← Back to experience2025 · Antler
Entrepreneur in Residence
In 2025, I joined Antler to build my own startup and find a team. I worked through three ideas, each killed or parked for a commercial reason, not a technical one. The real return was a different way of working: look at a project commercially first, start fast, fail fast, build a POC instead of a full system, and work all sides of the business, not just the product. One of the three ideas I still believe is the number one idea, and I am building it now.
Venture ValidationPOCsAI/LLMComplianceCommerce
Selected projects
01
EU AI Act compliance management
Situation
Every EU company that uses AI now has obligations under the AI Act, and the mid-market has no tooling for them.
Task
Validate whether companies would buy a compliance management system for it.
Action
Researched the obligations, built the concept, and tested it with potential buyers.
Result
The need is real, but compliance bores most buyers, which makes it a timing and packaging problem rather than a bad idea. I remain convinced it is the number one idea of the three, and it is the one I am building now.
02
AI agent orchestrator for ecommerce
Situation
Ecommerce operations are full of tasks that fleets of AI agents could carry out if they were coordinated.
Task
Build the orchestrator and find out who buys it.
Action
Took the system far into development while testing the sales motion in parallel.
Result
The technology worked; the market did not: it only makes sense for enterprises, and enterprises were not ready to bet on running fleets of AI agents. Killed for commercial reasons, and the orchestration lessons carried straight into my later work.
Situation
Cross-border sellers struggle with VAT, which looks like a classic compliance software gap.
Task
Size the market and validate willingness to pay.
Action
Tested the idea with the buyers who felt the pain.
Result
Real appetite, too small a market: the EU's One Stop Shop regime shrank the target audience dramatically. Killed by market math, exactly the kind of call I would previously have learned after building for a year.