# Jason Cohen Built Two Unicorns. Here’s The Only AI Startup He’d Build in 2026

- Channel: [Rob Walling](https://www.youtube.com/channel/UCHoBKQDRkJcOY2BO47q5Ruw)
- Published: 2026-02-15
- Source video: [Watch on YouTube](https://youtube.com/watch?v=skmHKOjqWO8)
- Read on ReadTube: https://www.read.tube/p/videos/skmHKOjqWO8
- Language: Original (source language)

## Summary

**Jason Cohen's 2026 AI Plan**

Jason Cohen, who built two billion-dollar companies, advises that a successful AI startup in 2026 must solve existing customer problems by using AI to enable a dramatic 10x improvement, rather than selling AI as the product itself. He recommends building for expert users who can compensate for AI's current unreliability and creating an "absolutely amazing" product for a narrow customer segment to differentiate in an inevitably crowded market. This strategy prioritizes tangible value and user capability over simply applying AI.

Jason Cohen advises that while corporate budgets heavily favor AI, founders often misunderstand its role: AI is a solution enabling new possibilities for existing problems, not a problem people want to solve. A critical challenge is AI's current unreliability; it's frequently wrong, a factor new companies must account for.

He identifies three types of AI products:
- AI for incumbents: Adding AI to current products, often with limited usefulness.
- AI for experts/professionals: Aids skilled users who can correct AI's mistakes; preferred due to AI's unreliability.
- AI for "noobs" (non-experts): Empowers beginners but leaves them stuck when AI fails due to their lack of expertise.

For an AI startup, he'd solve a real, existing problem with budget, using AI to achieve a 3-10x improvement that was previously impossible. The target would be experts, who can mitigate AI's current flaws. He would focus on building an amazing product for a narrow, ideal customer, rather than worrying about inevitable competition using similar underlying tech.