Below the Approval Line

Every large company is now trying to find the AI its employees use without approval, and the tools to do it have become cheap.

In this episode I take the manager's side first, because wanting to see what's happening is a reasonable reaction. Then I look at what the research says happens next. Being watched on its own changes little, but expecting to be judged makes work safer and less creative. Decades of studies on office monitoring can't find the performance gain the tools are bought for. And when a channel gets watched or banned, the work moves somewhere you can't see. So the number in the report goes down, and it means the opposite of what it seems to mean.

This episode builds on chapter 11 of Submerged, "Stop Hunting the Fleet". Maria, the quality engineer in the story, is a composite character from the book, not a real person.

Research mentioned:

- Bernstein (2012) on the transparency paradox

- Amabile and colleagues (1990) on evaluation and creativity

- Meta-analyses of electronic performance monitoring

- US regulators' off-channel communications fines (September 2022)

Both books are free:

Submerged: https://submerged.lerias.org

Leapfrog: https://leapfrog.lerias.org

The views in this episode are my own and not those of my employer.

© 2026 Hugo Lerias. Licensed under CC BY-NC-ND 4.0.

Most innovation programmes in large companies don't get cancelled. They get a sign-up form, then a status report, then a review meeting, and a year later nobody can tell you when they stopped.

In this episode I look at why good programmes die without anyone deciding to end them. The cost is easy to see on a budget sheet and the benefit isn't, and every sensible question asked to protect the programme slowly rebuilds the approval process inside it. I talk about Google's twenty per cent time, Ethan Bernstein's curtain in a phone factory, and the protected window, a simple mechanism from my book Submerged that replaces status reports with a time limit and a spending cap.

This episode builds on chapter 7 of Submerged, "The Graveyard of Good Programs". Maria, the quality engineer in the story, is a composite character from the book, not a real person.

Research mentioned:

- Cyert and March (1963) on organisational slack

- Criscuolo, Salter and ter Wal (2014) on bootlegging

- Bernstein (2012) on the transparency paradox

- Google's twenty per cent time (press and talks, not a study)

Both books are free:

Submerged: https://submerged.lerias.org

Leapfrog: https://leapfrog.lerias.org

The views in this episode are my own and not those of my employer.

© 2026 Hugo Lerias. Licensed under CC BY-NC-ND 4.0.