Here’s How I Built an AI VC Agent to Negotiate Deals
I joined BCV as an AI Engineer in Residence to create some fun AI projects. First up is a bot looking for the next unicorn.
In Matt Harris' presentation at our 2023 Fintech Summit, shared here, he breaks down why he sees fintech headed to "whitewater rapids" after the "fog of war."
For the past seven years, I’ve cohosted along with Hans Morris of Nyca Partners the Fintech CEO Summit in New York, and each has brought me the opportunity to arrive at what I’d call the state of fintech for that year. At our last event, held on May 31, I was focused more than ever before on the macro, and to get an idea of where my head was at, the main symbols I played with were “the fog of war” and navigating whitewater rapids.
These were a few of my core takeaways:
In this post, you’ll find the full video of my presentation, the deck itself and a transcript. If you’d like to go deeper on how generative AI is poised to shake up financial services, I’d encourage you to read my recent blog on the topic (Attila the Hun makes an appearance), along with a comprehensive set of “field notes” spearheaded by my partner Sarah Hinkfuss.
Below is a full transcript, with light edits for clarity.
For those of you who’ve been here seven years, you’ve seen quite a parade of different topics. But mainly they have to do with, let’s call them micro topics within technology, within financial services, within the specifics of fintech, topics around decentralization, around embedded financial services. Lots of things that are on my mind are on people’s minds over time. But today what I couldn’t get out of my mind is the macro.
And as the agenda was coming together and I realized that I would be following Larry Summers and giving you my musings on macro, I started to grow very, very nervous about that. What if he came in and said everything was just fine, for instance. That would run certainly afoul of my feelings, thoughts and observations. … I won’t quote the off the record remarks of Larry Summers, but suffice it to say he doesn’t think things are just fine. So I think I’m OK. I think he and I will directionally agree.
This talk’s going to have three chapters. The first is really to characterize the facts on the ground as they stand today. We’ve been calling it the “Fog of War” throughout the day, and it does feel surely, at least in the board meetings that I’m in and the investment committees I’m in, it feels like a maybe not unprecedentedly, but disproportionately, foggy time. We all had a very thick fog for a couple of months in 2020 in March, April and May, and then we were off to the races. This has been a more durable sort of fog and doesn’t seem to be lifting. So I’m going to characterize the state of uncertainty that exists.
I’m going to make what you’ll see is somewhat of a prediction that what comes when the fog lifts will look like whitewater rapids, that things when they become clear will become clearly disconcerting. And then a few remarks about why that is.
And then finally some thoughts about what you can do about it in your seat, primarily here, thinking of the CEOs who have this incredible responsibility to run businesses both in uncertain times and then sometimes in certain and in scary times. So those are my comments.
We asked — well, I should say when I say “we,” I mean Noah’s 15-year-old son asked generative AI to put together a picture of, I think the exact prompt was “nerds walking through a fog of war.” And I think this is a little too bullish, I would say. I mean, looking around the room and then looking up at the picture … I’m not sure they nailed the nerd thing, but maybe it’s the visor the one guy’s wearing? I don’t know. So this is where we are. We are dressed for battle and we can only see 10 feet ahead of us.
Zooming out a little bit, what we do know is that there didn’t used to be a fintech industry. This is the history of fintech investing over time. For a long, long time, when Hans was running financial services and fintech at General Atlantic, for instance, and when I was working at my prior firm, fintech wasn’t really on the radar screen. And then boy did 2021 come along and change that. And we’ve already started to see again — as everyone in this room surely has felt — a retreat from the heights of 2021.
There was a funny dinner that I went to hosted by Silicon Valley Bank before the Tim Mayopoulos weeks, and Greg Becker, who was the then CEO of Silicon Valley Bank, had a group of 12 New York VCs in a room. And he was citing numbers around just US venture capital — not just fintech, all of US venture capital. His number was $360 billion in 2021. And we’re having this dinner in May of 2022. And he said, “I want to go around the room and everyone has to guess how many months or quarters will it take before the US venture business gets back to the 360?” At that point, his numbers were predicting that US venture in 2022 would be around 180, so it’s sort of fallen in half. How long before we get back: how many months, how many quarters?
People given that prompt were sort of saying, “I don’t know, 15 months?” Or if they were thinking longer timeframes, they would denominate in quarters. “Maybe it’ll be six quarters, maybe it’ll be eight quarters.” And then they get to me and I said, “A minimum of 40 quarters.”
And everyone was like, “What are you talking about?” That wasn’t even really on the menu.
When people talk in quarters, they don’t usually expect you to get to 40. And I pointed out that it took 18 years for the venture industry to get back to its peak in 2000. It wasn’t until 2018 that it reached that peak. People have short memories sometimes, but limited partners have a way of really remembering times they find painful, and they found 2000 really painful. And something tells me that when all is said and done about the vintage year 2021, they’re going to find that pretty painful too.
So I don’t think — notwithstanding Greg Becker’s boosterism, which we’ve all learned a little bit about the consequences of — I don’t think we’re going to be back to that anytime soon. And in fact, since the annualized numbers of 2022, over the course of that year, we saw the pace go from whatever the annualized number was in the 80s to an annualized pace of 44. A little bit of a bump in Q1, particularly with one notable financing at Stripe.
But we’re certainly going to get cut in half again in 2023, notwithstanding generative AI. And so we’re seeing just a continued drop to what I think will still feel like a healthy plateau of funding. Venture firms have raised a lot of money. They’re not going to give it back. They’re going to invest in companies. We at Bain Capital Ventures raised $1.9 billion in Q1 of this year. We fully intend to invest it and are pacing to invest it in three to four years. And that’s what we hear from our peers, as well.
So money is not evaporating and the fact that it’s not going to quickly return to the peaks of 2021, I think everyone should take actually with some dose of comfort. I think we all look back on 2021 as an exciting year, but it didn’t feel like the most rational year in the history of capitalism and not one that again is going to produce a lot of fantastic companies or fantastic investment returns.
We’re also seeing, I think — very, very encouraging from my perspective — a return to a focus on seed and early stage investing, where in fintech I almost got the feeling during the COVID boom there were no really new great ideas. Because all of the deal flow activity, certainly by dollars, were just going into later and later rounds of the same companies that were terrific companies. They may be less terrific now having been foie gras-ed hundreds of millions of dollars, but it was not about this Cambrian explosion of new ideas and the way that I experienced, for instance, fintech in ’15, ’16 and ’17.
And so if what we see is in fact this increasing return to the dollars and deals being concentrated at the very seed and early stage, I think we can all be really optimistic about the medium- and long-term future of financial services. That will mean the founders, like the ones in this room, are once again feeling inspired to start novel companies. And you have to have more to raise money now as a startup than you did for any time in the past five years. And the fact that folks are still bringing themselves to do that, and that they’re finding capital that’s interested in their causes, I think is very encouraging to me.
So this is a different chart. This doesn’t talk about volumes, this talks about returns. This chart shows every vintage year, this is how we venture capitalists think about our industry. You start a fund in a given year, you are starting in a cohort of other funds that started in that year and you’re going to generate a certain return and your cohort is going to generate a certain average return. And it’s also going to generate a certain top quartile return, which is to say that firm at the very bottom of the top 25% of that cohort, they earned X. And so this chart shows what that X is looking backwards at these given years.
… What does it tell us about the relationship between economic cycles, technology cycles, and venture returns? Well … these charts don’t settle for a long time. So as you look at this chart, think about this. When I looked at this chart in 2002, a long time, 21 years ago, and I looked at the line that was the year 2000, so two years prior — as measured in ’02, the top quartile was defined by 83% in net IRR. The best venture year in history is the way it looked in ’02. Ultimately that settled at -7. So these things have a way of shifting. Not once in my history of studying this, have they ever shifted up. But they do sometimes settle down and sometimes they just totally invert. And so when you look at the areas marked “bubble building,” well, maybe they don’t go to -7, maybe they go to -17, maybe they go to 14 — I don’t know where they go, but they’re going to come down.
And when this is all baked, we’re going to look at that middle box across this 20-, 25-year period and say that was the single best time to be a venture investor in this couple of decade span. When all that stuff to the right of it settles down, it’ll settle down well below the average of that box. And so when we look at this, the boxes overlay the technology trends. That tends to be what we think about as the most important driver. In the early part of this century, there wasn’t much going on. It was still very much an establishment phase for technologies that didn’t have commercial applicability. So much happened between 2010 and 2015 in terms of new platforms, and that led to incredibly exciting new companies.
And so the question we ask ourselves as an industry is which is more similar to today? Are we in a period where generative AI and low-code and no-code tooling and all the other things that are many of the themes people in this room are working on, are they similar to mobile, social and cloud? Or are they in their awkward adolescence and unlikely to produce great companies? And that will tell us where these years shrouded in mystery end up landing. So this is our current fog.
But there are some certain predictions we can make. And this leads us to our thinking on whitewater. Is there a realistic plan to a soft landing? … We would say, our macro team at Bain Capital, almost certainly no, very unlikely that we have a soft landing, almost certain that we have a recession.
Where can inflation be heading? I think down. I think our macro team feels down. Probably some stickiness. But I think already if you look at the signs of the signs of the signs, all the indicators are headed down. We’re not going to have terminally sticky inflation.
But this is the slide that I’m stuck on. This is from our macro group and I’ve got a bunch more slides that back it up. But just the rate crisis causing the banking crisis, these shocks could create an epic and long-lasting credit crunch. Larry Summers again said, “In a credit expansion, almost no one defaults.” And the inverse of that is wildly, painfully true that in a credit contraction, good borrowers sometimes default. Because when you get to term, you can’t for love or money refi.
And I don’t know what probability is associated with it, but I think we may finally be coming to the end of the American banking system with its 8,000 banks. And the end, the death rattle of that odd system could produce quarters and quarters and years and years of contracting credit for companies and projects of all types, and have this really massive, and in the near- and medium-term, tumultuous impact on everyone in this room.
So that’s what I want to talk a little bit about. These are bank failures over time, and it’s a pretty striking chart. Striking, based on the label, in that not just any three banks failed, but three huge ones. But also in that we haven’t seen any of the little bubbles yet. In my experience in economic history, bank crises don’t just come and go. They come in chapters and waves. And we are very definitively at the end of the first chapter in what is likely to be a multi-chapter story.
And now we’re going to get into very nerdy slides. And again, I have a captive audience and so I thank you for your forbearance. But this chart to me tells an incredible story about the American banking industry. This is percentages of risk-weighted assets, and this is a historical chart from 2019. And it buckets the banks in the country by asset size. The circle is the profitability, so the return on risk-weighted assets. And you can see even then, the “now” part of this chart reflects the fact that five years earlier the most profitable banks in the country had been the middle-sized banks, and by 2019 it was the largest banks that were the most profitable, at 243 basis points return on risk-weighted assets. And all the other buckets were worse, even though historically they’d always been better in terms of their profitability on a return-on-asset basis.
The primary reason, and let me just define this chart a little more, the dark blue is … think of it as net interest margin, so the spread on your loans relative to your cost of deposits. The light blue is your non-interest income. So fees, basically. In certain cases, those would be fees associated with late fees and nuisance fees. For the big banks, there are fees associated with very healthy fee businesses that are non-leading, but small banks don’t tend to have those businesses. And then finally the gray bar is the expenses. And the circle is the net of the two blue bars against the gray bar.
So what happened between 2015 and 2019 to swing the profitability far to the right is an expense story. Post-financial crisis, and increasingly over time, it’s just become much more expensive to be in the banking industry. And so those costs that are fixed somewhat, at least in nature for small banks, became harder and harder to bear as a function as a percent of assets.
And so how are they surviving? They’re surviving because they have this incredible NIM, they have borrowers in, generally speaking, sort of quasi captive geographies who don’t have a lot of great options, and they have depositors who they treat very poorly, who they pay very low interest, even relative to other banks. And you just look at that as a function of size and you would just say, wow, those spreads are highest in those smaller banks, who have these captive geographies, who pay less in the way of deposits and who charge more in the way of interest. And then even with that said, they’re still the least profitable. So this is what we had before SVB.
And this again, a fascinating chart to me and perhaps only me, but if you look on the left, this speaks to this historical geographic advantage. There used to be a linear relationship as recently as 2012, if you look at that basic straightness of the 2012 lines in these two geos — which are quite representative — which basically says your percent of deposits in a market will basically correlate with a very high R-squared to your percentage of the branches in that deposit. So local, physical market share will determine how many of the deposits you get.
But now 45% of new checking accounts are opened with people who don’t have any branches, who are just these mega banks who are in some markets, not in others, but who have digital marketing and brand name and much better digital product. Such that they are on a greater than 2x basis poaching deposit accounts from these local markets that historically, as recently as 11 years ago, were basically physical geographic markets, like supermarkets. … It was retail. It’s gone from retail to e-commerce. And these local and regional banks have no corresponding strategy to deal with that from a deposit gathering perspective.
And so we’ve seen that show up already again … but well in advance of that future driven by the episode at Silicon Valley Bank, or exemplified by it, we saw this massive shift, where it used to be in recent memory — 30 years ago — 80% of the industry’s banking was done by smaller banks, and that’s been cut in half.
So now you have this situation leading into 2023 where those 8,000 or so small and medium-sized and regional banks outside of the top 20 … their market share has been cut in half, their economics have gone from the most favorable to the least favorable, all the while defending the widest spreads relative to the large banks. And then these things happen.
You have these new conditions where — [it’s] been reified in laws as far back as the enabling legislation of the CFPB — but finally, open banking has come in a relatively vigorous way to the United States. So consumers can more seamlessly tap into their bank accounts and allow for automated actions on those bank accounts. You finally have real-time payment rails. Been talked about for a decade, RTP has been around for three or four years, but very few banks enabled it until the last year. FedNow launches in two months.
Those two things will hit maturity very soon. And we generally have a consensus — I bet in this room, certainly at BCV — that the idea of a concierge who can manage your financial life powered by generative AI is no longer a pipe dream. It is very likely to be available and it’ll be instantly an extremely high ROI concierge because the things it will do will help you avoid fees (you should really never pay a fee again), it will help you refinance your credit (not just your mortgage when rates are low, but all of your credit on your balance sheet), you will be able to refinance in an automated fashion whenever it is advantageous for you to do that.
And all of your deposit accounts will find the highest yield available to them frictionlessly without your even having to know about it.
And the amount of money that you think you need to keep in your checking account will be way lower than what you currently think it is because this concierge will make sure you always have the right amount of money in your checking account.
So the shift between checking and savings broadly defined, and then the frictionless availability to seek yield on that savings, this is not even in the numbers yet, but we can see it happening very clearly in some number of quarters. And so think of all those banks whose entire business was based on branch density and, frankly, whose entire business model was based on taking advantage of lazy customers — which is not the most noble way to put it, but it is the case. That’s all gone.
There is no reason to have regional banks anymore and they will be adversely selected against. And the large scale digital banks or hyper-focused banks, not by geo but by industry or other psychographic or demographic, will be all that’s left. And the death rattle of those 7,000-plus financial institutions is going to be very rough on the economy.
So what do we do about that? I’ve got two minutes left. I’ll probably leave some open questions. But resilience. Resilience. Larry Summer said “robustness.” I refer to it as resilience. …
The right choice is to balance the highest chance of winning with the strongest chance of survival as you make these choices. And there are trade-offs — these are not easy choices. But playing always to win at the expense of resilience and robustness in a whitewater rapids situation is a losing strategy almost all the time, and certainly in any Monte Carlo simulation across the probability spectrum.
So this is an analysis that my partner Sarah [Hinkfuss] did of all the public fintech companies and basically which ones survived the downdraft in public markets of fintech companies from the fall of ’21 until more or less the fall of ’22. And you can see resilience: How well did your stock price do as a category? And then where did you end up in terms of year multiple? And you see a bunch of characteristics that all to me correlate with a robust business model. Do you have high gross margins? Does your revenue reoccur? Is it hard for your customers to change? Are they enterprise or middle market B2B customers who made hard ROI decisions and are committed to your product? Are you in lines of business like payments and wealth that do not have credit risk, liquidity risk, meaningful amounts of reputational risk? So some of these you can’t change. Many of these you can change as you think about building resilience and robustness into your business.
So I’m going to come back to this slide because … today’s all about predictions. You can take them all with a grain of salt. Certainly if they’re inconvenient to you, you can ignore them entirely. But my prediction is these years are going to be fantastic for venture returns. And we’ve already raised our money, so I don’t have to say that. Because the thing that I lived through during that incredible period in that middle box was, this new red set of words that I put up in this second version of the slide, “The banks were KOed.” It was a period of massive distraction for the incumbents across banking, insurance, asset management, wealth management, broker dealers. All of the incumbents forgot about technology, they forgot about innovation, they focused on regulators and compliance and survival. And it was a field day for every company that … I was lucky enough to work with during that period.
And it wasn’t easy because those same environmental considerations had to be considered. But everything gets better when your core competitors are massively distracted. And let’s face it, we’re in the business, all of us, of trying to change the system, of trying to disrupt or overthrow or create a financial system in 10, 20, 30 years that is better, more technology-driven, more fair, more efficient, more advanced, more progressive than what we’ve all inherited.
And I’ll leave you with this for all the new Formula One fans in the crowd. [Ayrton Senna quote: “You cannot overtake 15 cars in sunny weather… but you can when it’s raining.] The conditions that allow things to change dramatically, the conditions that allow startups to beat incumbents, that force customers to do not the safe thing, but the right thing — which is often trying the new product — are messy conditions. They’re difficult and challenging conditions. And if you build a resilient business, what you should be hoping for is market conditions that test less resilient businesses. And so I couldn’t be more excited and also grateful to you all for permitting me 25 minutes of musings and ramblings. So thank you for that. I appreciate it.
I joined BCV as an AI Engineer in Residence to create some fun AI projects. First up is a bot looking for the next unicorn.
Technology development is unpredictable. You have to build things if you want to understand them.
Together with Bain & Company and Coresight Research, we are proud to present the inaugural CommerceTech Power List, recognizing the most innovative marketing, merchandising and loyalty related technologies.