A Plain Memo
If your team trusts AI more than they trust themselves — the talk won't fix it
If a junior on your team keeps handing in confident, well-written, AI-assisted work that's quietly wrong — and you keep running the values conversation about it — the conversation isn't broken. The diagnosis is.
The problem isn't that they're trying to get away with something. The problem is that AI is great at some tasks and terrible at others, the worker can't tell which is which, and the wrong answers come back sounding more persuasive than the right ones. Your job, as their manager, is to teach them where the edge is — and that isn't a discipline conversation. It's a calibration one.
ℹ️ AI voice disclosure: This episode uses an AI-generated voice (ElevenLabs). The host is a pseudonymous byline persona (Reid Conlin).
In this memo
Sources
- Dell'Acqua et al., "Navigating the Jagged Technological Frontier" — BCG / Harvard Business School, Organization Science 2025:
- Randazzo et al., "Persuasion Bombing: Why Validating AI Gets Harder the More You Question It" — HBS WP 26-021, 2025:
- Cash, Oppenheimer et al., "AI Chatbots Remain Overconfident — Even When They're Wrong" — Memory & Cognition, 2025:
- Bastani et al., "Generative AI Can Harm Learning" — PNAS, 2025:
- Lancet 2025 — AI-assisted endoscopy and unaided polyp-detection skill loss: )00133-5/fulltext
- Anthropic Economic Index, November 2025 — directive task delegation rising 27% to 39%:
- Pew Research, October 2025 — about 1 in 5 US workers now use AI on the job: