mindsets.

So the data that we've acquired has been significant. Um, you know, not only the doctorate, but books have been written on it. Um, podcasts have been talking about it. So we, we were sitting on, and, and also we had quantitative data through the sales mindset survey. So we have, we're sitting on a huge amount of source information, all supporting this notion that if you live the right mindsets, will sales performance follow?

I think what's been exciting now is sort of translating that into an AI environment. Now, is it possible to be able to train, uh, sort of an AI agent to be able to look at a conversation and from the conversation, be able to interpret either the positive or negative sales mindsets? Um, and so for me it's been a very interesting, uh, case of, of, uh, trial and error, if you like, over a period of time.

It's, uh, I don't know, been four or five months, maybe, maybe six months actually working on it now. Uh, sort of trying to, or getting to a point where one felt comfortable. That the accuracy of observation in the AI agent, um, was where we wanted it to be. So that, that for me has been a, a very interesting kind of process.

And also the reason why I personally, uh, just wanted to do it. I didn't want anyone else to be involved in the training of the AI agent. So it was a one-on-one Jesus, me and the AI agent working together, trying to work out if the age AI agent was performing or not performing according to the standards that we would like to see.

Jesus: So how do you feel at the very first, you know, there were, uh, several phases in the project. There

Phil: were,

Jesus: right?

Phil: Yeah.

Jesus: How do you feel in the first phases where maybe the agent was not so good responding to expectations?

Phil: Well, there, I, I think there were, there were two things right at the very beginning, which.

Which is interesting. One was that when we looked at training the AI agent, um, we, we, we needed to have data to look at, you know, and the, the original data were fictitious conversations going on between sellers and buyers. And, um, it just didn't feel right. Didn't feel that, that, you know, that was the right.

Kind of starting point. So I think that one of the things when we sat down and looked at, at the initial cases, uh, where the AI agent was looking at analyzing conversation, we felt there was two things that were a challenge. One was the cases they weren't real enough, and secondly, the AI agent, um, was. I didn't know what's the right word, wasn't quite getting it.

Jesus: Mm-hmm.

Phil: You know, it took it and, and it, it, it's interesting and some, some of the, some of the times it might have got it or, you know, I know there was a lot of, there was a lot of questioning I think, around the way the AI interpreted the conversation and correlating it back to the mindsets that made us, or made me question.

Um, whether or not it was possible, but there was enough in those early stages to suggest that actually with a bit of, um, patience and with more coaching and support and to be honest, to Aviso and the, the team at Aviso who are brilliant in supporting us on this journey. Um, you know, they said it's gonna take.

Seven or eight iterations before we get to a stage where we feel comfortable enough that we've got something that is, is, is accurate.

Jesus: Yeah. Yeah. I think you're completely right. Right. So we are, the problem we are launching now, it's, uh, it's focused on, on, on soft skills, on sales mindsets.

Phil: Correct.

Jesus: So it's much easier to train an AI agent when.

They have to follow a process. They have to follow a certain amount of information you need to fill and account, things like that. Right? But here it's about evaluating how a human is interacting with another human.

And providing. A, a feedback that sounds like a human. So it's, it's one of the most complicated scenarios you can think about, right?

Phil: Yeah.

Jesus: And I believe that, uh, in the first, uh, in the first, uh, times, in the first, uh, phases, I think we look into it as, as, as, as it was like a baby. We need to train. Right? And I think we find this in many companies that sometimes. People have the tools, but they don't get the training or do, they don't do the training to the ai.

I mean, and as much as you train, as much as good data and clean data, you have the best, the best will be the outcome. No. So I think we had the patience to go through the different phases. And I guess that there, there was some moment where you started to see it click, right? A click that, uh, things started to work

Phil: well, Sort of having gone through the case material, for example, the basis of which the agent was then analyzing conversation and realizing actually you've gotta work with live data, uh, or real conversation. The next phase was really quite interesting because we were able to get live, live conversations. Yeah, so these were live conversations taking place between.

Um, sellers and, and buyers, you know, uh, real conversations, customer support, conversations, and so on and so forth. Um, when you, when you work with live conversations and I'll, I often come back to the. Reference to Professor Liz Ko at, uh, London School of Economics, who's a conversation analyst, and she talked about the Methodist of language.

And I remember Liz had spent some time, I think, in Spain, uh, sort of trying to work with technology companies to help. Her approach to conversation analytics transcribed to a tech technical environment, and I can understand the challenge that Liz faced because. When you have conversation, you have uhs and ums, and you have interruptions and, and pausing and so on and so forth, and how could you expect an AI agent to pick up on the subtlety of language?

And that, that, that was interesting. I mean, even, even humans can misread. A pause, you know, even humans can misread and own, um, you know, they can interpret that as not listening or, uh, where the opposite may be true. So it's actually, it, it, when you started to look at the, the conversations taking place, um, that the AI agents were analyzing, um.

You found yourself having to train the AI to deal with the messy bits of conversation somehow?

Jesus: Yeah,

Phil: to train it to say that if there is, um. If there is an, uh, a pause or an No, it doesn't, it's not necessarily negative, you know, it might be positive. The word okay. Might be appropriate because the word Okay.

Might be encouraging a customer to talk more, which is what you need to do when you qualify a sales opportunity.

Jesus: Yeah.

Phil: Um, so what I found interesting in the early stages when we were looking at the conversation analytics was. Was, um, working with the agent in the inter their interpretation across the sales mindsets of conversations and.

Questioning their interpretation in, in a feedback loop. So what I was doing was taking a conversation. I would either accept that their interpretation was correct or not.

Jesus: Mm-hmm.

Phil: In the early stages, there were many knots. Correct. And I was then. Writing some sort of commentary about what I thought would be a better way to interpret the, you know, that aspect of the conversation.

And there were many conversations that, um, that we were looking at. You know, there were many. Liz, Liz Doko talks about turns in her conversation analytics piece. Each turn is a, is a, is a, a buyer and a seller communicating. So if you start to look at how many turns did we actually look at when we, when we did the training, it was, I don't know, I haven't counted them many, many different terms.

Um, but what we wanted to see was a sort of progression. You know that the corrections would become fewer and fewer. And it was interesting over time that we began to see that actually, um, the AI agent was doing a much better job of, uh, of interpreting back to those sales mindsets. The conversations, um.

In an accurate way, which made us more confident that we had something that would be quite interesting for, for companies to follow up on.

Jesus: Yeah. Yeah. I think, I think, I mean, uh, we, of course, we don't want to unveil everything that will be launched by end of May.

Phil: Yeah.

Jesus: Uh, but, uh, you know, the, the ability to have a 24 by seven.

Uh, I mean, the fact to have a 24 by seven, uh, kind of intelligence on the sales mindsets, uh, helping you is something that is, is gonna be probably very, very recognized and valued by our customers. So, um, I. I mean, I know with, uh, during the different phases and the stages, and I know that you've dedicated, uh, a lot of hours for, uh, training.

The training our, our, um, solution, our joint solution with Avis, right? So we've been working with the, with the Aviso team, which we need to thank you all, everyone, uh, because they've, they've been ama amazing with us and a lot of patient because we have. Uh, we are very strict and we are very rigorous, uh, on, on doing things.

So, um, as it cannot be or there any other way. So what do you think is the academic level? I know it's difficult, right? But we've always, uh, make a joke about, uh, having a, an AI to go through a PhD or to go through a. Through our Masters? No. So, uh, I mean, is A BSC is more or less or what, what, what do you think is, I mean, we know, I know that we are happy to launch to, to our customers.

Yeah. And, and, and potential new customers because it's ready Right. Has been trained enough. And of course we will continue working on it to improve, uh, on Yeah, on a daily basis. But what, what do you think is, is the level right now?

Phil: I would say probably is at that level if you wanted to give it an academic, uh, sort of rating.

But, um, I would say that. Um, the, the tool is only gonna get better the more it's used, you know, the more, more it's used. So there's a lot of, um, there's a lot, lot more to come from the tool. You know, the correlation of the, the linking of the outputs of the. Of, of, of, of the intelligence you've got around the mindsets with performance, you know that there's so many, there's so many things that are, are going to surface from this, which I'm really excited about.

And so, um. You know, would it, would it rate, is it possible, I mean, we did talk about whether or not it's possible for an AI agent to, to get a, an MSC in agen analysis, maybe, whatever. Yeah. Um, but you, you know, I, I would say we're still way off that particular point, but, um. You know, who knows what the future's gonna hold.

I, I've been really amazed at it's not just the, the, um, it's not just the accuracy, it's the, uh, it is also the suggestions, the coaching aspect. It's quite interesting, you know, that it's one thing to say yes. Uh, this is not as client-centric. as It could be, uh, it's another thing to say it's not as client-centric as could be.

And perhaps if you phrase the question in this way, you could, you could, you could get a better result. So I think that, uh, the AI agent is taking it to that level. It's not just analyzing, but it's actually suggesting what could be a way to improve, um. Uh, to, to, to improve a response to a customer by reframing, re re-asking the question in a slightly different way.

Um, so it's, it's really interesting.

Jesus: Yeah. I, I think, I think you're fully right. Right? So, and at the end, uh, once, uh. The solution is deployed in, in the customers, if they have a hundred sales reps. If they have a thousand, 10,000.

Phil: Yeah.

Jesus: I mean, it'll have all the feedback. I mean, it will get more data that will serve as training and will be more accurate because they will be, you know, this, um, intervals you mentioned, like Hmm, et cetera, or things like that will be different depending on the culture and depending on the geography, depending on, on the company.

Right. Yeah. So, um, yeah, it's, it's, I mean, I think, I think it's, it's. We, we just pushed the, the, the solution to a level that is, it starts to add value of using it. Yeah. And this moment where, of course we'll continue moving forward with version two, version three as, as any software tool. But

Phil: yeah.

Jesus: Uh, in the moment that this is deployed into a customer, then.

Uh, you know, uh, it starts to gather all the information about how people is performing. What are the best performers doing, how they compare to your performance, how this translates into the sales mindsets. I mean, things we will get much deeper into, into the launch event, um, that is scheduled for end of May, but I think it's, it's, it's gonna be quite amazing.

Um. So Phil, how, I mean, I know that all your, or most of your life, or all your life has been, you've been either doing training or consulting with humans.

Phil: Mm.

Jesus: And now it was a machine on the other side. How do you feel about that?

Phil: Well, I feel quite good about it 'cause they don't answer back.

Jesus: They were not complaining.

Phil: They,

they don't complain. They just listen to what you have to say and then work on it, come back with something else. So it, it, it, so from that aspect, um, I think it was quite, it was quite an interesting process and, and, uh, and so I was. I, I, I was really interested, um, to see its evolution over time. You know, it's, it's, to what extent does it take note of suggestions?

Uh, what, um, and its ability to. Uh, take on board those suggestions and listen, you could argue they do a better job sometimes than humans. So, as

Jesus: humans. Yes.

Phil: Uh uh So yeah, I didn't find there's a problem at all, and I've just found it, I suppose is I found it more. you know, it's like an assistant in a way.

You know, you've got someone who's there to assist you. In my case, there's an agent here who is assisting me or assisting Consalia in being able to provide, uh, an accurate, um, account of the sales mindsets. And not only that, also. Uh, provide insights at a, at a individual contributor level, insights into what they could do differently.

But like you mentioned earlier, Jesus, when you start to look at the, um. The opportunities of analysis at scale, um, then you've really got some, a very powerful tool. Um, the other thing that I'd just like to, to, to, to mention is because I think you've got scale and speed. You know, I think there are two things, and the scale is volume and speed is how quickly you can, you, you can get, get data.

Um, I think the speed element, of course, it it's instant, you know, being able to get analysis of conversations, but it, I'd just like to talk about the difference between conversation analytics and the mindset intelligence, because the thing about the mindsets is that. The mindsets are, if you like, the operating system that underpins how salespeople will behave either positively or negatively according to customer feedback.

If you are able to develop a tool that can help address the system of selling the way you think, then for the. Contributors who are trained in in that system and they're aware of it and they're coach to it, the chances of you getting positive impact from that are going to be much quicker than dealing with something at a.

As a symptom. It's the, you know, it's the difference between giving someone a plaster or giving someone a, a Nurofen if they've got a headache, or being able to understand what's causing the headache in the first place. If you can get to what's causing the headache, you solve the problem, the headache for the future.

If you give them a nen, you give them some peace of mind. You know, they, they don't have a headache, you know, 20 minutes later when it starts to kick in. Um, and I think this is why, where the speed and scale dimension of what we're trying to do here with the intelligence studio, becomes interesting.

Jesus: Yeah. I think, I think also if I can add this also the, the option to customize so you, you get customized, fully customized feedback so that that contributes to this feed. And the other thing that I think is also important is that. As you almost mentioned, right? So, um, uh, it's a machine, so it's completely objective.

There is no subjectivity or you know, it's objective and it's based on, uh, tons of data and tons of training. Yeah. So, and, and normally people react differently to feedback coming from machines. That feedback coming from, from humans. Yeah. Depending on, depending on things. Right? So it's a 24 by seven personal, impersonal environment.

Um, yeah. And, and also, uh, using coaching techniques as well, right? So it's not just an advice. So it, I mean, we'll embed much more, but, um, so it looks like you, you enjoyed the experience. So, uh, do, do you think, uh. You will do it again in the future?

Phil: Not

Jesus: if we have another,

Phil: I'm not sure the next iteration will, will, will, uh uh, I'm not sure what the, the next in inter iteration will be actually.

'cause I, I, I, I'm going to look forward to, The possibility of studying the data at scale, I think. Mm-hmm. And, uh, whether that's going to involve training the AI in a different way, or it may be that the data suggests that we need to change, perhaps the, when we start to correlation between sales performance and mindsets, it may, may have an impact on.

The nuance of the mindsets that we're talking about. You know, maybe they'll change as a consequence of that, so therefore would need more.

Jesus: Yeah, maybe they did. Through this, through the studio, we are able to detect, uh, trends, new trends or new changes, new sales mindsets. I also think that we focus a lot on the technology.

Yeah. Uh, for the humans, but now maybe we need to focus on the humans for the technology as well. Yeah. People using it and people adopting not just the, the studio, but also some other AI tools. Right. Yeah. So I think, I think that there, there's a lot of work on that. Uh, I know I'm conscious that we have like, uh, like 30 minutes, uh, like a cap.

Yeah. So, uh, is there any other final thought you want to mention or.

Phil: No j just that it's, um, you know, it, it's an interesting, it, it is certainly a, a really interesting area to explore and for any organization that's has the ability to be able to have recorded conversations, it's a, it is a fantastic.

Opportunity to be able to link that to sort of an academically proven framework, uh, around the sales mindset. So, um, I and I, I think it's, uh, the partnership we may have with future clients as we start to explore the power of the mindsets and, and sales performance. I just think it's going to be a very exciting time ahead, you know, for what we're doing here and what we've started.

So yeah, super excited that we have got to this stage. It wasn't quite the journey we thought out at the beginning. Jesus was it we, it wasn't quite how we thought it would be.

Jesus: I think, I think the, the, the main concept or the framework was similar, like, you know, to have like a brain. Yeah. With, uh, all the experience, knowledge, we can train on all our, you know, yeah.

We have a lot of assets, a lot of, you know, research. A lot of projects we've done and a, a lot of experience, right. Recorded, let's say. Yeah. Uh, and put an intelligence to work, uh, objectively. 24 by seven. That can scale up because we, you know, at the end, a coach, a one-to-one coach, we feel is quite or has mathre.

Okay? We have limited time and limited amount of time, right? But at the end of the day, this is like making a small avatar of us.

Phil: Yeah.

Jesus: Helping them, right? Uh, maybe in the. Maybe, maybe now, I don't know, maybe better than us. Who knows? Right? So it's, it, it, it's, it's where, where the AI comes when, when you really focus and dedicate and, and, and, and focus on something and, and train and, and do the right resources and time.

And this is what we've done with all the care, right? So, um, I believe it's gonna be very well perceived by many people, you know? And yeah. And I think we've really trained something that is unique. 'cause when we see, even, even all the. Coaching agents or things like that I've seen, uh, they are quite trained or on, on, on, let's say, on, on documents, right?

Yeah. So we are, we are talking about mindsets, we are talking about emotions, feelings, uh, behaviors. And I believe this is probably one of the most complicated kind of agents we could train. And probably that's the reason why it took some times for us that we are also. We, we like to be very, very professional in everything we do.

So it took a little bit longer than, yeah, than expected, but now we are happy and we are launching end of May, and I'm pretty sure we'll have, uh, existing customers and, and new very interesting people that will love to, to hear about that. you know, sales leaders, uh, sales engagement, sales enablement, uh, transformation people. So people that is really looking into transforming the people, the salespeople. Through, through technology and through AI and connected to the sales mindsets, not connected to, you know, to all the things that you can study on books, et cetera.

This is something that you really need to train, you really need to practice in order to, um, to excel. So, well, I think, I think that's all for, for today, Phil. So thank you. Thank you so much for, I'm pretty sure we'll have more conversations on our new baby. That is gonna be launched, so thank you very much.

Phil: Thank you.

​