What Changed My Mind About Niche Scoring for AI SaaS

I used to score niches by what the model could do. The capability was the niche. Two failed launches later, I changed my mind.

What Did Not Change My Mind

Better prompts. New model releases. Faster inference. None of it. The capability was always the wrong starting point.

What Did

The third candidate scored highest on builder-fit and reachability and lowest on capability fanciness. I almost skipped it because it was boring. I built it instead.

It worked in three months. Customers paid. The two earlier capability-led niches still have not.

The Realization

Capability is a multiplier on top of a real niche. It is not the niche. The niche is the audience, the pain, and the spend. AI is how you serve them faster or cheaper.

What I Do Differently Now

• Score builder-fit by audience access, not technical interest.
• Score crowding before scoring capability.
• Run an outside reviewer before committing.

What This Means for You

If you are scoring niches by what the model can do, you are scoring the wrong axis. The capability is the easy part. Reaching the customer is not.