AI Work Diary

The diary isn't busywork or insurance theatre. It serves three real purposes: protection against future contract or dispute scenarios where the AI question gets asked, internal data on where AI actually saves time versus generates rework, and a credible answer to clients who ask "how did you make this?" The format is deliberately fast — a 2-minute entry per deliverable, not a 20-minute essay. If it takes longer, the format is wrong and people will stop using it within a week.

The TRACE framework structures each entry. Task definition (the human strategic thinking that preceded AI involvement — who set the brief, what context shaped the approach), Raw AI contribution (what tool, what prompt, what was produced), Adjustments (what was added, removed, rewritten, what errors were caught), Checks (factual, brand, client-fit, peer review), End product assessment (an honest percentage estimate and a specific note on what value the human layer added). The percentage estimate is meant to be honest — over-claiming human contribution is exactly the thing that looks bad in a future dispute, so the skill flags entries where the maths doesn't add up.

A five-tier classification covers the realistic spectrum: Human-Created (no AI), AI-Informed (AI for research or ideation, deliverable fully human-written), AI-Assisted (AI generated initial draft, 50%+ revised by human), AI-Produced Human-Reviewed (AI generated most output, light editing under 50%), and AI-Automated (minimal human intervention — automated reports, transcriptions). The taxonomy matters because contracts and policies increasingly hinge on which tier the work falls into, and "we use AI sometimes" doesn't survive a procurement review.

A simplified team template keeps the format consistent across multiple operators, and the monthly aggregation view surfaces patterns the individual entries don't — where AI is genuinely saving hours, where rework eats the savings, which client work is most AI-leveraged. Rapid-turnaround work gets a 30-second micro-log (date, client, deliverable, classification, one-line note) because the goal is sustained habit, not perfect documentation of every keystroke.