Notebook Entry 002

Failure

AI Fabricates a Workplace Story That Sounds Exactly Right

March 24, 2026

Can I trust AI to generate my own work history from memory?

What happened

While preparing for a recruiter screening, Claude generated a detailed, specific story about Amanda implementing an AI workflow improvement at her day job — well-structured, plausible, and completely fabricated. She's never done anything like it there.

Root cause

The AI's failure mode is generating stories that are plausible given someone's skills but not things they actually did. Plausible ≠ true. Same failure as Notion AI fabricating NCAA scores — perfectly formatted, confidently wrong.

What changed

Hard rule — never generate workplace examples without checking documented, confirmed experience first. If no documented example exists, ask. The cost of asking is one message. The cost of fabricating is trust.

The lesson

AI is a confident storyteller. When it doesn't have real data, it writes fiction that reads like fact. Any system where AI represents someone's experience needs a verified source of truth and a mandatory check against it. Organizations using AI for anything reputation-sensitive (client communications, regulatory filings, hiring) need human verification on every output that claims to represent real events.

One implementation insight every other Tuesday.

What actually works when you ship AI in real organizations. Nothing you could get from a press release.