I shared a series of X posts in September 2026 that distilled my thinking on AI‑first product management, from decision‑making and metric‑first design to concrete checklists and AI‑agent reliability models.
The hardest part is often deciding what NOT to build.
A feature is not successful because users can use it. It’s successful when it changes a meaningful user behavior or outcome.
Metrics should be discussed before building, not after launching.
- ✅ Clear user story ✅ Acceptance criteria defined ✅ Designs/API requirements available ✅ Dependencies identified ✅ Edge cases discussed ✅ QA expectations clear ✅ Story small enough for the sprint ✅ Team understands what “Done” means
Prompt - what I want Context - what it needs to know Tools - what it can access and do Constraints - what it should and shouldn't do
I also codified a productivity workflow: Context → Goal → Constraints → Output format, which I applied to meeting transcripts, email drafting, and technical summarisation.