# From PM at Stripe to Building an AI Startup: A Recent Founder's Journey

Mounir Mouawad, Porsche AI | AI Engineer World's Fair 2025 | 11:59

Source: https://www.youtube.com/watch?v=HIGpxVjGFBw
Channel: AI Engineer (https://www.youtube.com/@aiDotEngineer). Summarised by AIE Talks.
Page: https://aietalks.com/talks/from-pm-at-stripe-to-building-an-ai-startup-a-recent-founders-journey
Published: 2025-06-03
Tags: engineering-culture, product-strategy, startups

## TL;DR
- AI startup users often do not know their problems yet, so founders have to form hypotheses, release products, and learn from reactions.
- AI product development can move from an idea to code and release within hours or days, while fast, well-timed releases are required to catch short-lived opportunities.
- A new founder loses the reach and credibility of a large company and must build awareness through advocates, partnerships, cross-marketing, and integration channels.

## Summary
Mounir Mouawad describes the move from product roles at Amazon, Google, and Stripe to co-founding Porsche AI with Emma. The company builds an open source SDK for production agents in regulated industries, but the talk focuses on what changed when he began building independently. AI users often know they want to use AI without knowing the problem they want to solve. Mouawad says founders must create use-case narratives with users, make directional bets, and iterate through deliberate, hypothesis-driven releases. Product development is faster than at a large company, yet that speed is required because opportunities around technologies such as MCP can disappear quickly. The hardest change has been distribution. A startup has no established brand, audience, or senior executive who can make a product known with one post. Mouawad is candid that outreach and adoption have been difficult, and that he is still learning how to find advocates and partner with other startups.

## Key ideas
### AI users often discover their problems while using the product
[01:56](https://www.youtube.com/watch?v=HIGpxVjGFBw&t=116s)
Mouawad says conventional product work usually starts with a problem that has a recognizable shape. At Stripe, for example, a customer might want card issuance connected to an existing Stripe balance. AI startups often begin with users who know they want to use AI but cannot yet define the problem. The possible applications keep changing as the technology changes, so the problem itself emerges through experimentation. He compares this to starting a difficult Soulslike game without knowing how to configure the character or combine its gear. The founder has to help users find a concrete use case before ordinary product iteration can begin.

### Hypothesis-driven iteration replaces the comfort of a fixed roadmap
[03:27](https://www.youtube.com/watch?v=HIGpxVjGFBw&t=207s)
Because the problem space is still forming, Mouawad says a three-month or six-month roadmap is less useful than directional bets followed by fast iteration. The team needs to put products in front of users and study their reactions. He calls hypothesis-driven, deliberate iteration the anchor for the founder and team. Familiar tools such as user storyboards, critical user journeys, and clearly defined pain points do not disappear entirely, but they provide much less structure at the start. The work is to create enough focus to learn which use case deserves further product development.

### Founders have to help users create a narrative around AI
[04:35](https://www.youtube.com/watch?v=HIGpxVjGFBw&t=275s)
Mouawad compares early AI adoption to starting No Man's Sky without a clear mission or story arc. Users can be interested in AI while having no idea what they should do next. The founder therefore has to reduce the inertia around a first experiment and work with users to establish a narrative for a specific vertical or use case. Once a use case exists, the company moves into more familiar territory. It can refine the product against concrete challenges and gradually narrow the problem space.

### Small teams can turn new technical developments into releases very quickly
[05:53](https://www.youtube.com/watch?v=HIGpxVjGFBw&t=353s)
Mouawad says the product process is unusually satisfying because an idea can move from a new development to a specification, design, testing, and release in hours or days. He uses MCP becoming a standard as an example of a trigger for this sequence. At a large company, the same work may require many departments and approvals before launch. At a startup, the connection between the founder's thought and code in the hands of users is much shorter. He describes that speed as exhilarating, especially when the team can act on a new technical opening.

### Speed is required because AI opportunities can disappear quickly
[06:55](https://www.youtube.com/watch?v=HIGpxVjGFBw&t=415s)
Mouawad argues that velocity in AI is a basic requirement rather than something a team can treat as an optional advantage. Technical moments such as MCP, agent-to-agent protocols, and diffusion models can create openings that appear and fade quickly. A startup needs to react while a development is still attracting attention and build something that reaches the front of the conversation. He connects this to growing a brand within a developer or user community. Releasing quickly matters because a delayed product may arrive after the relevant moment has passed.

### Distribution is harder when the company has no larger brand around it
[07:59](https://www.youtube.com/watch?v=HIGpxVjGFBw&t=479s)
The hardest part for Mouawad has been building awareness, traffic, and adoption without the scaffolding of a large company. AI hype and memes make it difficult to separate useful attention from noise. At Stripe, a tweet from a senior leader could make people aware of a product and prompt inbound interest. A new startup has to find credible advocates who will speak about it and lend their own reach. Mouawad compares the situation to playing a racing game without its boosters. He is direct that this area has been a steep learning curve and that Porsche AI has not always performed as well as he wanted.

### Partnerships can replace some of the reach a startup lacks
[09:19](https://www.youtube.com/watch?v=HIGpxVjGFBw&t=559s)
Mouawad says a young company cannot rely on the website, brand association, and distribution channels that come with a larger employer. One response is to find allies among startups at a similar stage or companies slightly further ahead that have a reason to work together. Possible arrangements include partnerships, cross-marketing, and cross-selling. He gives the example of having a product appear in the integration documentation of Browserbase. That placement can send developers to the startup's website or GitHub repository, where they can discover and try the product.

## Notable quotes
- "They sort of know that they want to use AI and they know that there's some opportunity there, but they don't know the user problems are almost like an emergent property of the space." (03:08)
- "Your anchor for you and the rest of your team is really just hypothesis-driven deliberate iteration." (03:47)
- "In the span of a few hours if not a few days, you'll go from something happened like MCP is out and becoming a standard to having an idea, a spec, a product spec, a design, some testing and release." (06:26)
- "People follow people primarily." (09:04)
- "I'm no expert. I'm still learning." (11:18)

## Tools & references mentioned
- Porsche AI
- Stripe
- Google
- Amazon
- Amazon Deals
- Google Pay
- MCP
- No Man's Sky
- Browserbase
- GitHub
- Mario Kart
- Crash Bandicoot
- Neo

## Who should watch
- You are considering leaving a large technology company to start an AI business and want a candid account of what changes after the move.
- You are building an AI product whose users have interest in the technology but cannot yet describe a concrete problem.
- You have a working product but little distribution, and need ideas for advocates, startup partnerships, and documentation channels.

## Related talks

- [The AI Pivot](https://aietalks.com/talks/the-ai-pivot) (Brittany Walker, CRV & Chris White, Prefect & Bryan Bischof, Hex, 35:16)
- [From Hype to Habit: How We're Building an AI-First SaaS Company](https://aietalks.com/talks/from-hype-to-habit-how-were-building-an-ai-first-saas-company) (Rossella Blatt Vital, Sprout Social, 19:29)
- [Shipping something to someone always wins](https://aietalks.com/talks/shipping-something-to-someone-always-wins) (Kenneth Auchenberg, AlleyCorp, 16:17)
- [Reverse-Engineering the AI Buyer](https://aietalks.com/talks/reverse-engineering-the-ai-buyer) (Aliisa Rosenthal, Acrew Capital, 19:10)
- [Your AI Product Will Fail Unless You Can Explain It](https://aietalks.com/talks/your-ai-product-will-fail-unless-you-can-explain-it) (Veronica Hylak, Hey AI, 06:02)