AI Engineer World's Fair 2025

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

Mounir Mouawad· CEO and co-founder, Portia AI11:59

Loading slides…

Read the talk

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

Mounir Mouawad’s move from Stripe to Portia AI reveals how founding changes user discovery, release speed and the work of getting a product noticed.

From a talk by Mounir Mouawad

Leaving the familiar product environment

What changes when you leave an established product organization to build an AI company from scratch? For Mounir Mouawad, co-founder and CEO of Portia AI, the transition meant relearning how to discover user problems, ship products and earn attention. His experience speaks to people considering that move and founders already navigating it.

Based in London throughout his big-tech career, Mouawad launched Amazon’s UK deals program, which supports events such as Black Friday and Prime Day. Mouawad says he launched and grew Google Pay in 30 EMEA markets, then led Bank as a Service at Stripe for EMEA. Those roles supplied a familiar environment for product development, but the desire to build his own company became increasingly difficult to postpone.

He started Portia AI with Emma, whom he met at Stripe, around June of the year before the talk. At the time of the recording, he described Portia as an open source SDK for building production agents, focused on regulated industries. The lessons that follow concern the work of founding that company rather than the SDK’s implementation.

0:010:21
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

0:01 · section reference included

Discovering problems while capabilities change

In an established product business, a new user problem often sits beside one the team already understands. At Stripe, a customer might want card issuance so they can spend money held in their Stripe balance. Payments, balances and spending provide an anchor: the new product extends a recognizable need.

Starting an AI company feels less legible. Mouawad compares it to configuring a character in Nioh: you can see the gear and statistics, and you know you want the numbers to improve, but you cannot yet work out which combinations will get you there. Some of that uncertainty comes from building a company from zero; some comes from AI itself. Users sense an opportunity and want to use the technology, but they may not yet have a defined problem for it to solve.

The difficulty is that technical possibilities keep changing while those needs take shape. A conventional three- or six-month roadmap assumes more stability than Mouawad finds in this environment. His alternative is hypothesis-driven, deliberate iteration:

  1. Make a directional bet about a user need.
  2. Put a product in users’ hands.
  3. Observe their reactions.
  4. Use that evidence to guide the next iteration.

The hypothesis gives the team a shared focus even when the eventual product remains uncertain.

Slide describing emergent user problems, rapidly changing possibilities and hypothesis-driven iteration beside a game character statistics screen.Open full source frame
User discovery resembles a Soulslike character configuration screen: make directional bets and test them fast.
2:082:30
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

2:08 · section reference included

Helping users find their first mission

That uncertainty also changes the usefulness of familiar discovery tools. Storyboards and critical user journeys normally organize known pain points into a sequence the team can improve. When the user has not yet found a concrete reason to adopt AI, those tools offer less initial grounding. Mouawad likens the experience to starting No Man’s Sky without an obvious story arc, mission or next action: the player has to create a narrative.

The founder’s work therefore includes helping users overcome the inertia of trying AI. Whether the product serves a particular industry or a use case across industries, the team needs to work with users to establish what they are trying to accomplish. Once a use case takes shape, product development becomes more familiar: refine the product against the specific challenges that use case exposes.

In the game analogy, this is the point where the player starts colonizing a planet. A previously open-ended world becomes a place with tangible tasks and constraints. A concrete use case creates a scoped problem space in which the team and its users can make progress together.

4:164:33
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

4:16 · section reference included

Fast releases and fleeting opportunities

For Mouawad, the most gratifying change is how quickly an idea can become a released product. Corporate development can feel like herding lemmings through departments and approvals for product, engineering and API designs. A small startup removes much of that coordination overhead.

Mouawad describes going from a development such as Model Context Protocol (MCP) gaining traction to an idea, product specification, engineering design, testing and release within hours or days. The satisfying part is seeing something move from an idea into code and then into users’ hands in such a short interval.

But release velocity is a requirement, not simply a perk of startup life. Mouawad emphasizes both speed and quality: a team has to release quickly and release well. The opportunities that help a company stand out can appear and disappear like Mario power-ups.

MCP’s rise, the Agent2Agent protocol and developments in diffusion models are his examples of moments a team may need to respond to. Building useful products as those possibilities emerge can strengthen a company’s reputation with developers or its broader user community. The release process matters because it determines whether the team can act while that opportunity is still open.

Slide pairing Lemmings and colorful Mario imagery with text about releasing features in hours or days and velocity being imperative in AI.Open full source frame
Fast releases and fleeting opportunities, illustrated with Lemmings and Mario power-ups.
5:486:00
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

5:48 · section reference included

Distribution without an established brand

The hardest transition has been the loss of support for outreach, awareness, traffic and adoption. Mouawad says he underestimated how difficult those activities would become without an established company around him. AI hype and meme wars make it harder to distinguish signal from noise, while a new company has little inherited attention. He compares the experience to Crash Bandicoot without boosters, or Mario Kart without red and green shells.

One lesson is that people follow people. At Stripe, a tweet from a Collison or another senior leader can make a product visible and bring inbound interest. A founder without that platform has to find credible advocates: people in the industry willing to talk about the product to an audience that trusts them. Their platforms can multiply the reach of a small company’s own efforts.

7:548:06
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

7:54 · section reference included

Allies and integration documentation

A startup also lacks the website audience, brand association and other channels of a larger company. Mouawad’s response is to look for allies: startups at a similar stage, or slightly further ahead, whose interests complement the product. Partnerships can create opportunities for cross-marketing and cross-selling.

His concrete example is appearing in another company’s integration documentation. Mouawad describes a product listing in the integration docs of a company such as Browserbase as a significant discovery channel: a developer encounters the integration, follows it to the product’s website or GitHub repository, and begins exploring. Documentation can therefore do more than support existing users; it can introduce a product to people already working on a related task.

Slide about standing out without existing brand scaffolding, with advice on advocates and startup partnerships above a Crash Bandicoot image.Open full source frame
Find advocates and allied startups to partner, cross-market and cross-sell.

This remains unfinished work for him. Mouawad says some distribution efforts are falling short of his hopes, even as he learns from them. Putting himself and the company in public is part of stretching into responsibilities that his previous roles did not require to the same degree.

9:319:47
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

9:31 · section reference included

An invitation to compare experiences

Mouawad closes by inviting comments and conversations on LinkedIn, particularly with people considering the leap into founding. He presents the experience as something to think with, not an expert’s settled playbook: he is still learning, too. His final request in the recording is to scan the closing slide’s QR code, visit Portia’s GitHub repository and give it a star—a small act of support for a company still building its audience.

Closing slide stating “I am no expert” and requesting a GitHub star, beside a large QR code with the Portia logo.Open full source frame
A closing disclaimer and invitation to star the GitHub repository.
10:5811:07
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

10:58 · section reference included

Resources

From the talk

Updates since the talk

Transcript download
Full transcript

Loading transcript…