From PM at Stripe to Building an AI Startup, a Recent Founder's Journey - Mounir Mouawad

Talk transcript

1. 0:01

   Hello, and welcome to this presentation. My name is Mounir. I am the CEO and co-founder of Portia AI. Um, I'm here to, uh, share a bit about my journey as a, a relatively recent founder of an AI startup, uh, coming out of a variety of, of roles, especially in product at, uh, big tech, most recently Stripe, and

2. 0:21

   before that Google and Amazon. Um, I'm hoping that this-- the presentation, uh, is useful in particular for people who wanna make that leap and start their own company, or people who are in a, a similar space as, as sort of, uh, uh, stage as me, uh, starting their own company in AI at the moment.

3. 0:40

   Um, and, and hopefully you'll take some, some useful notes from this. So, uh, before jumping in very briefly, um, I, uh, have done a variety of roles I mentioned in big tech in particular, um, always based out of here in London.

4. 0:54

   Um, so I launched the deals, uh, program of Amazon in the UK. It's the thing that they use to run Black Friday, Prime Day, and all these big deal-driven events.

5. 1:03

   Uh, I launched and grew Google Pay in thirty markets in EMEA, and I led most recently Bank as a Service, uh, at Stripe for EMEA as well. But at some point, I always wanted to do my own thing, and at some point I felt like, um, you know, like it-- I really need to, to, to get to

6. 1:18

   it. And I felt like nothing else made sense anymore, and that frankly, I felt a bit like a cop-out. [chuckles] I felt like, "Damn it, I need to do this."

7. 1:26

   Um, and so fast-forward to, uh, whatever it was, June last year, uh, so a year ago, um, Portia AI became that, that thing that, uh, that I started with my co-founder, Emma, that I met at Stripe.

8. 1:38

   Uh, it's an open source SDK for building production agents,

9. 1:42

   um, with a focus on regulated industries. And I'm not gonna dwell too much on what, uh, Portia itself does, 'cause it's not really what we're here to talk about.

10. 1:49

    There'll be a link at the end of this presentation, uh, if you're interested to learn more about the product itself, which I hope you will be. Um, but anyways, let's, let's focus on, uh, on the current topic.

11. 2:00

    So, uh, let me share what, what I learned so far, and I hope you find this useful, and I'd love to, to hear your thoughts on it in, in the comments as well.

12. 2:08

    Uh, the first thing I wanna talk about is what, what user problems and user discovery feels like at, at the moment, and how freaking different that is from what I've done, what I've seen so far.

13. 2:19

    Uh, I'm going to use a lot of video-- I'm a gamer. I'm gonna use a lot of video game analogies, um, uh, to, to describe this. I hope some of you can relate to it and find it, uh, find it a bit fun.

14. 2:30

    Um, but like the first thing I'll say is, uh, normally user problems are... Uh, they have some shape to them, right? And they have some adjacency to other user problems you're, you're used to.

15. 2:41

    You're at Stripe, you do payment products, and a user wants to do card issuance to, uh, uh, you know, use the money that they have in their Stripe balance.

16. 2:49

    I mean, it's-- there's some adjacency, something to anchor to. Uh, in the current setup, uh, partly because it's a startup, you're going from zero to one, and partly because it's AI, it really feels like that first time you try out a, a sort of Souls-like game, and I'm using here a, a screenshot from Nioh, which is one

17. 3:07

    of the worst ones at that particular challenge. And you're trying to configure your character, and you can't figure out how to, [chuckles] how to combine, uh, the gear you sort of directionally know that the stats all need to go up, but you don't know what-- how to, how to get there.

18. 3:20

    Uh, right? And so nothing makes sense at first. And this is a bit how-- what AI feels like at the moment because users don't really know what their problems are.

19. 3:28

    They sort of know that they want to use AI, and they know that there is some opportunity there, but they don't-- the, the user problems are almost like an emergent property of the space.

20. 3:36

    They are not defined yet. Um, a-and what is possible and what the technology allows you to do is changing very quickly. And so instead of your more conventional three-month or six-month roadmap that you sort of revisit, um, you just have to make directional bets and iterate very fast on them, um, uh, through putting products out there and,

21. 3:57

    and looking at the user reaction. And so your, your, your anchor for you and the rest of your team is, is really just hypothesis-driven, um, deliberate iteration. That's the bit that gives you the focus that you need as opposed to, you know, the conventional roadmaps, um, that we're used to in the trade.

22. 4:16

    Um, the other thing is like, uh, all this like [chuckles] this really reassuring, uh, idea of using, uh, user storyboards and having sort of user-- critical user journeys for your pain points, et cetera, that you can anchor on.

23. 4:33

    Uh, those things all sort of feel like they go out the window to, to a large extent. And so it feels like for those of you who've tried that game, No Man's Sky, where like you start the game and you're like, "Huh?

24. 4:45

    Like, what am I-- what's the story arc? What am I-- what's the mission? What am I supposed to do next?" And instead, you really are creating your own narrative here.

25. 4:53

    Um, and so th-this oper-- this-- you really have to stretch that muscle where you have to help your users out of that initial inertia of trying AI for their, uh, you know, if you're in a particular vertical or you're horizontal for a specific use case.

26. 5:10

    Um, a-and you have to work with them to create those narratives and help them anchor on a use case. Uh, and once you're there, you're starting to gra- to gradually move to more familiar territory where you're iterating with them on the use case, and you have something to anchor on where you can refine your product against that

27. 5:28

    use case and the problems that it-- the, the challenges that that brings. Um, and that's a bit like when you start colonizing a planet in No Man's Sky [chuckles] as well.

28. 5:36

    So you sort of start to, to know what your actual problems are, uh, and, and have a more tangible problem sp- and more d- more scoped problem space to work with.

29. 5:48

    Uh, and that's, that's, uh, it's an opportunity and a challenge at the same time, of course. Um- The third thing I wanted to, uh, to talk about is, uh, I actually love that bit, and that's probably my favorite bit.

30. 6:00

    It's just how gratifying the product development process is because, you know, unlike this sort of sensation of herding lemmings that you get from having to, like, speak to 100 different departments and get 100 different sign-offs on your product design and your engineering design before, and your API design before you can go and launch the damn thing, here

31. 6:22

    you're sort of in, in the span of a few hours, if, if not a few days, uh, you'll go from, like, something happened, like MCP is out and becoming a standard.

32. 6:31

    You go from that to having an idea, a spec, um, a product spec, an eng design, um, uh, some testing, and a release, uh, in a very short period of time.

33. 6:41

    And it's just amazing to see, to see that stuff go from your head into, into code and into, and into the wild so quickly. Uh, it's exhilarating. Uh, it is worth noting, though, that, like, in particular in AI, velocity is like, uh, a staple stakes.

34. 6:56

    It's not a nice-to-have or, or something you pat on your back, on your back for. Uh, you're, you're out of the game if you can't, uh, if you can't release quickly and release well.

35. 7:07

    Um, and that's because some of the opportunities that allow you to stand out and, and sort of get those power-ups that accelerate you, uh, like, uh, good old Mario here, um, they come and go very quickly.

36. 7:19

    A- again, I always go back to MCP, uh, when it had a moment, but you can think of agent-to-agent protocol, you can think of diffusion models. All of these are moments that come up, and it's important that you're able to react and build products that, uh, that are, uh, at the forefront when it comes to these, to

37. 7:38

    these trends and to these breakthroughs, technology breakthroughs, uh, to basically continue to grow your brand and, and cement, cement yourself in the developer community or in the user community more generally, uh, whatever your target, uh, customer segment is.

38. 7:54

    Um, lastly, and this is-- my God, this is, like, the toughest part for me, uh, if I'm honest, um, but also where I'm learning the most, um,

39. 8:06

    i- it is something I, I have to admit I really underestimated how difficult that bit would be, and that's, like, how little scaffolding you have around you when you're, when you're doing your own thing from scratch, uh, when it comes to sort of outreach, awareness, traffic, adoption, right?

40. 8:23

    And, uh, we're in the middle of, like, full-on hype and meme wars in the AI space, uh, and it's hard to tell signal from noise. And so in that particular race, and without, uh, any of the sort of, uh, advantages of, of being a-attached to a bigger brand, uh, it feels like you're playing Crash Bandicoot, but without

41. 8:44

    any of the boosters. Or if you're more familiar with Mario Kart, you're playing Mario Kart, but you don't have any of those, uh, red and green turtles to fire at your friends, so you can beat them.

42. 8:54

    Uh, [chuckles] so, uh, this for me has been, like, the steepest learning curve. Um, a-and so one of those is, for example, that, that, that you, that you learn is that people follow people primarily, right?

43. 9:06

    And, and then, um, if you were launching a product at Stripe, all it takes is a tweak-- a tweet from one of the Collisons or one of their seconds in command, and, uh, you know, people know about your product and, and start reaching out and, and you're, you're off to the races, as they say.

44. 9:20

    Uh, whereas in this case, you have to find advocates, you have to find credible voices in the industry to speak for you, uh, and to use their platform to talk about you and act as, as force multipliers.

45. 9:31

    Um, the other thing is in terms of just brand awareness and traffic, um, you don't have the advantage of, of the br- uh, like I said, this sort of brand association and, and the, the website and the other, uh, channels of, uh, of a broader company, right?

46. 9:47

    And so the way you, uh, the way you deal with that is you find allies, and those will be either startups at the same stage or startups that are slightly ahead but that are interested in the space you're working on and where they see synergy.

47. 9:59

    You find ways of partnering, cross-marketing, cross-selling. Even things like getting your docs to show up in the, um, getting your product to show up in the integration docs of some company like Browserbase, for example, uh, i- is a significant channel to get people to then land on your properties, your website, your GitHub repo, and, and, uh, start

48. 10:21

    discovering your product. Um, and so I, you know, we're getting pretty beat up there. At least I'm getting pretty beat up on, on that front and, and, uh, uh, and even some cases not, not doing as well as I would like, but, uh, also learning a ton.

49. 10:36

    And so it's, it's part of why I put myself out there. It's to sort of, uh, really stretch myself. And so, uh, I'm sure we'll figure it out, but it's been, it's definitely been a challenge and a massive learning curve.

50. 10:48

    Uh, so I wanted to call that out. Uh, so there you have it. I won't take much longer of, uh, of your time. If you made it this far, I really appreciate you listening.

51. 10:58

    I hope that you find this useful. I would love to hear from you, uh, not just in the comments down below, but if you wanna reach out on LinkedIn, have a chat about your experience.

52. 11:07

    If you're trying to make that leap, and you think a conversation would be helpful, uh, you know where to find me, uh, on LinkedIn, I mean. Um,

53. 11:16

    and, uh, and look, I'm no ex-- I'm no expert. I'm still learning. The idea, uh, the purpose of this, of this, uh,

54. 11:24

    talk is to just, uh, share my experience and, and for it to maybe act as inspiration or get you thinking about what that looks like for you. Um, [clears throat] if you like the content of this video, the only thing that I ask, if you can find it in you, is to scan this QR code and land on our

55. 11:40

    GitHub repository. Uh, and, uh, if you would kindly give us a star, that goes, um, a, a very long way. Um, thank you for listening. Thank you for making it that far.

56. 11:52

    And, uh, wherever you are, I hope you're well, and I hope we one day meet in person. Thank you very much.