Case study · Career, setbacks & entrepreneurship
Killing My Own AI Golf Coach Idea
I validated an AI golf coach before investing further and stopped when the market and cost model did not support a serious business.
The context
As a keen golfer, I saw an opportunity for an AI coach that could combine swing video and shot data with a virtual teaching professional, recreating the way a real coach supports a player on the course.
The challenge
I wanted to find out whether the idea had enough demand and commercial potential to justify building it into a business.
How I approached it
I built a prototype in Lovable featuring a live HeyGen avatar of a golf professional, including a cloned version of his voice. I interviewed golf professionals, surveyed golfers and set up a Wizard-of-Oz test through WhatsApp. AI ran in the background, allowing me to benchmark the accuracy of the coaching experience without distorting the demand signal.
During prototyping, I noticed that the product had quietly shifted from an asynchronous experience to a real-time service. That fundamentally changed the economics: instead of paying per analysis, the business would incur API costs for every minute of use. I incorporated that insight into my assessment rather than treating it as a problem to solve later.
What it delivered
I concluded that the Dutch market was too small to support the serious business I wanted to build, particularly given the real-time cost model. I stopped before investing in a full product, learning cheaply and preserving the option to pursue stronger opportunities. I still use a self-built version of the AI coaching app with my launch monitor at the driving range.
Looking back
The most important lesson was to test demand and economics together. The prototype, interviews, survey and Wizard-of-Oz test helped me learn before committing to a full product. I would pay even earlier attention to shifts in the experience: moving from asynchronous analysis to real-time coaching fundamentally changed the cost model and, with it, the commercial case.
I’m glad I stopped when the evidence showed that the Dutch market and per-minute API costs could not support the business I wanted to build. The principle I carry forward is simple: make key assumptions tangible, test them without distorting demand, and be willing to stop early while preserving room for stronger opportunities.