From: aidotengineer
Munir, CEO and co-founder of Porsche AI, shares insights from his journey as a recent founder of an AI startup, drawing on his previous product roles at big tech companies like Stripe, Google, and Amazon [00:00:03]. Porsche AI, co-founded with Emma, is an open-source SDK for building production agents with a focus on regulated industries [00:01:38].
User Problems and Discovery in AI Startups
User problem identification and discovery in AI is significantly different from traditional tech [00:02:14].
- Undefined Problems Conventional user problems often have a discernible shape and adjacency to known issues [00:02:31]. However, in AI, users frequently don’t know what their problems are; instead, user problems emerge as a property of the space [00:03:25]. This is akin to starting a “Soulslike game” where “nothing makes sense at first” [00:02:59].
- Rapid Change The capabilities of the technology are evolving extremely quickly [00:03:41]. This means traditional three-to-six-month roadmaps are not effective [00:04:49]. Instead, companies must make directional bets and iterate very rapidly by releasing products and observing user reactions [00:03:51].
- Hypothesis-Driven Iteration The anchor for focus becomes hypothesis-driven, deliberate iteration, rather than conventional roadmaps [00:04:06].
- Creating Narratives Unlike traditional user storyboards and critical user journeys, these tools often feel obsolete in the AI space [00:04:32]. It’s more like “No Man’s Sky,” where you have to create your own narrative [00:04:41]. Founders must help users overcome initial inertia, work with them to create narratives, and anchor on specific use cases [00:05:00]. This process eventually leads to more familiar territory where product refinement can occur against tangible problems [00:05:19].
Velocity as a Table Stakes
The product development process in a small AI startup can be incredibly gratifying [00:06:01].
- Rapid Development Cycle Unlike large organizations requiring numerous sign-offs, a small team can go from an idea to a released product—including spec, design, testing, and release—within a few hours or days [00:06:21].
- Critical Importance of Velocity In AI, velocity is not merely a “nice to have”; it is “table stakes” [00:06:54]. The inability to release quickly and effectively means being “out of the game” [00:07:02].
- Seizing Opportunities Opportunities that allow a company to stand out and accelerate its growth appear and disappear very quickly [00:07:07]. Examples include new standards like MCP, agent-to-agent protocols, or diffusion models [00:07:20]. It’s crucial to react swiftly and build products at the forefront of these technological breakthroughs to grow brand and cement presence in the developer or user community [00:07:31].
Outreach, Awareness, and Adoption
One of the toughest aspects of starting an AI company is the lack of scaffolding for outreach, awareness, traffic, and adoption [00:08:11].
- Navigating Hype and Noise The AI space is currently in “full-on hype and meme wars,” making it difficult to discern signal from noise [00:08:25].
- Lack of Brand Advantage Without the advantages of a larger brand, it feels like playing “Crash Bandicoot” without boosters or “Mario Kart” without items [00:08:39].
- Building Advocacy Unlike launching a product at a major company where a single tweet from leadership can generate awareness, startups must find advocates and credible industry voices to speak on their behalf and act as force multipliers [00:09:09]. People primarily follow people [00:09:04].
- Strategic Partnerships To build brand awareness and drive traffic without a large company’s established channels, finding allies is crucial [00:09:33]. These allies can be other startups at a similar stage or slightly ahead, interested in the same space with synergy [00:09:51]. Partnering, cross-marketing, and cross-selling are vital [00:09:58]. Even having a product listed in the integration documentation of another company (e.g., BrowserBase) can be a significant channel to drive users to your website or GitHub repository [00:10:08].