E676 | Poone Mokari, ewake.ai & Pietro Bezza, Connect Ventures: Building the AI Teammate for Software Reliability

0:00Welcome back everyone to another episode of the EUVC podcast. It's great to have you all tuning in today's conversation is very much for our community of European founders, operators, and funders, and it centers on a fascinating bridge between product reliability in software engineering, a topic that gets a lot less spotlight than it would be needed in the European venture ecosystem. I'm thrilled to introduce our two guests, of course, Pietro Bessa, managing partner at Connect Ventures, a London-based early-stage VC known for backing strong B2B software companies like Typeform and True Layer. And then Punei Mokhari, and sorry if I got that a bit wrong.

0:36You're nodding and saying, Andres, you did an okay job. CEO and co-founder of Awake, a Paris-based startup which has just raised 2 million euros in a pre-seed round led by Connect Ventures to launch an AI team made for software reliability. Before we start the show, a quick note. If you're building or running a fund, you know it takes the right partners. At EVC, we only work with sponsors we truly believe should be part of your tech stack. Please do take a moment to hear about them. And if you do, reach out, mention your VC. It's the best way you can support what we do. Thank you so much. First up, Ace Alternatives.

1:11Every fund manager needs clean operations behind the scenes. From fund admin to tax and compliance, Ace handles it all across VC, PE, private debt, and real assets. They're trusted by some of the best investors in the world. And if you want peace of mind and a scale-ready back office, Ace should be part of your step. Finding deals and managing your portfolio is at the heart of running a fund. Synaptic helps you discover startups before others do and PortfolioIQ keeps your portfolio data sharp and ready for LPs. Together they're essential tools for modern fund managers. When it comes to legal you need a team that truly knows venture.

1:46Hainspoon supports LPs, GPs, startups and scale-ups across the full fund lifecycle. Smart managers make Hainspoon part of their stack. We have two at EUVC. Tech BBQ. Oh my God, who doesn't love barbecue? Europe's startup scene meets the loudest, friendliest family reunion ever at Tech BBQ. From Nordic founders to global VCs, this is where ideas catch fire and relationships get real. If you're building or backing in Europe, Tech BBQ is where you want to show up. And hey, if you've got a big fun announcement coming up, want to hit the headlines or just want to tell you a story about, do reach out to us because we'd love to help.

2:20And we've got some pillar partners to help you get in the right media places. They've held us land Bloomberg, CNBC, Financial Times, Forbes, and many more for the EUVC Summit. And we'd love to do the same for you.

2:54Welcome to the pod, both of you. Thank you, Andres. Happy to be. I think we should start, Pune, by having you just describe a little bit more clearly what you do more succinctly than what I was able to in my short intro. Yeah, of course. Yeah, so hello, everyone. I'm very happy to be here today. Yeah, actually, we are building AI agents for software production reliability. Developers and engineers are overwhelmed by the amount of data. And then when they want to just fix an issue or understand what is going on on the production, they need to go through all these data, correlate all of them together to just understand what is going on.

3:33And then, OK, of course, fix it. And all that just led us to just think differently, of course, thanks to AI, to build AI agents for just improving reliability of the software production. Pietro, you can comment a bit as the investor here. Why do you love this opportunity? me? So first of all, the interest came from two highly trusted sources and then people I, angel investors that I have great, great respect for. One is famous for his taste for beautiful products design. And the other one is because you really understand observability and he was a founder in the monitoring of observability industry.

4:18And so when they introduced me and they came from different parts. So as a VC, you have two referrals, both from great people, you jump on the call right on. Like, yes, that's where we met with Pune. And so to me, it was like finally having a shot to invest in the observative space, which is probably the biggest after cloud infrastructure is the biggest enterprise software industry, but it's also a very hard to craft kind of software categories, right? And I thought that with Bune and her co-founder Omid Vision, there was a way to wedge in. That is, of course, is the first part of the excitement.

5:04That is a space I was looking and spent time on it. So I had opinions about the opportunity space. And then, of course, it's all about the founders, right? Like we invested at Connect at Pre-Xeed. most of the time. So pre-product, pre-revenues, and everything that we assess is the founders. And Pune and Omid are great archetypes and examples of the product founders we loved back at Connect. Should dive into that a bit closer, just for everyone to understand how you think about founders. But let's stay on the space that you're in, Pune. And I would just love to ask you, like we're all seeing AI going through everything.

5:45And the big question is, of course, then for you, how does this impact observability? And why does that maybe deliver you an awesome wedge as a new company versus the old incumbents? Could you talk a bit about that? For years, they tried to become the data platform that just give full visibility. I would say, yeah, the full visibility is there. But then when you need it, when you need that data and that correlation, It's so hard and time-consuming for developers and SREs, of course, to understand what is going on. So engineers are needed to go through all those data, which is like a huge amount of logs, metrics, code changes, and then understand what is going on and then react on that.

6:28And at the same time, because the observability that we are seeing today in the market is so, let's say, reactive, meaning that engineers in advance need to think about what they need to measure what they need to monitor. And if they don't estimate something like that, probably an incident comes up and they don't have an alert for that. And of course, the company is just losing money every minute that they have incidents. So our mindset was building kind of operator, because of that we call it teammate, which is using the data that is already there, not adding another dashboard or another just data, and just being in hands of engineers when they need the top-notch data to just resolve an issue or even proactively understand what is going on to prevent issues.

7:14At least for me, right? Knee deep in the tech stack of a company, which means that it's something that's a bit removed from what all of us understand when we talk about AI agents all the time and hear about how it impacts consumer. And we can all understand what it will look like when companies trade with each other and those types of things. And Pietro, maybe you can comment a bit on why the a bit more hidden areas of tech are so interesting as an investment opportunity. To build an obscenity player, it requires a lot of technology. And basically, the only space of improvement is be better at collecting the data.

7:56throw tons of sensors and agents, not AI agents, also AI agents, into the infrastructure and start collecting the data in order to enable people like engineers and reliability engineers to have full visibility on what's going on, right? But again, there is a Red Ocean business. There are great companies out there doing great data collection and data monitoring. What was the missing part and is the shifting paradigm that excites us is actually now LLMs are very good on collecting not only structured data, which is what obviously tools are doing, but also on structured data, like your knowledge bases, your GitHub, conversations.

8:38And by blending this together, you create this intelligence layer on top of the data. So you're not just adding more noise, but you just extract the signals and you go straight to the root causes, which is the job that normally Pune used to do at Criteo as a reliability engineer, site reliability engineer, which is a very hard skill to have. These camera engineers are very rare. And again, they can't work 24 hours a day. And I love this idea that Pune and me, they were basically on call and they spent a lot of nights to try to fix things because when an accident happens, basically you're losing money as we speak.

9:17and so that is a very hard and critical mission piece of infrastructure that companies have to have and so this idea that actually LLMs and AI can improve the velocity of resolution that's in a reactive mode but also and this is another great part of EUA vision is we can use this to be proactive and not only give this extra intelligence to the production engineer but we can give it to the entire engineer organization and even to the product team and the sales team and the customer support team. Because again, when the product is down, customers are calling, partners are calling, everyone is panicking.

9:58Pune, you've been nodding a bunch here while Pietro is talking. Maybe you want to come in and add some of your perspectives. Yeah, I do agree that this is not a problem that because we have AI agents, you create a problem. This is a problem that has been there over years and everyone wants to solve incidents so faster. because it's important and it talks to everyone. But then AI agents are so powerful in reasoning on top of the data that you have because data of observability tools without any reasoning, it doesn't mean anything for the engineers. It doesn't understand the business of that company, the special things of that company.

10:37For example, if it's a Black Friday in ad tech company, of course, you have different production environment And observability tools is just like forgetting a lot of data, which is, as Pietro said, is so complex. But if you go further, it's developers and engineers that need to understand what is signal among all those data because a lot of it is just noise. It doesn't speak to them because it's not their language. It's not human language. It's not semantic. It's just metric and logs. And AI agents are very great in just doing and doing and redoing and retrying to just get the best data from the huge amount of data and then do a reasoning on top of that because context is the most important thing when it comes to a production.

11:23Production is not just code in your laptop. It's the whole business of your company when we talk about it. It's where your product is facing the users. And of course, it talks with everyone, with CTO, with CFO, with customer support. And then it's only developers' job to just understand what is going on, which is, as I said, it's so overwhelming now because they have a lot of noise among all those data as well. We're all using ChatTBT in these tools. And what I'm finding is that it baffles me how I can still have problems with reliability in terms of it not making stuff up and so on. And how do you, and it pulls the right things and it doesn't pull, if I run the same thing twice on the same data set, it doesn't come up with two different things.

12:08How do you protect against that? Where are we in that? Because at least as someone using the more, should we call them generic AI tools, it feels like we're still quite a far away from being able to fully trust. It's important that when we are talking about AI agents, it's kind of different with GPT. We are not training any model with AI agents. When we talk about chatGBT or any kind of those generate models, they are trained on a huge amount of data. So in two different iterations, they can come up with two different answers. But when we talk about AI agents, which are specifically used with one intention, for example, understanding what is going on.

12:50And after this incident, you have the timeframe and the data is limited. So agent tries to understand from that data to what is going on. So it's not a huge, like unlimited amount of data that is trained on. It's a amount of data that it gets as a context of that specific incident within that time frame, which means that probably you don't have different answers. You might have different language or tone, but you don't have different answers because your context is same if you just look at that incidence. And the other thing is, of course, hallucination, which is always like a question when it comes to AI, but with AI agents, you can just put the guardrails.

13:31What we tried a lot to build is like just saying, I don't know as an AI agent, because you are a teammate as well. So you can say, I didn't find enough context to give you reasoning, which is kind of scoring. And if it is more than, I don't know, 70 % sure, you just share that answer. If not, you won't share it. But there are many ways to just tackle this problem as well. And it is kind of different with how we use chat GPT in day to day. So now the audience has gotten to know you a little bit, Pune, and I have as well. And obviously, one thing that you have is incredible subject matter expertise, and you've been experiencing the problem that you're solving as well.

14:10And I would then go to you, Petro, to dive a bit deeper on what you said in the beginning, which is why you got excited about Pune and her co-founder as a team to build this and maybe also expand a bit on the investment thesis of Connect, which is of course that you're very product oriented, given that you're a pre-seed, that you're also finding conviction in founders even before there's a product. Yes. I like that you use the word thesis because we are one of the few thesis driven firm, and it's something that we have honed over time. But fundamentally, our investment thesis has been worded and declined in different shapes.

14:47But fundamentally, the underlying belief is that the best technology companies are product companies. And the reason is because product is the highest and best level for winning big adventure. A product that the customer loves fundamentally is your bigger engine for growth and for value creation. And companies and founders that put the product at the center of their strategy and execution are very different beasts than companies that are marketing-led, technology-led, or sales-led. And the reason why is that it's something that we learned over the years is that great products, they don't grow on trees like apples or oranges, right?

15:38So you don't build great products and great product experiences because you want, but it's because you create and foster and design the condition for great products to be built. And these conditions are several and it starts, of course, with the mindset of the founders. And we know that founders that are not compromising or winning through great products, experiences, are very singular. They have very common traits that we have selected and identified over the years. So if you pick connect, it's because we align on the value of product as a core business strategy. And when it comes to Iwake and Pune, I think they embody these two great traits that for us are necessary for building great product companies.

16:29One is they are very opinionated about the problems. So when we invest, there's no product. So it's very hard to assess an opinionated product, but there is always available the insights and how the founders thinks about the problem. In the past, I always spoke about new category myself, but eventually new categories are unsolved problems. And so when you find Pune, they have insider's view into the observability world and how observability software was working. but also they have an outsider point of view on how to solve it. And so they are a classic example of a founding team that like a singular insights of a problem.

17:16Now there is a why now, there is a new solution that was not possible before to solve this problem in a better way. So for years, the way to cope with preventing incident was like collecting trillism and petabio data to be able to then basically retrofit and reverse engineer where the problem is, it can take two or three days, right? And what I said, like, mission that product founders like Pune, then they say, okay, there must be a better way to do it. And then the LLMs and the generative AI was there to offer them a better way to solve the problem. I want to ask you one question before I go to Pune, and that is just you're product-centric.

18:03And there's also a concept called product-led growth, which when you say that you're product-focused and product this and product that, you do not mention product-led growth specifically. And maybe just you'd expand on that because it is a specific... Yes, Andrew. Those are two different things. Thanks for asking me because this product-led, it creates confusion. So now we talk about product companies or products first, because product-led, people then automatically confuse with product-led growth. Like, are very different things. Product-led growth is how you grow your business, and having a great product is very efficient, right?

18:43Sometimes is it not enough? Like, you also have to do marketing, you also have to do content, you also have to do sales, you have to do lots of things that you have to do. Of course, if your customer can find their product in an organic way or in a viral way and then use their product without hours and hours of demo and top-down, that helps a lot. But product-led company and product-led founders, that we call product-first to avoid the confusion, is a way higher abstraction. It's basically, how do you think about the business? And it's not just about strategy. It's just like, where do you spend the money?

19:23We believe that always, always, and of course, at the early stage, it's precede C, C, C, because you are building the product. But even when you are way scaled up, you always have to focus most of and allocate most of your capital in product development. And in those companies that we saw in direct experience that didn't manage to go beyond$50 million or$100 million in revenues, is because they failed to build product two at some point. So the problem is always there. It's never solved it perfectly. So the great product founders, the best product companies, they always find a better way, a new way to solve the problem better.

20:05And then you basically go through a cycle of innovation and you keep building product one, product two, you expand to additional problems. That's where you create multi-billion dominant product companies. And for my job is retrofit all this thing at day zero. And again, when you talk to Pudela, you see, they don't talk about, oh, we need to market and positioning, how to bring this to market. It's like, look, we know there is a problem. We will tinkering the technology that is very novel to find a solution. And if tomorrow does it come out better models, great. They will help the company, right?

20:43Even to be better at what they're doing. Hopefully, I clarify the two differences between product-led growth and product companies and product-led founders mindset. Yeah, exactly. Because product-led growth is a very specific go-to-market approach. It's not to be confused with that. Right. Thank you. Thank you. Now I'd love to ask you, you've managed to pull together what many founders out there would love to do, which is a 2 million pre-seed round before you have a product. Could you talk a bit about, and there's eons and thousands of advisors out there that claim they have the strategy and it's all about engineering, FOMO, and I don't know what.

21:21So maybe you can talk a bit about what you think were the most important ingredients in being able to do that. I think it's a big success to convince investors before having the product. I would say the first and the most important is finding those that are believing in you and your mindset. I mean, it's not easy, but the most important thing is that they are hands-on, they are ready to help. And what really attracts me from Connect was, first of all, they were super fast. That is what I really like. But the second thing was, like, they choose few over here. And of course, if you talk with a VC, they tell you that we invest in, I don't know, 200 companies in a year.

22:04It doesn't mean that they are a bad VC or there is no good and bad, but just means that it doesn't fit my thesis as a pre-seed company to get help from a VC. And I think from my point of view, what is important is if you are solving a problem, you know why this problem exists and where it exists. So I think one of the privileges that we had was, as Pietro said, when we talk about the problem that we are solving, it's so natural for us because we lived it. We lived it over years. We had the privilege again to work in one of the biggest ad tech companies in the world, Crito, where reliability is everything.

22:44And when you know the problem on that level, you can convince everyone. You can convince the first employees. You can convince investors. We even knew from first that where we want to get these insights, which is kind of interesting. We were like, maybe in the future it changes, but we were like, oh, we are always doing a lot of things on Slack as an engineer. And if I want to have a teammate, it shouldn't be just opening another tab because I'm already overwhelmed with opening many tabs at the same time. So I want to have it somewhere that I am talking with my other colleagues during incidents or whatever.

23:20I want to say it was so overwhelming for me, the problem. And it was so painful as an engineer that I was sure that if I don't solve it, there are other people that will solve it. So it should be me that I want to solve it. Because why not? Because if it's me that on like New Year's Eve at 3 a.m. I got called and I just didn't know what to do. I need something that just helped me as an engineer to decide. And AI is built for that. is built to empower human to decide faster. What use case exactly AI fits very well is where you have a lot of data, a lot of context, and you want to decide faster.

24:01And this was like, okay, this should get fixed with AI or AI agents. The technology can be different, but AI is there to just understand huge amount of context and then just empower human to decide faster and understand better what is going on. And here's a Stella example of the product-focused founder that is asked about how you raced around and ended up talking about the problem and the product. She touched on another great point that is for us is a great sign of those kind of like mindset, problem mindset that I was talking about. And it's not about the opinionated about the problem, but also obsessed about the user experience.

24:40Because they were the users of the sensitivity tools. You can see the thinking of like, okay, how does that should work? It should be another tab or should be into the Slack? It should be another data or should be more than... And for me, as an investor, I really, really value these two things. YC, two or three, four, similar media company that out of YC, one entrepreneur first. And eventually, you go and see the DNA of these founders, right? And there are people coming from FinTech. There are people coming from consultancy. There are people that built, that were excellent repeat entrepreneurs.

25:22And so it depends what is your lens on as an investor. Like we always prefer backing people with insights. They understand the user, they're empathy for the user. And they're like overly technical and can take on the technology out of the product rather than on the execution. and it's very different investment style. And so, yes, Pune is her first rodeo as a founder. She's a first-time founder. And so, yes, we overweight insights and mindset and what we call founder market feed over execution. Maybe you could just dive a little bit deeper on that. And I know that may be asking you to restate some of what you said in the beginning, Pietro.

26:08But I think that many would posit that if you had two teams working on the same thing, one come with Poonis background and one come with a serial entrepreneur background, the VCs will go for the latter. So talk about what makes you feel comfortable and why are you ready to discount the serial entrepreneur's experience in scaling and so on? Versus why do you say that the insights into the market is so important? It's so like this is where you're betting. And maybe that also gives you a natural way to sell Connect's value at us in the master because I'm sure that's also part of it. We think that everything starts with the product.

26:51So the execution, the company building and the go-to market comes after. You can't start from them. You need to start with building something the customer wanted. Again, if you break it down and unpack, it's not about what are the founders' vision, what do they know about the problem, what are their opinions to build the product. These are things you can't outsource, you can't buy. Actually, when you have product marketing, then you can go and buy execution. You can buy, you know, people that have already done that, you know, senior engineer, senior go-to-marketer, and you can grow as a founder of seed into that, right?