Ethical AI Leadership Decision Toolkit

Four pillars

1. 01

   Human moral agency

   The person remains the one who can be asked why. Systems execute; leaders answer.

2. 02

   Psychological safety

   Status, certainty, autonomy, relatedness, and fairness are the first things a new system spends.

3. 03

   Fairness beyond data

   Opportunity, voice, and social signaling matter as much as the confusion matrix.

4. 04

   Trust as capital

   Trust is deposited, withdrawn, and called. Treat it like a balance sheet, not a vibe.

Five non-negotiable questions

Ask them before the dashboard gets to sound like a colleague. Ask them in the room, out loud, with a name attached to the answer.

1. 01

   Who remains accountable when the system acts?

2. 02

   Whose status, certainty, autonomy, relatedness, or fairness is being spent?

3. 03

   What would we owe a person this system affects that we would never owe a model?

4. 04

   How will we know if trust is being withdrawn — this week, not next quarter?

5. 05

   What human overhead are we not counting, and who is paying it?

Field sheets

One-page PDFs you can take into a meeting. They come from the same instruments as the full Toolkit — Risk Radar, fairness audit, trust triage, health check, stakeholder map.

• 01 · Boardroom pause

  Five non-negotiable questions

  Ask them out loud, with a name on the answer, before a system starts to sound like a colleague.

  Before a pilot, a vendor demo, or a go-live.

  Download PDFThe rest of this is in the complete file

• 02 · Pre-deployment

  Agentic AI Risk Radar

  Nine checks on autonomy, psychological safety, and operational ethics before an agent is allowed to act.

  Before you give an agent spend, mail, or hiring reach.

  Download PDFThe rest of this is in the complete file

• 03 · Beyond the spreadsheet

  Bias and fairness audit

  Source, signal, and outcome. Catch social signaling and opportunity hoarding, not only dirty data.

  Before launch, then at 30 and 90 days.

  Download PDFThe rest of this is in the complete file

• 04 · Triage

  Trust Withdrawal Matrix

  Plot attachment against status threat. The cell tells you whether to monitor, reassure, calibrate, or pause.

  When trust is already moving and you need a next step this week.

  Download PDFThe rest of this is in the complete file

• 05 · Monthly pulse

  Human–AI health check

  Four scores on dependency, agency, reality testing, and emotional distance. Under 12, retrain on autonomy.

  Once a month with the team that lives beside the agent.

  Download PDFThe rest of this is in the complete file

• 06 · People map

  Stakeholder impact map

  Anxiety versus reliance. Threatened, ambivalent, distant, or attached — each cell has a different ritual.

  When you are naming who will pay the human overhead.

  Download PDFThe rest of this is in the complete file