"Human in the loop" is failing in two opposite directions at once. Novices over-trust AI outputs. Experts are skeptical but get overruled by leaders who prioritize outcome urgency over precision.

The issue isn't that companies lack a human checkpoint. It's that "human in the loop" has become a phrase that lets leaders feel covered without building the infrastructure that actually catches errors.

Three big ideas to ponder:

The moral crumple zone — one person held accountable for outputs they can't actually verify.

The expertise paradox — novices trust the AI more than the experts who know where it breaks, so the reviewer least equipped to catch failure gets nominal responsibility for catching it.

The ownership gap — leaders blame the expert in the loop when the system was never built to catch failure in the first place.

The fix isn't more humans. It's process evals and real feedback loops for error correction, paired with clear ownership and enablement of who's accountable for what.

If your team's AI governance plan is "we'll put a human in the loop" - watch this before you ship the next release.

h/t to all our thought partners (and idea friends) on this topic Steve Smith (here) Stacia Garr (here) Madeleine Clare Elish (here), PhD Kieran Snyder (here) Tom Fishburne (here) John Sumser (here) Dave Ulrich (idea friends) Master Burnett (here)

____

Thanks for reading, listening, sharing, and helping us think better. We love hearing where you agree, where you disagree, and what you're seeing from your seat.

If someone forwarded this to you, join the Insight Loop to get the next issue directly in your inbox.

And if you're already in the Loop, help us grow: forward this to someone who would enjoy it, and follow/subscribe to The Meg & Amy Show wherever you listen or watch.

Have a question you'd like us to tackle in a Leadership Corner? Send it to: [email protected]