AI product practice
Read before you write
AI can accelerate reading and reasoning. It cannot replace source evidence, customer judgment, or accountable product decisions.
Operating loop
Before asking AI to draft anything, I give it the material that should constrain the answer: research notes, requirements, technical documentation, decisions, and known evidence boundaries. Then I separate facts, inferences, and open questions. That makes uncertainty visible instead of letting polished language hide it.
Next, I use AI to compare options, surface dependencies, challenge assumptions, and turn rough information into testable hypotheses. It helps identify what deserves a closer read.
Evidence gates
AI output is not customer discovery, technical validation, or proof that a recommendation is right. I verify consequential claims against original sources, protect confidential material, and review outputs before they shape a roadmap or reach a customer.
Closing principle
A human remains accountable for priorities, tradeoffs, and published words. AI helps compress the work around those decisions; it does not own them.