AI series: discussing the opportunities and pitfalls of AI with Bridgepoint’s Head of Digital and AI

Paul Jevons:

Hello, and welcome to this Analysys Mason podcast series dedicated to the topic of artificial intelligence. My name is Paul Jevons, and I'm a Director at Analysys Mason. During the series of podcasts, I'll be joined by business leaders from across the TMT landscape to hear their thoughts and gather their insights on AI, and we'll be exploring what it means to them, their organisation, and the industry. 

Today, I am delighted to be joined by Dominic Gallello, the Managing Director and Head of Digital and AI at Bridgepoint Group PLC. Dominic has a career growing consumer tech companies and most recently was the Chief Marketing Officer at Bumble Inc., a well-known internet dating organisation, and currently within Bridgepoint has a focus on partnering with the investment teams on how to unlock growth through digital and AI. So, Dominic, welcome. Thank you very much for joining me. 

Dominic Gallello:

Thanks for having me. Pleasure to be here. 

The impact of generative AI on investments 

Paul Jevons:

Dominic, I wanted to start just from an investment perspective, an investor perspective. How do you consider investing in AI when you look at the market and the opportunities for investment, how do you view that and how do you assess what you're looking at? 

Dominic Gallello:

Certainly. So, Bridgepoint as a middle-market private equity investor is really looking to understand the risk-reward of any situation that we encounter and where there is an opportunity to partner with a business or an organisation. That being said, generative AI, specifically, has introduced a new variable that our deal teams really have to account for as we think about approaching any situation or any asset. So, now this is a discrete topic that we diligence with our deal teams and one that we are addressing before any deal really is signed. And that dedicated focus may not illuminate all the answers to the unknowable questions around how generative AI will unfold and impact different markets or different businesses. But we've started to develop a few frameworks that we use to really categorise the businesses and start to understand what are the key questions we have to ask around the way in which generative AI may impact that business and how that impact may unfold. 

Generative AI, specifically, has introduced a new variable that our deal teams really have to account for as we think about approaching any situation or any asset. – Dominic Gallello

So, it's not a perfect crystal ball at the moment, but it's a way that we can help frame the problem, to make sure we have as much clarity as we can despite the fact that these technologies are quickly emerging. 

Assessing AI's risk and opportunities 

Paul Jevons:

When you're looking at those companies that you're looking to invest in and the potential impact of AI, are there particular segments that you think are more open to risk or opportunity? 

Dominic Gallello:

Yeah. So, we ask a sequence of questions, really, around the business and the potential for impact. So, that starts at the market level, whether or not demand volume in a market may fall over time with generative AI, or whether the margin structure of the business might change because it gets so much more efficient. And who captures that margin? Is that the client or is that the business? 

So, what we think about are what are some of the qualities that have higher exposure to generative AI impact? And that could be delivering services or software. It could be any type of generation of any kind. It's scenarios where you are operating off of public, not proprietary data. And so, there are just a number of these attributes that we look at to determine the level of exposure a given market or a given business might have. 

I think one of the ones that has started to really emerge through the course of our diligence is the tolerance around the error rate in a given business. As one can imagine in certain sectors like healthcare, you have a very low tolerance around errors. And so, some of the weaknesses of generative AI become potentially less impactful to some of those sectors, despite being useful across a number of operational use cases. 

And so, I think we have to take it in some ways, question by question around, what is the risk of disintermediation of that business by the AI technology itself? What's the risk that barriers to entry around that market may be lowered as a result of generative AI? So, new entrants coming into the market. And based upon those questions, we have a number of attributes that we look at within a given business to determine the level of exposure. It doesn't mean that that exposure is necessarily bad, that could be an upside. It just means that there is likely to be some change in the market or to the business, that we have to understand. 

As one can imagine in certain sectors like healthcare, you have a very low tolerance around errors. – Dominic Gallello

Paul Jevons: 

I mean, it's really interesting in terms of looking at how you diligence a company that you want to invest in. Do you apply any different criteria to an AI company that you might be considering investing in? 

Dominic Gallello:

As a private equity firm that's interested in a risk-reward equation, AI-first technologies are not necessarily our primary domain of investment, given the rapid change in the landscape and the difficulty of really picking winning technologies or management teams at this point. But we certainly do invest in companies that have significant AI technologies sitting behind them. And so, a good example of that would be MIQ, which is a programmatic advertising business, that has developed significant amounts of artificial intelligence technology originally in the machine learning domain, but now really starting to experiment with generative AI capabilities. And what they're doing is helping an advertiser buy media across a number of different technology points, these demand-side platforms that they're buying on. And typically, a human was involved orchestrating a very complex ML algorithm to find the most effective buy, right? Simple optimisation process. 

And now, where generative AI is starting to come in is that we can actually not only have a human go into each of these platforms, but we provide a single interface where through natural language, they can say, "Please update my campaigns for this brand." And it's able to connect directly with those underlying technology platforms to make the changes. 

So, we look at these companies all the time and are trying to determine whether that AI, if it does have that capability in a native way, provides some sustainable differentiation. And in the case of MIQ, we found that this core component of its strategic differentiation as a business and as a whole, was underpinned by some incredible technology talent that they've nurtured over a decade, and that is leveraging the best in class of available technologies out there. 

But I would say, generally speaking, what we find in the landscape is companies like MIQ tend to be the rarity. And so, a lot of my time is spent really working with portfolio companies in terms of how do they, despite not necessarily having the abundance of technology talent available to them, make the most of these very democratic technologies available to them? 

AI's influence on portfolio management 

Paul Jevons:

And that leads neatly on to following the investment diligence decision to then your existing investment in your portfolio companies. How do you think of and assess how AI is either a threat, or potentially an existential threat to some of those, but also an opportunity of how do you really drive incremental growth? And I'm guessing that as an investor, there's your view that may or may not be 100% aligned with the management team, but there's an interesting tension there or an opportunity, I guess. 

Dominic Gallello:

Yeah. So, our approach towards our own portfolio is not dissimilar to how we look at a new investment opportunity in the way that we ask similar questions, but maybe to be more granular in terms of the type of analysis that we're doing. 

If we think about this question of disintermediation, of a portfolio company, of a new deal, where we think there's a potential that artificial intelligence, whether it be in some service or software that could be disintermediated by the technology itself, what we are asking is what do customers really value? And oftentimes, what we find is that what customers really value is not actually the component that can be disintermediated by AI. 

So, a service business is a very good example. A lot of the things that a service business does, you might say, well, generative AI can write reports and it can help synthesise data, and it can do a lot of the things that maybe a traditional consulting firm might do. But what we find is that oftentimes the humans that sit inside that organisation are often extracting insights and information from sources that can't be accessed by AI, or those sources may not trust AI to provide that information. And so, it suddenly changes your view on whether or not it's impacted or not impacted, depending on whether that core value is actually at risk of disintermediation. 

And we spend a lot of time doing customer insight work to really understand when you're paying the thousands of dollars that you're paying for this service, what is it you're actually paying for? And oftentimes, what we find is that the mechanism by which that value is delivered may potentially be at risk of generative AI, because it's a report, it's some language, whatever it may be. But actually, the core value itself is really challenging to disintermediate. 

And I think the second component is what we want to see in our new investments, I think is not dissimilar to what we want to see in our portfolio, which is tangible proof that the company is ready to deploy these technologies. This is a technology that will certainly have broad impact across a number of sectors. And if it doesn't necessarily disrupt the market, we do think it will increase the intensity of competition in a given market, right? Given the efficiency unlocks that it can provide to companies. 

And so, what we've been spending a lot of time trying to develop is the development of tangible proofs of concepts within companies. And when we think about the exit scenario for any of our portfolio companies at Bridgepoint, we want to be able to address two critical questions. Right? There is this question of the real versus perceived impact of AI on a given asset. Exit tends to be the main question, not only at the point of exit, but also at entry. We're often thinking about when a buyer is coming to this in five years' time, what are they going to be thinking? What questions will they be asking? So, we want to get over the strategic considerations around the business, but it's not only important to have that strategic lens, it's also important to have that practical lens. 

And so, we are really quickly developing, and we have been developing for the past year, a number of proof of concepts that help to show the well-defined, well-scoped application of these technologies to drive a business outcome. And most of those tend to be in operational efficiencies, whether that be in customer support, whether that be in the core operations of business, but we're looking for those use cases where we can unlock top-line revenue, not just drive a bit of margin saving overall. So, that's been our general approach with the portfolio. 

Future developments in AI technology 

Paul Jevons: 

The crystal ball question is, artificial intelligence itself isn't new. From your background in the dating technology space, there's a lot of machine learning and algorithms that have been around for a while. Obviously, we're going through gen AI and everybody understanding what that may change. Do you have a view or an opinion of where it all goes next? Or, is it quite literally just too early to call, do you think? 

Dominic Gallello: 

I think there are a couple of developments that businesses should be very mindful of, that are likely to evolve over the next five years, but I certainly wouldn't go so far as to say generalised artificial intelligence is something we think about. I think we'll be in much different positions apart from being on an AI podcast, if that were to be the case. 

But I think the two things that are really multimodal are very important for businesses to get their heads around, the ability to generate voice, video, text on the fly is going to transform the way businesses relate to customers in particular. And in particular, the ability for any business to pick up and run with models like OpenAI Sora, or even their voice training module that they recently released, similar to what ElevenLabs does, really transforms the front office in a way that I think is challenging for businesses to appreciate today, until they encounter a competitor who is leveraging this technology at scale to deliver personalised communications with their customers. I think it will have transformative impacts and we'll start to see not necessarily the transformation of the back office in the way that we've seen over the past sort of 18 months or so, but much more transformation of the front office, that I expect will have profound impact. 

And then, the second component is really the ability for these technologies to have agent-like behaviours and the ability to take action. The software engineering field is experiencing that to some degree. There's a phenomenal, I don't even know what to call it anymore because it's not really a model, it's an agent called Devin, this new software engineer. And we have a couple of portfolio companies that have gotten early access to experiment with it, but it's able to take a number of sequence reasoning steps, as well as start to actually execute code in a way that is differentiated from things like GitHub Copilot. It can go research and learn about different API documentation. It can start to go into your code environment and understand the context in which it has to write, and then it actually writes and compiles the code itself, as well as doing testing of that code itself. 

Is it perfect? No, but it's evidence of where things are moving. And so, as we think about what the transformation looks like for companies that are not necessarily augmenting humans in the way that we have been over the past 12 months, but start to think about inserting AI team members as a member of your team squad, who are self-determining active participants of your team, that is going to change our understanding of how we think about forming and assembling teams to tackle business problems. It won't just be humans with humans anymore. It will be humans with agents in some combination. That it is different from the makeup of human in the front with an augmentation tool in the back. 

Paul Jevons: 

Fascinating. Slightly scary for many, I'm sure, but no, really, really, really interesting. Dominic, thank you so much for your time. Thank you for joining me. It's been a pleasure to hear your insights. 

Dominic Gallello:

Absolutely. Thank you so much for having me. 

Paul Jevons: 

Pleasure. Thank you. 

If you would like to automatically receive future episodes, please do subscribe to the Analysys Mason podcasts. We also welcome your comments, feedback, and reviews. Thank you very much for listening.