Selected work
Evidence from execution
Each case follows the same structure: the decision relevant problem, the assumptions I tested, what I did, the evidence and its relevance to AI safety operations.
Skeepers / Octoly: Build a repeatable market entry system
Tested the model in Spain, built the local operation and codified the work into a playbook.
Reused for expansion into Italy and Germany.
Read case study Resource allocationSmily: Move capacity to higher value work
Made the case to close a small product whose maintenance burden exceeded its contribution.
Four staff redeployed and parallel maintenance removed.
Read case study Programme designCodeOp: Design training around outcomes
Used participant feedback to replace theory heavy teaching with applied cases.
Around 80% entered relevant roles within two to six months.
Read case study Team leverageQwist: Increase capacity through people
Mapped capability gaps and gave colleagues progressively larger ownership.
Two career transitions and a 40% rise in team satisfaction.
Read case study Cheap testsGetFound: Test the model before scaling
Built a lightweight service to test the weakest assumptions before committing engineering time.
Over 70% onboarding conversion and the first placements.
Read case study AI deploymentSmily AI Inbox: Design around human control
Changed the product from automatic replies to AI drafts after user research exposed a trust constraint.
Used by more than 50% of active accounts in its first month.
Read case study