Episode 40: Richard Lichtenstein of Bain & Company on How to Create Value Using GenAI
On this episode
Shiv interviews Richard Lichtenstein, Expert Partner and Chief Private Equity Data Officer at Bain & Company.
Shiv and Richard discuss how PE firms and their portfolio companies can leverage generative AI to improve investment processes and create more enterprise value. Learn how other companies are using this technology to increase efficiency, deliver better customer experiences and drive more revenue, how PE firms can streamline their due diligence, and the challenges and opportunities GenAI will likely present us with in the near future.
The information contained in this podcast is not intended to constitute, and should not be construed as, investment advice.
Key Takeaways
- The current state of GenAI usage in PE-backed companies (3:06)
- Some key opportunities for GenAI to upgrade process and experiences, using contact centers as an example (6:47)
- How the next generation of GenAI will differ from existing products and implementations (10:23)
- Why getting the right data to train AI can be a challenge for companies (15:46)
- The impact of context and how GenAI deals with answering complex questions — particularly in relation to marketing (18:57)
- How Bain uses AI to increase efficiency in the due diligence phase (28:36)
- Potential use cases for GenAI to improve sales processes (35:48)
Resources
- Bain & Company website
- Connect with Richard
- On Linkedin
- Via email - [email protected]
- Subscribe to Richard's Substack
Click to view transcript
Episode Transcript
Shiv: All right, Richard, welcome to the show. How's it going?
Richard: Going great, glad to be here.
Shiv: Excited to have you on and obviously Bain is one of the biggest consulting firms in the world. So why don't we start with your background and role and we'll take it from there.
Richard: Sure, sounds great. So I'm Richard Lichtenstein. I'm an Expert Partner at Bain. I've been at Bain for 20 years. I'm the chief data officer for our private equity business. So we are Bain, the consulting firm, but we work with a lot of PE funds on various issues. As chief data officer, I spent a lot of time thinking about advanced analytics and data, and lately in the last year and a half, GenAI. And so I've been working with a lot of funds to think about GenAI topics.
I also have a team of 100 plus people and we build stuff. And these days a lot of what we're building is using GenAI. And so we've seen firsthand what GenAI can do for us and our people and some of the things it can't do. And that's been interesting as well.
Shiv: Mm -hmm. And what where do you see the biggest opportunity? Especially when you think about GenAI and and how you're applying it to your clients and different PE firms that you're working with. Where do you see the biggest opportunity areas for that?
Richard: Yeah, so I think today - let me start with where we are today as opportunity. I think we would say most PE-owned companies should start with what's available off the shelf. And today's vended solutions in GenAI are still somewhat immature, but they're getting better all the time, right? And that's the place to start. And so that's going to be applications like contact center, where there's lots of applications around co-pilots to augment agents or automated chat, things of that nature. Developers and engineering, obviously there's tons of excitement around tools to generate code. I actually get even more excited about some of the tools to generate developer documents, right? To make developers - take the stuff developers don't like doing and take that off their plate. And then I think the other big one is sort of sales enablement, which is obviously your bread and butter. But I think there's a lot of tools that can make salespeople much more productive and get them in all aspects of get them more leads, give them better collateral, give them less windshield time, more time with clients, all kinds of things we can talk about. So those are the areas where I think there's good off the shelf solutions that exist. I think then there's also more idiosyncratic opportunities that a lot of companies have about how they can integrate GenAI into their products. That's obviously a bit less mature, but we've done some work with several companies where there are some exciting opportunities there, but it's harder. You have to build it yourself, and that's obviously more difficult and expensive.
Shiv: Yeah. Yeah. And, and just generally before we jump into some of those specifics in the areas that you mentioned, how far is this curve of companies adopting this? Like obviously things that are off the shelf are earlier stage and that is definitely a first step that we can take, but how far off are we from companies developing a lot more maturity and sophistication in terms of how they're leveraging AI?
Richard: We're in pretty early innings here. I mean, second inning maybe. There are very few companies I talk to that say, we have a GenAI tool, either that we built or bought, that's in production that we're using. We're using it every day. Our customers are using it every day, or our people are using it. It's not zero. There are companies doing that, but it's really the minority. Now, a much bigger group of companies will tell you, we've got a couple of PoCs or MVPs that we've set up. We're testing things. We're seeing good results. We're excited. You know, I hope many of them six months from now will be in the first bucket, right? But right now, most people are testing and learning at best.
Shiv: Yeah, that's what we've seen as well, even on the go-to-market side, where it is something that people think about, but they're not fully deployed or it's not a core part of their strategy. And so what are some of the - you mentioned contact center, devs and sales enablement. Walk us through some of those quick wins, maybe with a more specific set of examples. We'd just love to bring that to life for the audience.
Richard: Sure, yeah, I mean - so going one by one - and by the way, the reason I came up with these wasn't just me off the top of my head. We do a survey every quarter of, I think, 500 plus practitioners of companies and ask them, what are you doing with GenAI? Those three are the top use cases that they are doing. So, I mean, again, this is based on data. We love data at consulting firms.
So if we talk about the first one, which is enablement in contact centers, so I think that the first wave of tools that you see there are co-pilots. So these are tools that sit next to the agent and they might give the agent advice on what to say. They might allow the agent to ask a question, right? So if it's a company that has a lot of products, right? The agent may not have every product at their fingertips. And so if a customer asks a question about a product they're not familiar with, they can ask this guy, what is this product? How does it work? How is it different from the competition, et cetera, right? And get answers really quickly that can help them give a better answer to the customer. It can also help with troubleshooting. So the customer calls up and says, I've got a problem. My phone's not working. My whatever's not working. The agent can use a bot to help troubleshoot that and do that faster and more efficiently. We also are starting to see early ones that can do things like interpret people's mood and sort of say, this person seems upset, happy or whatever, which can help agents, especially if they're multitasking across lots of chats and things. And then I think the last one that I've stil -l I've heard is out there are tools to help agents be more effective at things like cross-selling. So you'll have a tool that will say to the agent, you know, Hey, this is a good moment to maybe cross on - like they seem happy. We solved their problem. They mentioned that they have this other need that is - that we have a solution for. Maybe mention that solution, maybe see if they want to buy it. Agents today are not always great at remembering when to do that sort of thing. It's giving them nudges. That's kind of what's existing today. What I think is coming next - and I did hear of a company doing this, so this is again a real thing, although it's a pilot - is instead of it just being a co-pilot, you now have an agent, a virtual agent that can actually do stuff. And so I talked to a company recently that has a chat agent and the chat agent is able to actually connect into their systems and take action. And so, you know, this - I don't want to get too specific about the company, but there are services type of company. And so if you want to cancel your appointment, you can chat with the bot and say, cancel my appointment. And it's able to connect into the database and cancel it. Or if you want to reschedule or whatever, I think right now it's - It's simple stuff, right? That like that - Like if - I don't know if you could chat and say recommend a provider or something, right? I'm not. I'm not sure how good it would be, but for those simple things it's it's already seems to be working. And again, I think if you fast forward six months, I think you'll start to see more people adopting that in chat, I suspect.
Shiv: So yeah, and the good place to jump in is, let's use this contact center as our example. There's a metric that support teams have tracked forever, which is first call resolution, right? And just knowing that has a customer's request been resolved after first contact. And we've seen companies like Zendesk or people that are in the support realm build more automations and predictive systems to help you reach resolution even before talking to a rep. I think a good comparison would be like a zero click search on Google, right? And so why is that feels like just a natural evolution of a product. So where's this going? I think bring that to life a little bit because so far it just feels like what would be next if you have a large database of great troubleshooting articles, being able to search through that and coming back with the right troubleshooting help article feels like it's less so an AI thing and more just like an automation thing. And even this example of tools to cross-sell, I think that's a higher level of intelligence where you're now able to get enablement for the team to be able to cross-sell and upsell more. But even that feels like more of an automation at play. So how do you bring that to life about where the future opportunities are or what this looks like as it evolves over the next couple of years?
Richard: Yeah, I mean, I think those are fair pushes. I mean, look, the fundamental technology at work here is different from previous automation, right? These bots are capable of performing reasoning tasks and things that the bots couldn't really do before very well. But as you say, there's a continuum of progress here. It's not like nobody ever thought of the idea of giving agents a tool to make them more effective, and now they did, right? So I think your point is fair. The couple of things I would say that get me excited about this, so it's different, is the ability to kind of more personalize some of those search type things, right? So yeah, sure, you've had a, there's a theory of troubleshooting database that's existed for a while, but what's different is you can say, you know, the user can upload a photo of the device that's broken and it could just look at the photo and say, here's what you're supposed to go do to fix it, right? And we actually have seen - we actually built a bot for a company that did that and it works really well. But you can also imagine giving just much more bespoke things, right? So you can say, actually, I, you know, I dropped my phone in water, and it's not working. And then I put it in rice. And now it seems to be working a little better. But now the screen is flickering slightly. And I want to figure out if I can get rid of the flicker. You know, that there may not be a thing in the troubleshooting guide that specifically - there's not a page in that that deals with that specific situation that you have. But but the GenAI is capable actually of sort of reasoning across these different things and giving you advice that's specific to your specific problem that you have in a way that a simple like find search, keyword search type of approach wouldn't be. So I think that's one thing. Now I think though in terms of when will this really feel different for some of those other things, there's going to be a moment - okay, this is my belief. I believe there will be a moment. I don't know when it will be. Maybe it will be in 2025, maybe it won't be 2026. There will be a moment where you go from, you call up for customer support and you're just pushing zero as fast as you can, right? You just want a person as fast as you can. That's how I think all of us sort of think about when we call someone. To a world in which we say, I don't want a person to pick up. I want a bot to pick up. Like when the phone rings in one ring and a bot picks up, or on a chat, a bot connects to the chat, and that bot - I have so much confidence that that bot can answer my problem that I'm happier because this bot is going to, they're going to give me for a better first call resolution than I might've gotten with a human or they're going to be, you know, or at least they got to me right away or they'll be friendlier or whatever the situation is, you know, for my preference. But you will reach a point where you are more happy to be in a world where you're talking to a bot than not for at least a majority of types of issues. Like that will feel very different, right, because that is a big difference in how people perceive customer support. I think that moment is coming. I don't know when, but it's coming.
Shiv: Yeah, I actually believe that. And I think that changes the dynamic of how a company thinks about investing in things because immediately you're more efficient as a business because you can allocate your resources differently. And then on the customer side, they're getting to resolution faster and hence they're more satisfied. And then also as a business, your margins are getting better because you're able to create satisfaction within customers for a lower cost base.
Richard: Yeah, well said. I think it's a win. And I think the other thing about this is, as you said, there are companies like big companies in the space like Zendesk that are investing heavily here. So it may be that this is a service that you get through a company you're already working with. This may not be, I got to go find this new startup that no one's ever heard of and put my whole operation over to some company that I don't trust. This may be something you can already get from your - or will be able to get from a trusted provider, which I think also will make this change happen a bit faster to some extent because it doesn't require, it may not require a huge switch for these kinds of things to start.
Shiv: Yeah, because you also need - just to come into your data point - you also need that source data to actually be able to train the GenAI to be able to give the right answers, right? So usually that sits inside systems like Zendesk or Salesforce or wherever to be able to get those insights back to the customer.
Richard: Yeah, although I'll tell you a funny one. I was talking about this yesterday. A big client has been working on this problem of how do we build this sort of thing. And they realized as they did this that actually the documents that the agents were using were all wrong. Like once they actually tried to go in and figure out how do we build an automated version to do this, they tried to figure out what is the source of truth within our company that the bot should be using. And they actually realized they didn't have one. That they hadn't - No one had bothered to keep all the documents updated and they didn't even have a good repository. And so now that's what they're doing. Now step one is just build the actual repository for the bot to feed off of. So some companies will probably figure that out too, but it's all, as you say, for what we're - if you have a good repository, as you say, it's already in the system, which helps.
Shiv: Is that the heavy lift that companies haven't done? Like we work with a ton of clients, they don't even have basic data in order. And then on the product side, I would imagine they have data, but when they actually - in bringing it in one place so that it can actually be leveraged for the, for AI to actually be effective. Is that the big lift that's missing that companies now need to invest in before we kind of hit that phase two, phase three of this?
Richard: Yeah, I think that's probably true. I think it depends what you're trying to do. So where I would draw a distinction a little bit is, I mean, on the one hand, you need some base layer information, right? So if you don't even have, you know, the frequently asked questions document that you're going to feed off of, then you got to go build that. And that may be a task that companies have to go do, although I think you could argue they should have probably done been on top of that already.
I think then there's a second layer, which is how do you have a bot that's able to answer a question kind of out of the box effect? And in that case, the nice thing about LLMs is that because they're pre-trained on tons of stuff, they've seen many, many, many millions of documents, billions of images, et cetera. You don't have to give them a lot of training data for them to perform very well. So like an early version of what we're describing, you might need to give it 100,000, a million chats that are annotated to teach it how to be a good chatbot. That's not necessary at all. It's a good chatbot already. So in some sense, you don't need a ton of data to gather to get to good performance. Now, what I think you're getting at, which is like the next layer, and this is what Apple was talking about yesterday, right? If you saw the Apple announcement, like a big thing they talked about is we're going to add context, right? And that context is going to let us give you better answers. You want to do the same thing. Every company should be doing the same thing, right? When I call up the bot, or even if it's an assistant bot to the agent, whatever, has some way of accessing my file and getting all the context about me. They know every issue, every product I own, how I use it, what I've called before, etc. etc. And it can then talk to me in a way that understands me. In order for that to be possible, then you really have your data in a good spot, right? And probably a lot of companies don't.
Shiv: Yeah, talk about that piece a little bit because I think context is just so important and just using your example of sales enablement, you know, or like cross selling, upselling to really know if there's a true cross sell opportunity. If a human was doing that analysis and let's say it's a multi-product line business, you may look at all the customers of product line A and then cross reference to see if they have product lines B and C but then also they need to meet certain characteristic criteria that is actually eligible for products B and C. And you can very quickly, as a human spot like are they actually a fit or not a fit on a one-to-one basis, if you will, but not having that particular context, you can have a model making recommendations that isn't actually accurate. So how do you develop or train these models to have more of that context built in so that the recommendations are more accurate?
Richard: Yeah, well, so I think you're asking a couple of questions, right? There's the underlying question of, I mean - and your point is right, of, you know - eligibility grids for a long time. I did a bunch of work in the cable and telco sector and eligibility grids are sort of the bane of that world because everyone's on a different package and that exact problem is so important. So I've lived that in the pre-GenAI world for sure. What I would say is there's a rules-based approach that you can always use. But what gen.AI can allow you to do is to extract the features to go into the rules more effectively. So you can use gen.AI to, you know, I mean - and I've seen this in, I mean, I can give you another example in a sec, but where I've seen this, like, or I'll give it to you now, where, you know, I did some work recently on software in the HR space. And one of the questions is, can GenAI be used to think about matching people to jobs? And so the way people are thinking about that though isn't doing it purely generatively because that creates a ton of risk of all kinds of biases and things that are frankly illegal, let alone unethical. But what they're thinking about instead is let's extract a bunch of features from the person's resume, right? Let's have a list of features and we'll say which of those features does the person have? Then we'll do the same thing to a job description, compare them and say, where's our match?
I think you could imagine doing the same thing with the product catalog where you say, we're going to extract a set of products. We give it the catalog of products. The bot extracts a set of features from each product. There are a set of rules that say, okay, well, if you're already, you know, this is, this package is on a discount. If you're already in a discount here, you can't get an A, you can't get discounted B or if you have a, you can't have C or whatever. And then the bot is able to, but then it's still rules-based like in the old days, but because the feature extraction is done generatively it takes a lot less time to get it all into the system and create the offer grid. And when you come up with a new offer, you can just come up with the offer and it - again, we'll be able to do the feature extraction, figure out the parameters of that offer and slot it into the grid without anyone having to do that automatically. So you can be much more dynamic in how you do things and how you train the sales force. So again, I think it's AI traditional stuff that's been around for 10 years or more with GenAI enabling those models to perform better is the way that those kinds of tough problems will get solved.
Shiv: Yeah, I think that's a great example. What about situations where it's more open-ended to figure out the context? Because let's say you're locked customer databases, you can kind of build Excel spreadsheets and kind of analyze and you can kind of translate that into rules that an AI tool can then interpret and come up with its own recommendations. And I'll give you an example of something that's a little bit different than that. In our world, we do a ton of management consulting as well. And as we work with private equity firms during diligence or post close, we're often asked, what's the right size marketing budget for this company? It's a very general question. And then in order to answer that, we have to then think about, how much do they spend on events? How much do they spend on paid media? How much do they spend on content marketing? And let's say we find a company that's a construction management software business. One construction management software company might be very different than another construction management software company because they're serving different sizes of customers or different types of verticals, or maybe they have a different feature set. And so very quickly, like if I was to just go into a general, like even using Chat GPT as a basic example and ask how much should a construction management software company spend on marketing, it would not give me a good answer. There's so much context and nuance there. And so how do you develop that in those instances so that you get to a better answer that is actually applicable to the user for what they're trying to figure out?
Richard: Well, it's an interesting question. So we've been experiment - I mean, so first of all, today, I - just to be clear, like, I don't think any of the stuff that's out there is able to answer questions as complex as that, you know, off the shelf. I'm sure, you know, you can build flavors of it. And I mean, I tell you a little bit what we did, and then I'll tell you kind of what I think the ultimate solution will be. Cause we're building something like this for a client so I can at least get you a basic idea. So first of all, what we're doing today is, when someone says, you know, I want to figure out if this company has the right marketing budget or if they're spending the right money on different types of channels of marketing. We have a tool we built where you type in the company, it says, here's some competitors. You can choose the list of competitors, whatever. Then you push a button and it will then pull data from a bunch of different data sources. So we'll pull in web traffic. It'll pull in SEO rankings, SEM rankings. It'll generate a bunch of keywords that it thinks are relevant and then search them. It will look at display, advertising, spend. It'll look at a bunch of these metrics and performance metrics and pulls from a bunch of, I think, four or five different third-party data sources that we have access to. Then it puts them on slides. So it will just make the slides. Now, it's not using a lot of generative AI for that. GenAI does the taglines. It's using traditional automation to make the slides. But it makes the slides, makes the taglines. Now you've got a piece of material you can look at that says, well, I'm looking at this company. I'm looking at some peers. How's it doing on these different metrics? And you can at least start to have a bit of an answer to say, well, they seem like their web traffic's a bit lower, so maybe they should spend a little bit more on driving traffic. And look, their SEO isn't very good, so maybe that's why their web traffic isn't so good. And maybe that's a place for them to invest. Or they're spending a lot less on display ads or less on social ads. You can come up with a story of why they should be spending more or spending less, and you can benchmark it. That's kind of how we think about that today, which does not translate. That's obviously not as precise an answer as a number but it allows you to get closer to where there's gap.
Now, that's how we do it today. And that'll get better. I think in order to be able to have the bot then get to the next level and get to an - in order to actually answer your question, right? That's an intermediate answer, but to actually answer your question, that's where I think multi-agent approaches are really gonna come in. And we've been playing around with multi-agent stuff.
We were working with one fund on a multi-agent bot that can help them look through documents from deals and better synthesize them and understand what's in them and answer super nuanced and complicated questions. And we're seeing pretty good results. I wouldn't say we're there yet. I mean, it's still in the very early stages, but we're seeing promising findings, I think, from a multi-agent approach. And so the multi-agent version of what you just asked would be able to say, OK, I was just asked whether company X, how much they should be spend - you know, this construction software management company, how much they should spend on marketing. So I'm going to have a couple of bots. I'm going to spool up now a couple of bots. I'm going to spool up a competitor bot. Competitor bot is going to go see what it can figure out about marketing budgets for other companies, which is more or less kind of what I was describing. You're going to have a, you know, performance bot. Maybe they go pull the performance metrics right along the lines of what I was describing. Then maybe you're going to have a, you know, an agency bot that's going to kind of go through and look at publications from different agencies and try to understand if there are any benchmarks they can find among agencies that suggest what company, SaaS companies of a certain size should do or something, et cetera, et cetera, et cetera, right? You know, maybe it's pulls up those three, maybe there's fourth or fifth bot that I could come up with if you gave me five more minutes of rambling. And then it gets the answers, right? So then it goes and breaks those questions down to a point where eventually it's just looking for facts. It finds the facts and then it should try to synthesize it together and try to say, you know, here's an answer. At this point, will the answer be good? Probably not. But that's, but there'll be a point where the answer is good. And that's, that's the future of where, you know, you know, we're, that's the future.
Shiv: Yeah, yeah. It's almost, you kind of have to go through all the stages in order to get there. Like we have a ton of automations and internal benchmarking that we've built out to handle these things and, automating it to the point where we can even like you described, like a initial readout, if you will, is definitely step one, but then getting to a point where the system can actually come up with an accurate recommendation. It might be multiple iterations, before you get there, but it's interesting to hear Bain's perspective or your perspective on that as well because it's kind of like tackling the same problem from different sides.
How do you see this when you are in diligence processes? How much of this type of analysis or how much are you bringing in AI tools to help you analyze target investments for private equity clients that you're working with?
Richard: So I think one of the tools that - there are a few tools we built at Bain that have been become very popular. One of them that I think is a good example is a tool called Dialog, which you use to synthesize calls. So often when we're doing a diligence, you know, everybody does a lot of calls and diligence, right? So we might go talk to 50 customers, 100 customers to get a perspective on what they think of the business or what they think of a competitor business or likely to switch, you know, the usual type of questions. So today we have this tool and what will happen is you do the calls, you get the transcripts. These days there's still a step where you have to upload them to the tool. I think we're working on making the workflow more seamless so that they just flow right in, but you know, these things take time. But then it will allow you to synthesize these calls and slice and dice them in all kinds of interesting ways. So one thing you can do is you can have a set of questions that are priority questions for you and it will just show you what did everybody you talk to say about that. So what did you know what were the - what were the different ones. You can also see things like you know it'll do summarize a call or clean up these notes so that I can send them to someone else where it'll do those things. The thing that's really neat about this is - And I think this is why, by the way, this is a tool that's become very popular, is on the one hand, it is able to save associates a lot of time. Our AC - what we call our analysts, our consultants, right - for them saves them a lot of time because they're tasks they were doing, synthesizing, et cetera, they don't have to do anymore. On the flip side, the partners though can get a lot more plugged into the work. And that's what's very exciting. So, you know, if I'm a partner - I mean I'm a partner at Bain, I'm not on a lot of those hundred customer calls. You know, maybe if I join one or two, that's great, you know. But, so as a result, the way I'm digesting information is it's summarized. And often it can even be a summary of a summary, right? What I can do now is I can go into this and just search and see what people said, right? So if there's a question about pricing, I can just start - I can say, what do people say about pricing? And I can start seeing what people said. And then I can say, well, did people say those price increases were like-for-like pricing? And see people talking about changes and like-for-like pricing. And I can see what people actually said. And I'm just that much more involved in the answer in the way that I was before. So I think that has been a big hit. So anyway, I can go on and on. But that's an example anyway.
Shiv: Yeah. Yeah. And, and there are tools out there like Gong and Chorus and things like that, that do record calls and turn them into insights. And, and so do you see this as like another level beyond that in terms of what's possible as you're ingesting the data in a different way?
Richard: So yeah, I mean, it's an interesting question. So I do think that for us, the reason we built our own special platform is because, I mean, we do - I don't know if I can say the exact number, but well over 100,000 calls a year. And so a lot. So for us, optimizing the workflow for the specific call, like for the exact workflows that we at Bain have was worth it to us, right? Because if we can summarize the call the exact way we want it summarized, or clean it the way we want it cleaned, or allow it to more easily answer the questions we ask, that saves a little bit of time. And that little time over 100,000 plus calls adds up to a big deal. I think for most companies, they're not doing a volume where they need to build their own tool. I agree. There's lots of tools that you can use. I feel like half the meetings I'm on these days, there's a little bot in the corner taking notes. I think that's great, and those tools are great. I think it's where I would say - and then honestly, you can just use them to summarize the specific calls, and then if you want to get insights across multiple calls, I mean, you can frankly throw them into ChatGPT or Claude or Google and just say, here's 10 customer calls that I did. Synthesize across them, point out where there are differences between them, et cetera. It'll be able to do all those things, right? So. But it's more than if you're just doing these things at scale. At some point, it becomes beneficial to have scale tools that are more bespoke.
Shiv: Yeah, so that's great. And I'd say that's more of an example of making internal Bain processes more efficient as you're looking for insights inside these companies. What about in situations where you're looking at companies and trying to analyze their performance and you kind of need to do an outside in analysis or you have access to a management presentation or a data room, where are you leveraging AI there to get better insights out during diligence processes?
Richard: Yeah, okay. So I can give you a couple of examples there. I mean, one, again, another one where we had to build our own tool, because we do this enough, is analyzing reviews, right? So, I mean, and this is a simple idea, but it's actually quite a powerful one. I mean, a lot of companies at this point have a lot of reviews out there, B2B or B2C, and, you know, thousands, right? And so analyzing them takes time. And, you know, it used to be when we were starting a deal, we'd download the, you know, you have a web scraping team. We'd scrape whatever, several thousand reviews. We'd send them to a team in India. The team in India would read them, synthesize them, put them on slides and send them back. And that might take 24 or 48 hours, depending on how many reviews you want to give them. Now we have a tool which we call Classify. And we put them in Classify and it synthesizes them in, you know, I don't know, five minutes, right? It'll be able to go through and synthesize them. And it can pull out its own, you know, you don't have to give it themes. It'll pull out which themes are important. It'll say which companies did better or worse across those themes. You can even chat with it and say, you know, what should company X do to improve their company? Give me five examples and give me quotes. And it'll just, it'll do that for you. And so that I think is actually really powerful because you can get those insights then in, you know, in the first hour or two of the case, right? You can know what people are saying about the brand and that, you know, having that early insight matters because it then colors as then you go talk to customers to, you know, that now you know what that - you know, now you push harder on these things that came up. If you don't get that analysis done until the end of the first week or something, you know, you've missed an opportunity to corroborate it on other, from other places.
Shiv: Right. And that can drive a ton of efficiency and insights as well. I think that's a great example. What about on the go-to-market side? Have you seen or have you looked at tools or ways in which yourselves or companies internally can uncover insights from the large data sets that they have by leveraging GenAI a bit better? And a couple of examples of places where this data sits is inside Salesforce or inside HubSpot and we see a lot of the use cases where we're brought in is the data is kind of all over the place and we come in and we organize the data. We gather the insights, we're making recommendations on making changes to go to markets, do main changes to the marketing team or their overall campaign spend and things like that. Have you seen use cases there where companies can do a much better job regarding marketing or sales in particular with the large amounts of data that they have sitting inside their own systems?
Richard: Yeah, and it's a great opportunity. One where I feel like we've only scratched the surface of what's possible because as you say, so few companies even have the data in the right shape to do this sort of thing. The use case that I am personally most excited about is this idea of a bot that pings all your salespeople on, let's say, Monday morning and says, here are five people you haven't talked to in a while. Those might be leads that have gotten a little cold. They might be customers that are closer to renewal, whatever it is the machine chooses and says, here's five people you haven't talked to in a while, you should go talk to them. And then says, and I've written an email for you to send to them. It could be here's a white paper that we've, you know, that might be interesting to you, or here's a new feature we just launched or, or whatever we had, you know, last time we talked about blah, blah, blah. And, you know, again, the more context you can give it, the better, right? The ideal is last time we talked about blah, blah, blah, you'd be excited to know that we've done, taken your advice and done whatever, right? Whatever gets them to respond.
Ideally, the salesperson is going to read that email before they send it because the way GenAI is today, it's going to give you an 80% draft. It's probably not ready to go. But even if the salesperson just clicks send, whatever, it's still way better than where we are today where those five people are probably never going to hear from you. We're not going to hear from you for another month or so. And so, just pushing the salespeople to just be more proactive, but giving them the tools to do it so that this doesn't waste half their day, I think that's a big unlock.
So that's, that's, that one's exciting for me. well, I will say it's exciting for me as someone who likes seeing companies be successful. It's, it's not as exciting for me as somebody who doesn't like to get lots and lots of email from, from vendors, but you know, whatever. I'll take, I'll take the other side of it.
Shiv: Yeah, it's almost like there's - using sales as the example there, there is a bunch of work that you need to do that's outside the value creation piece of actually messaging prospects and nurturing deals and actually closing deals, that it's almost like the sales operations piece of it. And the more of that that you can automate the more time that can be spent in actually creating value and, and driving revenue. And that example that you give is a great one. It's like every week, yeah, there is somebody that you haven't talked to in a while or you need to follow up with, or you just haven't checked in on based on the last conversation. And sales reps may often miss things, but if that's captured through great sales operations and an AI tool, then you can kind of skip past some of the grunt work of that or the administrative work and actually get to the value creation piece.
Richard: Yeah, I mean, I think sales ops is a place where that's sort of ripe for opportunity for these technologies. I mean, I think that's an example. But, you know, I think one of the things we often see, and this isn't a GenAI point, this is just a general point is that, you know, companies aren't always pointing their salespeople at the most promising leads, right? That their way of scoring and prioritizing leads is not as scientific as it could be. And then if they used more data or, you know, they could they could be more successful. I think GenAI can be a big help there in, you know, both figuring out more about companies and what those companies are actually doing. So just doing better lead scoring because you're able to extract information. But also you can imagine a GenAI that like helps create the schedule for the salesperson to help them or helps them or creates the prioritization list for them or just in general helps point them at the right opportunities. Because obviously the salespeople want to make sales. They want to be pointed at the good leads. Right. I mean so everybody wins there. But if they don't know who the good leads are, they're just going to do what they can.
Shiv: Yeah, 100%. I think that's a really, really great point. And I think it's a good place to stop for us just to be respectful of time. But before we end, what would be a great place for people to learn more about you, Richard, and also Bain and what you guys are up to?
Richard: Well, that's a great - I'm trying to think of what the right answer to that is. I mean, obviously we're a big consulting firm. So, you know, you can always reach out to me over email at [email protected] and send me an email. I'm happy to chat about GenAI and PE. I am actually planning to start a Substack in the next couple of weeks. So if you Google me, you will probably find my sub stack and I'll probably have one or two posts on it. And maybe there'll be something there that you want to read or subscribe to. And you can always find me on LinkedIn too.
Shiv: That's great. We'll be sure to link all of that in the show notes and this episode will go live after your sub stack is live. So we'll link that as well if you send that to us over email. So looking forward to it. And with that said, Richard, thanks for coming on and sharing your wisdom. I think a lot of the PE firms and investors that listen to this podcast probably will learn a lot in applying some of these lessons to their respective companies and how they think about even investing and doing diligence. So appreciate you coming on and sharing.
Richard: Great. Well, thanks for having me. I think this is a great conversation and I really appreciate you inviting me to join your podcast.
Shiv: Thanks, Richard. Have a good one.
Richard: Thanks.
Suggested Episodes
Ep.37: Matthew J. Safaii of Arrowroot Capital
Improving Alignment in Software Investments
Matthew discusses the alignment-focused strategy that helps Arrowroot separate themselves from a crowded market.
Ep.38: David Acharya of Acharya Capital Partners
Growing Portcos Through M&A
David shares his proven process for using acquisitions to grow portcos and generate greater ROI for investors.
Episode 39: David Mangot of Mangoteque
How To Leverage DevOps to Deliver a Competitive Product
Learn how PE firms can navigate DevOps and technological processes in their portcos.