Podcast transcript: What AI has done for Hg

22 min approx | 6 January 2022

Winna Brown     

Today I chatted with Nic Humphries. Nic is a Senior Partner and Executive Chairman of Hg and Head of Saturn Fund.  Nic is responsible for Hg’s strategy, management and governance. He focuses on Hg’s larger software investments that provide daily use, mission critical applications for professional users such as accountants, lawyers, tax compliance professionals, designers, engineers and scientists. We talked about how being an early adopter of AI has benefited Hg; how AI has transformed both their portfolio operations and their origination process, and the ways in which Nic expects AI to drive the investment landscape in the future. Nic, thanks somuch for joining me today.

Nic Humphries   

It's an absolute pleasure, Winna. Thank you for inviting me.

Brown   

So Nic, while Hg has been successfully investing in software companies since the early 2000s and exclusively focused in the sector for over 10 years now, do you recall the moment that you made that bold decision to invest in AI and advanced analytics and really become an early adopter of the technology at a time when AI was not that widely known.

Humphries          

Yeah, I would say it wasn't a blinding flash of light at one particular second in time, but certainly around about six or seven years ago we could see some of our portfolio companies that were advanced in kind of cloud technology, starting to experiment with kind of AI and machine learning and analytics, and we felt that it had to be something that we at Hg developed ourselves as an area of expertise because it would become applicable both to our own business and also to pretty much every business that we invested in due course. And so, it was around about that time that we started to get people on board to Hg that had the requisite backgrounds.

Obviously can a PHDs in kind of machine learning and AI. And we started to build a team and that's mushroomed today to be a team of nine, ten people, and also a very close relationship with a third-party organization where they’re major client and they do probably can have half or two-thirds of their work directly with us as well. So collectively you think about it as a team of kind of somewhere between 30 and 40 people that we can get a pull in this area.

Brown  

Do you think that your focus on AI has really changed your market position today?

Humphries          

I think it probably wouldn't be distinct and recognizable as like the number one, thing that distinguishes Hg. There are probably other factors that perhaps we develop slightly further ago and are more mature and tend to get recognized in the marketplace. So, I guess my opinion is things that you're developing and experimenting with are somewhat new, all those tend to become a recognized by the market more generally, a little later and that's fine from our point of view; it means we get to experiment. We get to going kind of trial, and we get to develop expertise may be a long while before the market recognizes it, and hopefully before our competitors recognized it as well. So, I think there'll be other things that right now the marketplace would say a distinctive about Hg. Our scale in Europe. Just thinking of volume of deals we've got in particular and markets or in sectors that we've been developing for 15 to 20 years, but certainly, from our point of view we think AI and machine learning and the things we're doing there are going to be formative for the way we think about investing, the way we think about the processes and what we do within our own business, and also, obviously the way that RN companies choose to invest. So, we think it's a very kind of lead indicator of what will differentiators, hopefully now, but very much in the future.

Brown  

So, when you think about the way you approached AI, is this how you approach all of the companies that you assess with this philosophy of curiosity and experimentation and you know, kind of giving it a go to see what could be that next big thing.

Humphries          

Yeah. I think it's exactly that. It's a kind of cultural kind of approach that says, when anything is kind of relatively new and relatively nascent, I think it's impossible to predict, you know, the one or two ways or even three ways that are going to be ultimately successful. Maybe other people can do that. I've never thought we could and so experimentation and trying things and being prepared to acknowledge that some things will fail, and you'll inverted becomes waste money, but ultimately, you'll learn, and you have a kind of culture of continuous improvement. So, you try, you learn. You improve, you drop those things that don't quite work. You push harder on those things that do work. I think that's all part of something that ultimately will then allow you to identify what things are working, what things you can scale, what things will be applicable across a multitude of kind of different areas. So yeah, that attitude of mine is certainly what we apply to cloud and kind of SAS ten plus, years ago, and I think it's been helpful to have that attitude of mind. Both for us as Hg of how we think about our business and also about how our end investing companies think. So, you know, we think it's not right for example, to run our companies, absolutely kind of optimal margins in the short term that may be applicable but in the medium to long term, if you want to continue to generate superior organic growth in your businesses, you have to be prepared to invest something today that doesn't optimize today, but it allows you to achieve superior growth performance five years, seven years, ten years from now.

Brown  

What are some examples of, maybe some non-intuitive or non-traditional data points, which because you have access to data, and you've been doing this for so long really helps distinguish you and your process around origination?

Humphries          

Tapping into social media. So, kind of LinkedIn, but also Facebook, Instagram, you know, a whole bunch of other kind of social media that isn't applicable for every type of business because it obviously depends on the end product. But basically, we found that there are links between degrees of social media activity and usually kind of success in marketing obviously. And that leads to usually if there's some efficiency success in sales and so there are kind of correlations between that degree of activity and ultimately the rate of growth and the rate of success of a company's revenue line. So, some of that sounds obvious when you say it, but it wouldn't necessarily be in the kind of things that we, as a B2B software investor would have been tracking kind of five or seven or ten years ago. We've also found, you know, correlations between a company's kind of activity in terms of product launches, in terms of kind of marketing announcements, you know, terms of attending, kind of conferences. So, there's again a number of things that sound reasonably obvious when you say them, but when you pick any one of those data sources alone, they're not necessarily indicative. It's a combination and understanding kind of the relative weight of those sources that leads us to some insights that perhaps we wouldn't have been able to gain in the kind of old non-data, non-AI world.

Brown  

There's a lot of concern out in the public today around AI. You know, they're saying it's going to take away White Collar jobs and other people are seeing AI as a significant opportunity and it's going to open up career paths that some career paths that we don't even know about today. So, what I was curious about is what are some of the applications that you're seeing across your portfolio companies of AI and which ones have been really transformational? And have you seen these impact the advances? Or have you seen these impact the workforce in those portfolio companies? I know you said they're engaged and they're excited and they're looking at the next big thing. But is it changing the jobs that exist and the opportunities as well?

Humphries          

I think they can pick a number of examples. And at one end of the spectrum would be the applicability of kind of AI and machine learning to let's call it support tickets, you know, when a customer has a question and they historically would have dialed into the going to customer support center. Today, they might be kind of doing that online. They might be using chat, etc., and how AI and machine learning has improved that part of both the company that we back their service to their customer, but also started to change the jobs of those customer service representatives. And then at the other end of the spectrum, we can talk about the way in which AI has actually helped to develop brand new products, and those brand-new products are actually creating both new revenue streams and new channels to market that didn't exist before for our portfolio companies. So, on the first of those put, very simply, we have a number of companies that have applied AI and machine learning to, as I say these incoming customer support requests and that can actually be in a variety of kind of in band forms from voice to email to chat. To be honest with you. I can't remember the exact step but it's something crazy like 80 plus percent of all support requests are about how do I remember my password? And how do I turn on the application properly? It's a stunning number. And I suspect that. I may be wrong here, but I suspect that if you're a customer support representative and 80% of your responses have to be to those two questions, that is not a particularly stimulating job to have. Whereas they're exactly the kind of thing that AI and machine learning can take away and remove those issues from you. That allows you to focus your time on the 20% of support requests that are much more complex. Probably have a much more meaningful impact on customer satisfaction and the customers you see your product and its utility and so in those businesses where we've had AI applied and they've cut down all those kind of very mundane AI support requests. What's happened is that we have introduced the number of customer support staff. We've actually kept or increase the number of customers support staff, but the job that they're performing is now we believe much more sophisticated. It's providing a better more stimulating job for those individuals that are doing it and it's also providing a better support for the customer in terms of utility of other software in the product that they're using.

Brown  

So what you're actually seeing is job retention versus attrition.

Humphries          

Yeah. So, across our entire portfolio we're trying to apply AI and machine learning wherever we can because we think it's applicable. We haven't decreased job count in any of our portfolio companies. We're actually increasing employment. I think last year by about 17% through the middle of COVID-19. So that's kind of thousands of knowledge economy jobs being added in Western Europe and North America. And we believe those jobs are actually stimulating, they’re well-paid kind of creative jobs. And frankly, we think that AI is generally removing the mundane, the kind of thing that those people wouldn't want to spend their time doing.

Brown  

I'm sure that's being very well received.

Humphries          

We hope so. Yeah, I think at the other end of the spectrum, we've got a number of examples from our portfolio of where AI, machine learning, big data is being applied to actually help improve the revenue profile in certain cases. And then actually to develop new products. Sometimes products that just could been satisfied without those technologies. They couldn't have been done manually. So, we have a portfolio company called Visma. A couple of examples from Visma. One is they provide a lot of home care, home healthcare, software and support for nurses within the Nordic region to provide home help to patients. In the past those nurses were scheduled to move between locations on a kind of relatively rudimentary basis. By applying AI and machine learning. We're able to actually increase the scheduling capacity, i.e., either provide kind of more visits per nurse or provide a longer period of duration of care per visit. Basically, we can be more optimized and that is to the tune of 25% or 30% increases in kind of productivity for nurses of the Nordic region. So, a very tangible example, where AI and machine learning Some that couldn't be done before. Literally it would take days to provide some of these kinds of optimizations, by which case, of course, the day is over. So, it's too late. It can now be done in kind of seconds or minutes, which means it can be applied every day and at repeated times during the day to reschedule and that's providing better patient outcomes or more frequent patient visits, has to be better for the whole community. Similar application in school timetabling. So, school timetabling that either couldn't be done, optimally or literally take days or weeks to do this. Visma is one of the major providers of going to school systems for the entire Nordic region. And again, it's applying machine learning and AI to school timetabling to enable kind of optimal kind of resource allocation.

Brown  

That's just some great examples your portfolio companies in different sectors. And I'm curious though how do you envisage you using and leveraging AI and all yourself and other private equity firms to drive that investment landscape in the next, you know, call it three to five years? How's it going to help you predict what are those next areas where AI can create value or unlock value?

Humphries          

So I say, I mean, the start point is back to the experimentation point which is I can't tell you exactly which areas because we're experimenting in many. And it's early days. I think we've given some examples where we can see applicability. But whether they are pretty interesting niches for one or two companies or whether they're indicative of things that can be rolled out to all 40 or 50 of our investing companies, it's a little too early to tell to be honest with you. We think some of them such as, you know, in a subscriber-based business that has thousands or tens of thousands of customers and ability to kind of look at lots of different data points and be able to predict potential customer churn and to spot that in advance and then to be able to apply algorithms and other things to be able to go to direct either a customer service team or an account management team or a new sales team to going to go and talk to that customer early and to prevent churn are you can have increased retention, I think that will probably be applicable to most businesses within our portfolio for example. So that's kind of churn propensity modeling. I think upsell/cross-sell modeling and those kind of things will also probably be applicable to most business in our portfolio because most of our businesses have exactly the same type of business model and revenue model. So, we see those things being applicable but then they'll be applications such as the one of the Visma school timetabling example, that will probably only be applicable in one or two of our portfolio companies. So, I think part of this is about experimentation and some of those things will scale. Some of those things will be quite niche. I think the things we described earlier around origination and how we're using kind of data ingestion and AI to help us identify interesting companies for us to potentially invest in - that will probably be applicable and scale and be more meaningful for us and probably you can have several private equity companies over the next decade or so, as we improve and learn and develop our own expertise around that. And I think the third area where we've applied AI and data pretty extensively over the last kind of five or six years is to really understand how we can ingest company data so everything about a company's customers, the customer database, the upsell, or cross-sell, the churn, those kind of things within our kind of company's portfolio, to be able to go and analyze that in a very similar way between Company A, Company B and Company C. And to draw conclusions from that has certainly enabled us to be better at predicting what our companies will do and where we can help our companies to grow and grow revenues more rapidly. And I think the more data we have the better we get at that. That's the whole point of big data and I can see applicability of that in many areas. So, we've developed a whole number of kind of tools and methodologies that enable us to now do what would have taken is literally days or weeks four or five years ago. We can now do that data ingestion data management and start to do the analysis in literally a couple of hours. So, there's been a very dramatic improvement in our productivity in those areas. And as we learn more about tools, we learn more about the processes we use, and we apply that across, you know, more businesses more systemically. I think you'll continue to see very big improvements in how we apply that certainly.

Brown  

So you know, as you said over the last few years, you've taken your tools and your access to rich data and you've been able to transform that to now be able to reach, do analytics and hours versus weeks. So, my question is looking into your crystal ball, in 20 years’ time do you think they'll be a point in time where perhaps there's even a private equity fund that's completely run by AI and you wipe out those few hours, it'll all be automatic, and you won't need people.

Humphries          

I'm an engineer. So, like I guess I've always probably had slightly more IQ may be slightly less kind of EQ. And at one point, I kind of thought that it would be great if everything could be run by machines. I'm, you know, I guess as I get slightly older, I don't really, that's necessarily true. I think data and analysis and really understanding kind of how businesses work is clearly vital to the investment process. It's vital to going to thinking about how you forecast returns and from our point of view, you know, it's also vital to delivering consistency. So, you know, over a 20-year period, you know, we're proud from a client point of view that we've delivered, you know, consistent kind of mid 30s plus percent IRR returns and 3X plus times our money for our clients. But actually, to be honest with you, we're probably most proud of the fact that we've done that with huge degrees of consistency and with very, very low levels of kind of volatility. So less than 0.7 percent in bed capital over a 20-year period. We think those kind of statistics is a very important and they've helped to be improved over time by analytics and data but and the big but here is ultimately businesses are really run by people and those people are just incredibly important. We found that much as we might try and analyze the traits of successful people, guess what, the human mind and the human emotion and those kind of things are not yet capable of being analyzed down into going to ones and zeros. And so, I think business and particularly going to private businesses, particularly businesses in fast-moving sectors like in a software and AI line analytics are always going to be a mix of analytics and data and then people and organizations and motivation and emotions. And so, I don't think the world of going to Hal and those kind of things is ultimately going to apply to private equity, into investment. I think there's always going to be judgment and people skills that are very important and understanding motivations and drivers of the management teams and trying to get a make sure you align with those and partner with them. I think is and continues to be and probably will be well into the future a very vitalkind of skill.

Brown

Last question I've got for you. I mean, clearly, Hg’s philosophy around curiosity and experimentation as really paid off. And we talked about that and you gave lots of great examples. I'm curious have you got any tips or insights as to what's going to be the next great, big transformation on the horizon?

Humphries          

Well, I think it is going to be AI and machine learning. I mean, big tectonic changes in kind of tech, kind of people think they happen every two or three years, but in reality, they happen about every can 20-odd years, you know, Mainframe to kind of personal computing. Personal computing to Cloud. And, you know, Cloud, you know, for those of us are old enough to remember, kind of like, was starting to be talked about in the mid- nineties to late 90s called it different things then, like ASP. But, you know, it's only hit the mainstream and become very kind of major driver in the last six, seven, eight years for the mainstream and it'll continue to be the driving force. The next can have five to ten years and in the same way that, you know, we started to talk about and think about AI and machine learning, you know, going to five, six, seven years ago, you know, that's the equivalent of like 95, 96, 97 from a Cloud point of view. I think, you know, the real impact of AI and machine learning will probably be felt 5, 10, 15 years from now. And so, I think that is really the next wave that obviously sits on top of Cloud, you know, one enables the other. So, I think that is the next big wave as to what happens thereafter. Pure speculation because it's very early stage. But my suspicion is it'll be Quantum Computing. Could be Blockchain, but I suspect it might be kind of Quantum Computing that is the next and a very big leap and that's probably in reality is it going to real commercial applications at scale again, is probably going to 20 plus years away. But of course, you'll hear a lot about it, and I'll be very successful in niches and it will be very exciting to watch some early-stage companies in the next kind of five to ten years as well as well.

Brown  

Nic, thank you so much. Very much looking forward to seeing how AI continues to evolve and change all our lives and very much looking forward to hearing about some of those innovations in Quantum Computing. I'm going to start looking at my Investment Portfolio, I think.

Humphries          

It's been a real pleasure. Thank you so much for giving me the time and thanks to your listeners as well.

Brown  

Thank you.