Deployment Labs
Module 1 · Lesson 3 of 5 · 14 min

The four promises

Why disclosure has to be safe before anything else can work.

Every step after this one depends on people telling you the truth about how they use AI: what they use, what for, and what went wrong. Right now, most of them have good reasons not to. The four promises change that, and they come before anything else in the framework.

90%

of companies have employees using personal AI tools for work. Only 40% pay for an official one.

MIT NANDA, 2025

48%

of desk workers are uncomfortable telling their manager they use AI.

Slack Workforce Index, 2024

29%

of employees admit to working against their company's AI strategy.

WRITER, 2026

Why people hide what they're doing

Some worry it looks like cheating, or laziness. More worry about what happens when they show how much faster they've become. When 69% of executives say they're planning AI-related layoffs (WRITER, 2026), keeping a shortcut to yourself is a rational choice.

The usual response is a ban on personal AI tools. It feels like control. It isn't: the use carries on, out of sight, along with whatever data is going into those tools.

Personal AI tools are bannedVisible to leaders5 of 19The dashed ones are still running. Leaders just can't see them.
Fig. 1.4Banned, not gone. A ban changes what leaders can see, not what's running. Illustrative.

The four promises

The promises are made before anyone is asked to disclose anything. Read them as a leadership team, out loud if you can. Your answers go onto your aim page for your sponsor to sign.

BuildMake the promises

The promises are a required minimum, and their wording is fixed. Your leadership team doesn't edit them; it decides whether it will keep them. Tick each one you will.

Who holds individual disclosures?

Someone outside the management chain.

These aren't new ideas

Aviation learned the first promise decades ago. Since 1976 the Aviation Safety Reporting System, run by NASA for the FAA, has given pilots and controllers limited protection when they report mistakes and near misses. More reports mean more problems caught before they become accidents.

Lean manufacturers learned the second. Improvement stops the moment people believe it will cost them their jobs. Wiremold's celebrated lean turnaround rested on an up-front promise that nobody would lose their job because of an improvement.

Test the promises

Promises are easy to make in a meeting. Here are two moments where they get tested. Choose what you'd do.

ScenarioA disclosure, three months late

A team member tells you they've been pasting client contracts into a personal AI tool to summarise them. It's been going on for three months. They came to you because of the new promises.

How much your people will tell you next quarter52%

What do you do?

Illustrative. The meter shows the direction things move, not a measured value.
ScenarioThe question from the board

Your operations team's AI work has freed about a third of their time. At the next board meeting, a director asks how many roles that saves.

How much your people will tell you next quarter70%

What do you do?

Illustrative.
The idea

You can't fix a bottleneck you can't see, and people won't show you what they're doing until it's safe.

CheckQuick check

Which part of the promises can your organisation change?

Finished this lesson?

Mark it complete. Your answers are already saved in this browser.