01
Problem and crux
GetFound was testing reverse recruitment, where employers approach pre-vetted candidates. A full product would have been expensive, while several basic assumptions were still uncertain.
02
Assumptions to test
Candidates needed to want the model and complete a demanding onboarding process. Employers needed to trust the profiles. The service then needed to produce actual placements.
03
What I did
Initial paid acquisition produced less than 1% click through and about 1.7% downstream conversion, so I changed the approach. I built a lightweight no-code candidate experience and kept parts of the service manual. This let us test the full causal chain before committing engineering capacity.
04
Evidence and result
The redesigned funnel produced hundreds of vetted candidates and the first successful placements. The evidence also contributed to a later CHF 2.2 million seed round.
05
Relevance to AI safety operations
New safety interventions often begin with substantial uncertainty. I know how to isolate a decision relevant assumption, build a cheap test and use the result to decide whether the work deserves more resources.