Skip to content

Rules of engagement

What to delegate and what to keep. When to trust and when to challenge. When to let it run and when to step in.

Whether working with AI leaves you more capable or quietly does your practice for you is decided by three calls you make every time you work with it. Capably Human calls them the rules of engagement. Each is tested by one question: does the help support your effort, or substitute for it? They are published as a one-page check, a model and a guide, and Humably coaches you by them on your own work.

The three rules

  1. Delegate or Retain

    Which parts of the work the AI does, and which you do.

    Keep the thinking you are responsible for, at least the first attempt in your own words; hand off assembly, formatting and retrieval without guilt.

  2. Trust or Challenge

    Which parts of the output you take as given, and which you check.

    Check the claim that would change your decision first; name the assumption; ask what you would ask a confident junior colleague.

  3. Let it run or Step in

    When the AI acts on its own, and when you take over.

    Step in before the point of no return; make sure you would see it go off course in time; keep the rep for anything you want to stay able to do.

The one test behind every call

Does this help support my effort, or substitute for it? Both are legitimate. Only one builds capability. The worst case is substitution that believes it is support.
The rules cut both ways: handing more to AI where it helps is a good call. The aim is deliberate, never less.

The stack

Humably

Practised on your own work.

Humably coaches you by the rules on the decisions and work already in front of you: your view first, the three calls made explicit, the record over time. The check is free in the app and on this page.

In the books

Heavy Lifting introduces the rules through a junior analyst’s Thursday afternoon (Chapter 14) and keeps the version for beside the keyboard in Appendix F. Load-Bearing carries them to the teams who design AI products.

The rules are working hypotheses. Capably Human’s trial program tests them on Humably Alpha data and publishes what it finds as Reports, whichever way the results come out.