Once several teams use AI, the same questions keep returning: which tools are approved, which projects go next, who checks the risks. A small steering group answers them in one place. These four steps set it up so it decides quickly, publishes what it decides and stays out of the way of day-to-day work.

Step 1 — Set the remit and members

Goal: Make clear what the group decides and who is in it.

  1. Write the remit: approved tools, priority workflows, policy changes and high-risk uses.
  2. Keep it small: an executive sponsor, an operations lead, a technology lead and someone covering legal or risk.
  3. Name a chair who prepares the agenda and records decisions.
  4. Agree what teams can do without asking the group.

You finish with: A one-page charter with remit, members and decision rights.

Step 2 — Create the intake

Goal: Give teams one simple way to bring proposals.

  1. Use a short form: the workflow, the data involved, the expected benefit and the owner.
  2. Sort proposals into low, medium and high risk using written criteria.
  3. Let low-risk uses within policy proceed without waiting for a meeting.
  4. Bring medium and high-risk proposals to the monthly meeting.

You finish with: An intake form and risk criteria.

Step 3 — Run the monthly meeting

Goal: Decide, record and move on.

  1. Review results from live workflows against their measures.
  2. Decide on new proposals: approve, approve with conditions, or decline with a reason.
  3. Update the approved tools list and policy if needed.
  4. Publish decisions to the company the same week.

You finish with: A decision log and a published monthly update.

Step 4 — Review the portfolio quarterly

Goal: Keep the overall programme focused and safe.

  1. List every live AI workflow with its owner and results.
  2. Stop or fix workflows that are not delivering or have new risks.
  3. Choose the next quarter’s priorities.
  4. Check that access, data rules and training still match how AI is used.

You finish with: A quarterly portfolio review and updated priorities.

Common pitfalls

  • A large committee that cannot decide.
  • Making every small use wait for a meeting.
  • Decisions that are never published.
  • No owner for each live workflow.

See it as a workflow

Leadership

A company-wide AI rollout with rules people follow

Move from scattered personal AI accounts to one approved setup, a short policy and a first set of shared workflows the whole company uses.

See the workflow

More guides like this

Related answers

Read as Markdown Fractional Chief AI Officer