AI's impact on private equity

Alex: Hey guys, I’m Alex Agron and this is the private equity technology podcast.

Lisa: Hello, and welcome to the private equity technology podcast. And this is Lisa Weaver-Lambert, co- hosting with Alex Agran. And today, we’re going to discuss official intelligence. It’s no longer a futuristic notion. It’s here right now. Pioneer companies are already using AI to innovate and grow fast. Will Lee and Neil Callahan of pilot growth, are here today with us to share their insights on the implications of AI and the shift and the transformation of all business processes within organizations, and their personal experiences working with AI in portfolio companies. They are using the deep industry analytics technology expertise to help their own portfolio companies get real value from these technologies. Welcome, Will and Neil.

Neil: Hi, Lisa and Alex, thanks for hosting us today. We’re excited to speak with you about how AI is impacting our business as a private equity investor, as well as how all of our portfolio companies are either beginning to use AI or have been using AI for the last several years to enhance their offerings to help their customers be more successful.

Will Lee: Thank you, Lisa. And thank you, Alex. Thanks for having us here. I’m Will Lee, and I’ve been an investor for 18 years, and also I’ve been creating software for over 30 years.

Lisa: Fantastic. And can you explain the origins of Pilot Growth? How did that start, when did it start, and what  were your motivations?

Neil: Sure. This is Neil Callahan. Pilot Growth is really out of a desire to help other entrepreneurs achieve the same successes that we have had as entrepreneurs. Will and I are both successful bootstrap software entrepreneurs where we created our own software companies really out of our own capital, or minimal seed capital and grew those businesses, and sold them, and had successful outcomes. And we really wanted to create a platform for other entrepreneurs to not only receive capital to help them grow, but benefit from the network that we’ve built, as well as our own experiences in building companies and growing them.

So, really, we’ve built a platform for entrepreneurial software developers and innovators to grow their businesses. So, we founded Pilot Growth about 10 years ago. Just finished investing second fund in our own to raising our third fund. So, we’ve been very entrepreneurial ourselves, and have built the platform to support entrepreneurs.

Lisa: And what is the AI component of your business?

Neil: Well, the core AI component of Pilot Growth business is our deal sourcing engine. We built NavPod out of necessity. Our focus is on finding companies that don’t need our money, that aren’t raising money, that are very successful bootstrap software companies. And we’re finding companies that we’re trying to introduce ourselves to and convince them to partner with us on a capital raise, as well as to partner with us on a long-term to help them grow their business.

And those companies are very difficult to find, as well as find the right time to engage with them. So, out of necessity, we built NavPod to really track the universe of companies that we’re targeting and really to help identify the right time to call. And that’s where Will and the team sat down and said, what’s the best way for us to leverage technology and machine learning and artificial intelligence to help us find and track and identify the right time to call these companies.

Lisa: And what results have you had from this investment on NavPod?

Will: So, we started building NavPod in about 2013. And now, it’s 2019, so this is about our fourth year of using NavPod. And so in total, she has about seven years of Pilot Growth investment data. And also, Neil mentioned that NavPod has become a better investment over time as she gets more data over the years. And the way she learns to find better investments is similar to how, say the Facebook or Apple news learned about you, like reading preferences and using machine learning. And so, NavPod has our own artificial neural network that we built in-house. And in terms of the sources of NavPod, I’m sure you’re all curious, like what does she look for?

So, just like a human analyst, the sources are the public filings, and news, websites, database screens, such as many publicly available databases, as well as CB insights, which is one of our portfolio company. And we also use IBM Watson to do some of the processing work. And so, I want to emphasize that NavPod is a deep learning AI system, so she learns from us the principles. There is no spreadsheet, there is no large copy and pasting or importing from other deal databases. Each detail is handpicked and filtered by NavPod just like a human would do, but it will do it in a very large scale.

There’s no relational database, no rules engine, and also there is no… If you’re into coding, there’s no ‘if then, else” conditional statements in the code to filter out deals. She really uses AI deep learning to learn to be a better investor over time just like how a junior person would do. But as you know, deal sourcing is, it’s always going to require an element of human judgment, but about 90% of the manual work, I would say manual grunt work that goes into deal sourcing we have been able to automate with NavPod using deep learning.

Lisa: And Will, can you help us define AI? Because we hear a lot about deep learning, machine learning, data science. How do all these terms relate to each other, and what is the significance of AI?

Will: Sure. Yeah. Well, as you know, AI techniques and AI algorithms has been around for many decades. You know, some say even over a thousand years ago by ancient philosophers. But really, the modern AI can go back to Alan Turing’s in Turing’s test in the 1950’s to assess whether a machine is intelligent enough to be human, since then, the AI field has gone through the ups and downs as a promising technology that could really change the world. But in the most simplistic form, AI is a subset of a broader data science. And data science is much broader set of technologies. And also, it has evolved as an industry.

It is until recently, in the past decade that AI is really becoming more popular again. And the main driver for this is the exponential gains in computer processing power, and also the storage capabilities that gives modern companies the ability to store huge amounts of data. And just as AI is a subset of data science, machine learning… You hear this term all the time, machine learning is a subset of AI. And then, machine learning is really that dynamic way for machines to modify itself without human intervention.

So, at Pilot Growth, we use something called deep learning. And deep learning is a subset of machine learning. I mean, it just means that the machine has multiple layers inside of the artificial neural network, similar to like a brain which requires more computation intensive processing power than the regular machine learning. And the benefit of deep learning is the high accuracy for the machine to find patterns. A the perfect example of this is Google’s deep mind, Alpha Go, which beat the world champion in Go, in 2016.

Lisa: Thank you, Will. That’s really helpful. And I’d be very interested to hear about your partnership with Neil. Because I understand that Neil’s background is different, and how you started working together, and the challenges that you had to overcome in building NavPod?

Will: Sure. Yeah. As Neil started out by describing what we have to do when we built NavPod. The challenges of building a growth equity business is to find great deals similar to the kind of companies that I co-founded. And there are thousands of these companies in the U S, and also, as well as around the world. And how do we find them? And these companies grow, and some of them…Actually, a lot of them become stagnant, and some of them die. And how do you know when to call? And even for a traditional human analyst actively keeping track of, I would say, 20, 30 companies is a full time job.

So, you imagine trying to keep track of 3000 companies actively, and then produce a report to your investment committee every morning about which companies to call, which companies that are interesting, and that will require an army of analysts. So, as you can imagine, we don’t have an army of analysts. We have NavPod doing all the work of maybe 20 to 30 analysts producing a written report for us every morning.

And then, when we talk to entrepreneurs, and they realize… We once, Neil and I, and our other partners, also former bootstrap entrepreneurs, and they are more receptive on hearing about how we can help them to grow their company with our business development team, which can help them open doors to customers and partners. So, this has been a very effective way for us to reach out and build relationship with entrepreneurs.

Lisa: And Neil, what are the things that you needed to change in the way that you were working when you started working with Will. And how did you compliment what Will was trying to build?

Neil: Well, yeah. I think the way in which we had to change working was that the traditional style of growth equity is to develop a thesis of what characteristics of companies you’re looking for, and then how do you go about and finding those companies, right? And so, it’s kind of a very simple process of, what’s the profile of the companies and how are you trying to find them?

What we did was develop a platform in a technology to really remove much of the manual labor and the guesswork in terms of finding those companies who are traditionally a private equity firm focusing on growth equity would need to go and look for companies that meet that profile. And they would do that through multiple different sources. And then the way in which they screen those companies is a very manual process.

So, what we did is developed our thesis, which was focusing on growth equity technology companies in the B2B space. I had very specific sectors that we’re interested in looking at, and we’re able to track all private US technology software companies, and use criteria to filter them down to the ones that we want to track. And what NavPod does from there is to understand really what’s happening in those companies by tracking multiple data points and flagging them to say, there’s something happening here. And really, the focus of our business is to engage with a company when they’re at a point of inflection of growth, right?

And so, if you’re engaging with them after that point of inflection, you’re too late. And if you’re engaging with them prior to that point of inflection, or they may never get to that point of reflection of growth, you’re really wasting time and resources. So, what we did to change was to say, look, let’s trust the system to essentially flag the companies we should be spending our time with, who we should be calling, who we should be gathering more information with, versus tracking down a whole list of companies in spending time with them until you find out that the answer is it’s not a good fit, right?

What NavPod is doing is saying, Hey, these companies are… Don’t spend time on them until we think it’s a good fit, until we have the data telling us it’s a good fit. And what that does is remove very common biases that we all have in our decision making. Things around confirmation bias, where if someone recommends a deal, or we find a deal that is relative to a deal that we had a good experience with, we tend to spend more time on that regardless of the underlying data, or recency bias, right?

The things we heard of just the other day, or things we’re focusing on now, versus a company we might’ve met two years ago, now they’re at that point in reflection. We really should be engaging with them. So, what NavPod does is try to remove those biases of how we assess an opportunity, and how we spend our time. And that’s a big difference than your classic private equity, which is really about who you know, it’s really about which network you’re connected to, and really, who’s making the most noise to get your attention.

Lisa: Thank you, Neil. I’d love to hear about how you’re working with your portfolio companies in the Valley creation stages, and how you’re bringing in AI practices and learnings, your learnings from creating NavPod into these businesses?

Neil: Yeah, that’s a great question. I mean, we really… The challenge of what we’re doing in private equity is not to find a partner with good companies, it’s to find a partner with great companies, right? And so, part of attracting great companies that again don’t need our capital, have lots of access to resources, and are deciding whether or not they should take growth equity because they have a successful business that’s break even a profitable, that’s growing at greater than 100% and a great team, is we come to them and say, look, we’ve been there, we’ve been in your shoes and we know that one of the most difficult challenges you have in terms of growth is how do you get more deals and how do you increase your win rate, right? and how do you do that in a very capital efficient way, and how do you do that quickly, right? Because the market is very competitive. There are folks who are now keen to your innovations, who might have more money or a bigger reputation in the marketplace.

So, we come to them with our business development advisors. We’ve spent time really curating and building a great team led by our partner, Rob Walker, of folks who have a tremendous reach into industry, and also have tremendous reputations in terms of having relationships with C level executives or general managers, or senior people who can be sponsors. And what we do is help our portfolio companies get introductions at the sponsorship level to introduce what they’re doing, and that sponsor that helps them find and navigate the right buying center or the right director of technology, or a director of AI who’s trying to help grow that business.

And for example, we helped one of our portfolio companies get introductions to the US Navy, right? which is a great prospect for them. We’ve helped introduce folks to places like Walmart, where people are trying to introduce what they’re doing, get the attention of a sponsor, and then then move into maybe a pilot stage or in a test phase, then into full production. So, that’s something that we actively do from a business development and value ad perspective.

From an AI perspective, we’re really sharing the knowledge of what we’ve discovered and the benefits we’ve gotten from from AI, and helping our portfolio companies integrate that more and more into their core software.  And we’ve done that by helping them understand… In many ways, it’s about workflow. It’s about once you find a pattern or once you identify that the AI platform or the machine learning platform is finding targets more often, more accurately and cheaper than the human would do, how do you then work that into the actual workflow of your software and the workflow of your business?

And for instance, with Velocidi, what Velocidi does is it helps e- commerce companies improve their return on marketing investment by creating better segmentation, identifying customers who are more likely to buy, more likely to not have to have shipping paid for, more likely to not return an item, so all sorts of use cases around buying behavior. And we’ve helped them think about classification, right? How can you, instead of identifying use cases on a one off basis, or identifying use cases as your customers are talking to you about them. How about we build an ontology and a classification of the most common e-commerce use cases?

And that’s an initiative that Velocidi is taking on to build a really a robust library of [inaudible] commerce use cases that are sort of out of the box, if you will, that any e-commerce customer could then take advantage of and say, look, we can identify these common traits in consumer behavior and then focus our advertising spend on those segments, and then remove the advertising spend on the segments that aren’t showing that behavior.

So, we’re really coaching these entrepreneurs who have incredibly disruptive and innovative technology, and bringing more of the AI best practices and learnings that we’ve had as practitioners of AI, and also students of it as well by seeing many, many companies on a daily basis using AI and deploying it to provide value in their software.

Lisa: And have you found that you need to change the skillset within the portfolio company to implement AI solutions? And if so, what type of skill sets do you look for?

Neil: I think the most important skillset in AI today is domain expertise. I think that data scientist is sort of the new computer scientists, right? We grew up in the IT boom of the 80s and 90s and 2000s, and it was really around computer science. And now, we’re in the age of data science. As Will said, because of processing power, because of cheapness of storage, because of ubiquity of data, making sense of that data, and using it to your advantage is what companies are trying to do. But the way to do that is to have the domain expertise. Understand what the best outcome is and what the best process is and then how do you automate that exponentially, right?

That’s really what AI is trying to do, is taking a salute problems and addressing them with solutions in a way that humans can do, but doing it at super scale. For instance, in the financial services industry, if I look at an application for a mortgage, if an expert looks at that, they can identify, Hey, this person is a likely fraud, so let me flag that. So, any human expert can do that and be extremely accurate. But how do you have a machine learning platform be able to do that without having to employ that expert and can move at a speed that that expert could never move, right?

So, AI is really automating tasks that humans are experts at, but doing it at scale and providing massive productivity for an organization. So, the most important aspect to our portfolio companies is that the team members really understand the problem they’re trying to solve. We see many companies in the AI space where they’re showing up and saying, Hey, we’ve got the capability to solve your problem. Tell us about your problem. Right? Versus, we’ve built ontologies and classifications around a specific use case, around auto stock management, around advertising demand, around understanding product to source assortment.

So, very specific use cases that by vertical or by industry, bringing that knowledge to the table and having that already classified and baked in, if you will, to the platform. So, that we believe is the advantage. The idea of trying to find data scientists who then can model something, or try to find value in patterns within data, frankly, that’s a commodity. And there are lots and lots of folks who will do that and can do that. The key is marrying that with the domain expertise.

And that’s what we did here at Pilot Growth, right? Where, as private equity and growth equity investors, Will, I and Rob, and our extended team really understand the software business and the B2B software business at its growth stage, because we’ve spent our entire careers in that. And then, having the ability to have the in-house data science and computer science expertise resident. And Will allowed us to really move forward very, very quickly, because we had that match where other folks within the private equity space have tremendous domain expertise with lots and lots of great competitors who were extremely successful.

But it’s very rare to then have one of your general partners be a world class data scientist as well, right? So, that’s the advantage we have. And again, that’s the approach in which we look for in companies, and we help coach or companies around marrying domain expertise in data and in computational science.

Lisa: Will and I were talking about one of your investments when we were chatting previously, r4. I wondered if you could talk a little bit more about that and the learnings that you’ve with that investment?

Neil: Yes. So, r4 is a very unique company that really is helping large legacy companies that have built over the years have invested hundreds of millions of dollars in transactional systems, right? So, if you look at a large enterprise across the entire world, there are financial systems, there are ERP systems, there are human resource systems, there are marketing systems that were invented to complete a transaction. And those transactions begin and end, and begin and end, and begin and end, but are not necessarily connected throughout the enterprise.

And what’s happened is there is trapped value between those transaction systems. We call it hidden demand. Where if you could look across all of those systems and understand what’s happening, you could see there are opportunities to serve your customer better, or present and offer better, or have a new product, or to reduce inefficient spend.

And so, r4 is that our approach, their approach is to say, we can look across the entire enterprise and look at these individual transaction systems that are connected to each other, but sort of bump up against each other and find hidden demand, and generate what r4 calls is rivers of cash, right? The ability to identify where there’s an opportunity and quickly recommend actions to take.

And the actions to take are, which inventory should you ship to which store, how much spend should you allocate towards a specific segment, which products should we start to highlight in our advertising, which employees should we be starting to support more and promote more because of their ability to contribute to the organization.

So, that’s really what r4 has done. And they’ve done that using machine learning and artificial intelligence. And the key there is that they’re not looking at data that’s been organized, data that’s been put in a data mart, data that has been cleaned up. What they’ve been able to innovate is their ontologies, their classification of business problems can understand data that’s not structured, data that is not cleaned, or dirty data, if you will, and be able to infer meaning from that.

And that’s the key of what r4 has been able to do, is look at data across the entire organization that’s in different formats, that is inconsistent. And using their machine learning algorithms to say, we can identify a pattern here, we can identify an opportunity here and then recommend something to go ahead and do.

And for instance, and one of their customers, they were focusing on which customers to engage with in their mobile advertising, right? Try to reach out to customers while they’re shopping, either in store, or out and about, or online, but they’re on their mobile phone and saying, which one should we be advertising for and to try to get them into an engagement, into an interaction cycle. And r4 was able to look at all the data across this organization and then make a recommendation of which targets to focus on.

And when they implemented that, the engagement of the folks clicking through and using the mobile app and the mobile experience went up by 7X versus control. So, that’s an example where they had the classifications of how to engage consumers on a particular product in a particular use case. Looked across that customer’s data came up with a focus area, and then we were able to identify actions and targets to go and actually put into production.

So, that’s how r4 is really approaching the market. And in a very unique way, it’s not a point solution where it can solve one problem and solve it very, very well. But it can look across the entire organization and solve many, many problems over and over again. And the value to an enterprise, like a large CPG company or a large retailer, or a large financial services companies saying, this is a great match.

So, instead of me having to go out and hire 5, 10, or 15, or 20 different individual AI companies that do one thing and do one thing really well, I can have an enterprise platform that I can solve many use cases using machine learning over and over again.  One set of data, I can use one platform versus having one set of data but using many, many platforms for different challenges.

Lisa: Awesome…

Will: And what’s…

Lisa: Sorry. You go ahead, Will. Go ahead.

Will: Yeah. Yeah. Sure. Yeah, I just want to add to that. And also, what’s really unique about r4 and this goes back to, Lisa, your question about what can portfolio companies do to implement successful AI product and technologies. So, what’s unique about r4 is that they have done this before. And these are former founders and executives of Priceline and IBM. And so, the way they changed the whole travel industry and turn Priceline into $100 billion company, they were using that same set of data science DNA from the travel industry that they disrupted over a decade ago, and we’re using that across different vertical industries.

And this is why r4 is able to really essentially create these rivers of cash, as Neil was saying for companies in different verticals, and really find these hidden revenue opportunities, versus a more traditional enterprise software they usually solve for better optimizations for reducing operating expenses. But r4, they are able to find hidden revenue opportunities, which is a much more difficult problem to do. And they have done that before successfully.

Lisa: What are the conditions for success if you are a portfolio company or a [inaudible] owner? What value is AI to companies that are not already on the data path?

Will: Yeah. So, the condition of success is really defined by the application of AI itself. So, there is the image recognition AI, cybersecurity AI, precision medicine AI, and self-driving car AI. They all have different definitions for a successful implementation of AI. And our case for Pilot Growth, the condition for success is for growth equity investment AI.

And one thing that is common is that algorithms that AI runs on, they all need data. And so, for growth equity investment, we need company data, we need market trends data, we need investment knowledge as data. And the quality of data really lays the foundation of a successful implementation of AI.

Lisa: So, Will, we’ve heard quite a bit about NavPod. Can you help us understand also what NavPod doesn’t do? Also, Neil, what are the sort of criteria for acceptance of the output from an AI system?

Will: Sure, Lisa. So, when one thinks about when is AI really applicable, we can really think about what AI doesn’t do for us. And so for NavPod, she actually doesn’t do many things, especially related to the soft human aspect of the task. So, NavPod does not build a relationship with entrepreneurs, because she doesn’t go out and have breakfast with them, or a lunch or dinner. And building relationship with entrepreneurs is very important and critical in getting a deal done.

And also, she does not convince the entrepreneurs, the value of us Pilot Growth as entrepreneurs, as well as the value of our business development team. And NavPod does not analyze financial statements or technology due diligence, or NavPod does not do customer references for us. We, the human, the managing directors, we still have to do all that work. It’s just that we can do it more efficiently, because the deals that we’re working on are highly qualified.

So, AI is really applicable when your company is dealing with a large set of data and you want to make use of that data to give yourself an advantage. And you want that data analytics process to be done in an efficient way. In the case of Pilot Growth, we leverage our AI or our NavPod deal sourcing and workflow engine to sift through tens of thousands of companies, tracking them like a team of human analysts would do, informing the managing director, which companies we should spend time on doing outreach. And we can do this without having a team of junior analysts, because we have NavPod. But she really does not do a lot of the relationship building, and what human we’ll do really well, and also the creative aspect of deal sourcing.

Neil: And Lisa, the idea of criteria for acceptance for AI platforms is a really important one. In your classic information technology systems, in many ways, the benefits of automation were pretty obvious, right? So, if I’m going to replace an invoicing system where my current invoicing system is, I create invoices and I mail them to customers, and then I expect them to write a check and send it back to me. Versus a platform where I can create an invoice on my system that connects directly through email or through electronic means shows up instantaneously at my customer and they can pay electronically with a payment system or a bank where I can receive payment almost instantaneously. Right?

So, the benefits of those systems are fairly obvious, because they’re transactional systems. And you’re trying to say, well, here’s the end to end transaction I have today, and the cost and time and error rate. And on a system, I can see that shrinking, and I can make a decision to go ahead or not go ahead. In AI, it’s a decision making platform. So, in every instance, you need to identify, well, here’s our current decision making process and the outcomes we’re getting, and here’s what the system is recommending, a different in terms of decision making and we need to see that play out.

And the key is to when do you make decision to go with it? Because one of the key concepts of AI is, it gets better over time, right? So at a certain point, you birth your AI platform and it starts to grow and feed off of your data. And as you have more interactions and more data, the actual system becomes more accurate and more dependable.

And so, the question is, how soon do you want to cut over to that system, where if it’s a small amount of data and sort of some inklings that it’s working, do you want to cut your entire organization over to that or do you want to let it grow and learn for another year before it gets fully competent and has the same level of expertise that a human decision maker would? And we’ve seen that in, say medical imaging, right? where the system alone has a certain level of accuracy, the physician or tech alone has a certain level of accuracy. But combined, they have a better level of accuracy.

And so, the key for criteria of success is a big one. Versus saying, Hey, let’s get AI here to make better decisions. Or you need to go decision by decision and say, well, here’s our current efficacy of decision making    and costs of decision making. And we want a platform that as at least as good as the human platform, or a card platform, hopefully better. But the challenge is we want to be able to do that at scale. So, if you’re not looking at that from a test and control perspective of what are we currently doing and how does AI help us improve that across many, many metrics, then you’re really not approaching it correctly.

Again, other systems, it’s very obvious. We don’t have to have that level of sort of oversight, because the benefits are clear and present, where companies need to understand, as am investing in AI, what am I really expecting to get out of it? Versus a platform that’s going to start making decisions and they’re really not beneficial, they’re really not better off than we were before. So, that’s a key point.

And for Pilot Growth, really the key measure of success for us has been really understanding all the companies that we’ve been looking at, and NavPod has been looking at over the past six years. And when we look at… We’ve had one company where we’ve identified it through NavPod and brought it all the way through to exit, which was Zenedge, which NavPod identified, Will cold called them within the investment, helped grow the company, and had a successful sale to Oracle.

But in that same time period, there were 100 other companies that NavPod identified as hot and warm that had been acquired by strategic technology companies, right? So, that’s about a 25% hit rate of our total of our hot and warm leads. About 25% of them have been acquired. So, we ourselves hold ourselves accountable to say, we believe it’s working, it’s helping us with our operational processes, but it’s also identifying high value companies, and the data shows that. So, that’s a key aspect to adoption of AI; is it really providing the value that we’ve set out for it.

Lisa: Thank you so much, Will and Neil. There’s so much to talk about on this topic. Where can our listeners reach out to you? What’s the best way to get in contact with you?

Neil: Well, I would say we’d open to anyone reaching out to us. You can contact us by email. My email is [email protected]. Will’s email is [email protected]. We’d be happy to engage with users, or technology companies, or other private equity firms, or what have you. We were very collaborative, and very curious to learn more from others. So, we welcome any feedback or engagement.

Alex: Thanks for listening to the private equity technology podcast. Please support the production of this podcast by subscribing in iTunes and leaving a review. If you want to reach out with any questions or comments, you can get me at [email protected].