Matt Henderson - Cofounder and CEO of Phoebe

0:00Matt Henderson:I am here with Matt Henderson of Phoebe. Welcome, Matt. How are you doing? Hi, Seb. Thanks for having me. It is my absolute pleasure. You've obviously just announced, well, recently announced a very big seed round. But before we get into all of that and all that good stuff, can you just give us a quick pitch, quick intro? Who are you and what are you building? Yeah, so I'm Matt. I'm a co-founder and the CEO of Phoebe. At Phoebe, we're building an immune system for software. And that is AI agents that constantly investigate and fix emergent problems in software. Amazing. And where did this vision idea come from?

0:35It was partly we'd been a kind of a recipient of the pain ourselves, sort of leading large engineering teams at companies like Stripe and elsewhere. and always getting frustrated at the way that our most experienced engineers were just getting pulled into reactive work in order to investigate problems and fix them before sort of things spiraled into serious incidents.

1:05Matt Henderson:Yeah, and I've seen this, it's kind of been referred to as like an immune system for engineering teams or for companies tech. And you've also been live with a couple of clients already. So Trainline, I think is one. So, TradeLine is one. I think PPRO is another one. Are you able to talk about some of those real-life use cases examples? What kind of results have you seen? Yeah. So, we've been working with 20 early access users as we've been building the product. We're now starting to move from that initial private beta into an open beta. And there's really been three main use cases. One is responding to symptoms of problems and being able to investigate them faster.

1:53Another is responding to alerts and being able to understand the root cause and the resolving action of those alerts as quickly as possible. And then the third is being able to analyze problems after the effect post an incident in order to determine how best to avoid the problem again in the future. And so across those 20 users, and P-Pro and Trainline are two of them, we've found that this can really help the engineers and the team to move faster and get to the bottom of an incident quicker. and also that it really drives down the recurrence of similar problems again in the future. And so those have been the key early findings.

2:42One of the things that we expressed in the vision as we announced the funding round recently was the degree that our aim is to move earlier and earlier in the life cycle of a problem. We think there's a lot of potential to diagnose problems before the actual outage or failure really materializes into a customer facing impact and implement preventative changes sort of that early in the life cycle of a problem. And so what you'll see from us over the coming months is more and more progress earlier in the life cycle like that.

3:20Matt Henderson:And at the moment, but your software isn't actioning any changes or fixes at the moment. It's just finding them, highlighting them, kind of helping engineers to fix and resolve them and learn from them. It's not actually doing the fixing itself. So actually, for some of those problems, it will generate a pull request with the code change. At the moment, they are all being reviewed by by an engineer. So it's not doing the final review, but it's generating the change itself. But there's also some actions where it will propose instructions for the actions that are going to resolve the problem. And then an engineer will go and implement that change.

4:04Matt Henderson:And do you see a future as you move earlier and earlier on in the problem? Do you see a future where your software, your AI will be able to do the entire process without humans? So I think there's different risks. You could think of a sort of stratified different types of actions based on how risky they are. When you are allocating additional compute resource to a particular process, whether you are, there's a type of release that is quite low risk to roll back a deployment. There's certain types of actions where we're implementing it without a human review is actually low enough risk that those will be the earliest ones that get adopted.

4:51And then I think there will be change over time. As Phoebe demonstrates very high precision with the actions that are recommended, then there will be demand from our users for some of those actions to be implemented right away, especially when they're in response to this sort of early indicator immediately prior to an outage. And that's the best way to be able to fix a problem before it starts.

5:24Matt Henderson:Got it. Okay, very nice. Can we touch quickly on the fundraiser itself? So I think it was$17 million, is that right, seed round led by Google Ventures and Cherry Ventures. What were the key sort of milestones or unlocks that enabled you to raise a round of that size? So, one of the things is the kind of scale of the problem that we're addressing. Another is the degree that we're an experienced team, repeat entrepreneurs, a network of colleagues from Google and Amazon and elsewhere we've worked that are coming together to work on this difficult but important problem. And so those things mean that you need to place a big bet as you're going after it.

6:14And so I think that's a big part of it. The other part is just the degree that we're making progress towards solving that problem. And so having these early users that can attest to the impact that we're already having on their reliability, I think gave investors confidence that there's going to be even greater impact in the years ahead.

6:39Matt Henderson:Yeah, I mean, I can imagine you are almost like a VC's dream founder, right? And for those who don't know, you were a product director at Amazon for a number of years, you know, at Amazon's relatively early days. You founded your own company, which was then acquired by Google. You spent some time at Google. You've then most recently been working kind of as a Stripe Europe CEO. So you have got an absolutely amazing experience, both as a repeat founder, but in some of the world's leading tech companies, you're almost like collecting them all. You know, you're not far off having the full fang on your CV.

7:15Matt Henderson:Are you hoping for like an Apple or Meta acquisition to sort of complete it all? You know, our aim is to build a big independent company. with our first startup range span we we it was acquired by google in 2014 and it was an early exit and it was a great experience and and you know it's also been great to work at some of these leading large companies but part of the draw to go back to early stage and being an entrepreneur is to build a big independent company. And, you know, we're also passionate about tech in Europe as well. And, you know, we have a great engineering team across the UK and Europe.

8:01And we want to build this big independent company that's really centered around Europe.

8:06Matt Henderson:I love that. And where do you see your future firmly rooted in Europe? Or are you already seeing pulls or trials or customers coming from the US? Yeah, so already sort of maybe 40 % of our customers are US-based companies. We will be global. We will have folks based in the US as well. Our team is all sort of concentrated in the UK and Europe now. And we believe that for product development in particular, there's really amazing talent here. and so for the foreseeable future, that will be the way that we build. Yeah, I hear that a lot from especially very early AI founders that are building their product team here in Europe where there is a very high density of talent that are generally cheaper and they're building their GTM function in the US because that's where most of their customers are already coming from or they see most of them coming from.

9:07Matt Henderson:Is that the split that you envisage or is it going to be slightly more blurred than that? I do think that we'll build up a commercial log in the future in the US. The further you look in the future, the more that can be blurring. You know, it's also true that Asia is a critical market and companies that do global well tend to have folks locally as part of that. And it's not just about selling. It's about understanding what's going to have impact for customers locally as well and, you know, the right feedback loops into product. So we want to be a Europe-led global company. Love it. You mentioned the size of the market, the size of the problem that you're tackling.

9:57Matt Henderson:Are you going after specific use cases? It looks like from some of the pilots, the clients you've already got, it's already quite a wide range of different companies. Can you roll out your technology at any sort of company or industry or are you focused on a few industries at first? So it is broad. So many industries now have become tech industries. They have their own engineering teams. they have their own um you know they're building in the cloud and um therefore they're also having breaking changes that are causing problems and so we expect there's going to be a variety of verticals we're already um the sort of three that that we have uh a sort of particular uh customers from are fintech b2b sas and then digital commerce and so we expect those to be our biggest three but it will spread beyond.

10:54You know, we've also found there are sort of some sizes of companies where it's natural to be pioneers in how they're achieving reliability and therefore work with early stage solutions like Phoebe. But over time, you know, we expect we'll be working with a lot of large traditional companies as well.

11:16Matt Henderson:At the moment, there's sort of the rise of kind of vibe coding apps and Vibe coding in general, whether they're kind of helping people who can't code learn how to code or kind of giving engineers the tools to become better coders, better developers. Do you see an increased risk of technical issues arising from the use of these tools? And can Phoebe help mitigate that? So AI can be a negative and a positive when it comes to creating risk of how it's affecting your systems. And it depends how you use it, essentially. So one of the things that will create more risk is that there are fewer people that have a strong understanding of everything that's going on in the system, especially when there's a greater amount of code and a greater velocity of change that's impacting that system.

12:16And so we believe that that will sort of increase how opaque a system is, which is part of the challenge when it comes to investigating potential problems. And so not only does that introduce some risk, but it also makes the investigation process harder, which is the source of frustration that we're addressing with Phoebe. um now it is also true that um ai can help to um rapidly identify some potential problems before they they get deployed um and it can um be sort of an unlimited set of second eyes on changes and so uh you know as it is uh improved and rolled out sort of at creating new software it also needs to be improved and rolled out at the checks and balances that help to make a resilient system

13:19Matt Henderson:got it okay so you're you're yeah the future looks like i guess both ai enabled engineers they're also they've also got these tools for essentially like qa and before they kind of push push go on on the releases um you know you're a very you know you've got a huge amount of experience in this space you're building a tool for engineering departments have you got any advice or suggestions to people working in engineering today on the types of tools that they should be using to either prevent these risks from arising or to help them mitigate them? One of the bits of advice is just experiment and evaluate new things.

13:55And we've already seen over the last year in particular, there's been a lot of change in the way that people are using AI to build new features. And if anything, there's this second kind of paradigm of AI engineering, which is reacting to the production environment and the way that that's not about building new features, but it's about using the presence of anomalies as this data set to then create more AI-driven flows. And so experimenting with new tools, setting up the right data environments in order to be able to use those tools in an informed way, you know, those things are going to be critical.

14:44Matt Henderson:Amazing. Yeah. Can I also ask about your general views on European tech? You said you're like an advocate, passionate about European tech. how did you find it building and raising money for a solution like this which is which i guess is is all about kind of like intelligent infrastructure it's not as sort of like as much there's not as much hype around it as some of the other ai tools where you can immediately see huge amount of revenue huge amount of users how how did you find explaining this proposition to to vcs firstly um you know it's actually worthwhile just taking a step back and looking at how tech in Europe has transformed.

15:24And it's vivid to me because I experienced starting a startup back in 2011 and doing it again now. And it's a totally different world. There's just dramatically more venture investing in Europe. There's a lot more strong, really product-oriented engineering talent and the ecosystem has changed and what people forget when they see that there's still some you know the presence of the big fan companies that are all listed in the U.S. and so on what they forget is actually the rate of change in Europe's being greater than the rate of change in the U.S. The percentage of venture funding that Europe and the UK accounts for is almost 20 % now.

16:16Back in 2010, 2011, it was less than 10%. So it's actually this incredible transformation. And for us as one data point, that meant that it was easier for us to raise money, but also the investors and potential employees that we were speaking to were all very ambitious and had the same kind of animal spirits that we have long admired about the US tech ecosystem. So it's alive and well here.

16:47Matt Henderson:You know, that's what I love to hear. I mean, it definitely feels like in the last 10 years has been a real change in both an approach and the level of ambition. And even the last few years, you know, people across Europe are saying, we want more, we want to do more. The ambition has never been higher. The drive has never been higher. And people are almost feeling patriotic about Europe. They want European companies to really succeed. And I completely agree. That's all very, very positive. What do you think we need to do to do even more or to take European tech to the next level? So one thing that I think has been present in the US for a longer period of time is the right alignment around chasing very, very large outcomes rather than the equivalent of what I myself did with my first startup where we sold it to Google three years in.

17:46And so what I think you see more commonly in Europe and the UK now is a bunch of people either like me that second time around are shooting for the moon instead of shooting for the early exit. But also you get a crop of a new generation of entrepreneurs that are aiming for that big independent company right from the outset. And so I think the more that people have seen examples of tech entrepreneurial success, the more that it feels possible. And you get this feedback loop that has enabled the catch-up trajectory that has been happening in Europe. And so it is about people, actually even more so than it's about money and venture funding.

18:45And, you know, I see a lot of graduates from the first startups or former employees from some of these companies that are now, going after new ambitious bets in the same way that folks are doing so in the Bay Area in California.

19:03Matt Henderson:Yeah. And that flywheel is definitely starting to turn. We're seeing these founder factories crop up from across Europe, whether it's Revolut or Monzo or some of the other European ones that, yeah, their ex-employees are now going out and starting their own and raising their own. You mentioned your previous startup a couple of times. Can I ask, Do you have any regrets about exiting when you did? Mixed feelings, you know, it was a great journey. It was also great being part of Google as well. And when we sold the company, we did it knowing that we were all still early in our careers and it'd be time to have another swing.

19:50and so that's what we're doing now. But it also, you know, it meant that there's folks in the team that have gone on to become entrepreneurs and so, you know, even those less notable sort of events in European tech have gone on to sort of influence the ecosystem and sort of breed more entrepreneurs.

20:15Matt Henderson:Yeah, okay. Yeah, no, it's an interesting perspective because you've done the entrepreneur journey, you've built one, you've sold one, you've been at the scale up as well as you've been at some of these tech companies as they go from being quite big to massive. And now you're back in the trenches again. I guess what would be the biggest lesson that you've learned from your career to date that you are now applying at kind of building Phoebe? I think it's about how you attract talent and inspire that talent to really keep learning from what you're building. And so we, you know, at RangeSpan, one of our sort of breakthrough parts of the company was an ML analytics long before AI was as sexy as it is now feature that we thought of having been inspired by the customers that we were working with.

21:16and we expect that at Phoebe our success will be both based on the vision that we have at the outset and how the talent that we're bringing into the team how they are inspired by the customers that we're working with and those engineering teams that are using the product that will influence what the next features are and the next ones beyond that and you know the the great enduring tech companies they tend to be built by these accumulation of many wins and many innovations and that's really what drives the moat and and success and so so that's you know perhaps more than just one takeaway but but at its heart it's a it's a kind of an attitude and

22:06Matt Henderson:a culture got it no that's um yeah it's an amazing lesson or yeah yeah like i said you You've had such an interesting career, but it's really interesting to hear your perspective on both tech early stage and now doing it all over again. But Matt, I think we're out of time. It's been an absolute pleasure chatting. Thank you so much for coming on. I'll be watching your journey. And if you've ever got anything that you want to talk about or announce, if you want to come back on, just let me know. I'd love to have you. Awesome. Thank you, Seth. Thank you so much. Bye bye. Bye.