Google AI Essentials certified, Google AI Professional Certificate underway, with a daily working practice across Claude, ChatGPT, and Gemini. Prompting is the easy part. The job is catching the point where a confident answer and a correct one stop being the same thing.
Years of enterprise IT audit and risk consulting teach one habit above everything else: nothing gets accepted on the strength of how confident it sounds. A control gets tested against the actual policy. A finding gets checked against the actual evidence. That habit transfers directly to working with AI — a fluent answer isn't the same as a correct one, and the gap between them is exactly what needs catching before it reaches a workflow, a dashboard, or a client.
In practice, that means rewriting a prompt when the first answer is confidently wrong, cross-checking a model's output against source data before it goes into a report, and treating AI reasoning the same way an audit treats a client's claim — as something to verify, not a conclusion to accept.
Cross-checking AI-drafted analysis against source data — GBFS feeds, Metabase queries — before it reaches a dashboard.
Rewriting a prompt when the first output sounds right but doesn't hold up against the actual workflow.
Applying the same audit-grade skepticism to AI reasoning that a decade of compliance work applied to business claims.