AI steering committee: who sits on it, what it decides, and how to keep it useful
How to set up an AI steering committee that funds, stops, and scales AI work, with members, decision rights, and a monthly agenda.
BY STEFAN MARITZ··7 MIN READ
TL;DR
- An AI steering committee is a small group of senior leaders who decide which AI work gets funded, stopped, or scaled, and who owns the risk. Its value comes from the decisions it makes, so a committee that only receives updates is a reporting line.
- Keep it to five to seven people: an executive chair, the leaders of the functions where AI is being used, someone who owns data and IT, and someone who owns risk, legal, or compliance.
- Give it explicit decision rights: approve or stop use cases, set the risk appetite, own the AI policy, and allocate budget across functions.
- Meet monthly with a standing agenda built around decisions, and review whether any workflow has changed since the last meeting.
- It works alongside a head of AI or an AI centre of excellence. The committee decides; the AI lead or centre does the work.
When AI starts appearing in several parts of a business at once, someone suggests a steering committee. It is a sensible instinct. The risk is that it becomes another monthly meeting where people present slides to each other and nothing is decided.
This post covers what an AI steering committee is for, how it differs from a centre of excellence or a head of AI, who should sit on it, what it should decide, and how to tell when it has drifted into theatre.
What is an AI steering committee?
An AI steering committee is the senior group that makes the cross-functional decisions about AI: which initiatives proceed, which stop, how much risk the organisation will accept, and where budget goes. It exists because those decisions cut across functions, and no single functional leader can make them alone.
It is the governing body in your AI governance framework. In the language of the NIST AI Risk Management Framework's generative AI profile, which groups the work into govern, map, measure, and manage, the steering committee sits squarely in "govern".
AI steering committee vs centre of excellence vs head of AI
These three get confused often, and many organisations need two of them. They do different jobs.
| Role | What it does | Who is in it | Typical rhythm |
|---|---|---|---|
| AI steering committee | Decides priorities, risk appetite, policy, and budget | Five to seven senior leaders | Monthly |
| AI centre of excellence | Builds shared capability, standards, and reusable tools | Specialists drawn from IT, data, and the business | Continuous |
| Head of AI or outsourced Chief AI Officer | Leads the agenda day to day and brings decisions to the committee | One accountable senior person | Continuous |
A mid-sized organisation usually needs the committee and an accountable lead before it needs a full AI centre of excellence. Whether that lead is a head of AI on the payroll or an outsourced Chief AI Officer depends on how much AI work there is to lead.
Who should sit on an AI steering committee?
Five to seven people, each of whom can commit their part of the business. More than seven and the meeting turns into a briefing.
- Chair: the chief executive, managing director, or chief operating officer. The chair needs authority to stop things.
- Functional leaders: the heads of the two or three functions where AI is being used most, such as operations, HR, or marketing.
- Data and IT: the person who can say whether the systems will support a use case.
- Risk, legal, or compliance: the person who owns data protection and regulatory exposure.
- The AI lead: the head of AI or outsourced Chief AI Officer, who prepares the decisions and runs the work between meetings.
Leave out anyone attending only to stay informed. Send them the minutes.
What should an AI steering committee decide?
Write the decision rights down. A committee without them defaults to receiving updates.
- Use case approval. Which AI initiatives start, continue, scale, or stop, based on agreed criteria.
- Risk appetite. Which tiers of AI risk the organisation accepts, and which uses need extra review.
- Policy. Ownership of the AI usage policy and any changes to it.
- Budget allocation. How AI spend is split across functions and initiatives.
- Vendor and platform choices. Decisions that commit the organisation beyond a single team, because those compound quickly and are expensive to unwind.
- Literacy and adoption. Whether people have the skills to use AI safely. For organisations operating in Europe, AI literacy obligations under the EU AI Act have applied since 2 February 2025.
If you want a recognised structure for these responsibilities, ISO/IEC 42001 sets out requirements for an AI management system that a steering committee can anchor.
How to run an AI steering committee
Run it like a capital committee in a manufacturing business. Proposals arrive in a standard format with the numbers attached, the committee decides, and the decision is recorded with an owner and a date.
- Meet monthly for no more than 90 minutes. Quarterly is too slow for the pace of AI decisions, and weekly turns it into operations.
- Circulate decision papers in advance. One page per decision: the proposal, the value, the risk tier, the cost, and the recommendation.
- Open with changed work. Spend the first 10 minutes on which workflows have actually changed since the last meeting. This keeps attention on adoption, where most AI programmes stall.
- Take the decisions. Approve, reject, or send back with a specific question. Record each one.
- Review one stopped or struggling initiative. Ask what was learnt and whether the lesson applies elsewhere.
How do you know the committee has become theatre?
Watch for these signs: months pass without anything being stopped, decks grow longer than the decisions they support, attendance drifts to delegates, and the AI portfolio grows while the share of organisations reporting that AI has added to EBIT has barely moved across the market. A committee that has never stopped anything is almost certainly not steering.
The fix is usually to cut the membership, rewrite the decision rights, and make "what has changed in the work" the first agenda item. An AI operating model that spells out who decides what, and where the committee fits, helps keep it honest.
Frequently asked questions
What is an AI steering committee?
An AI steering committee is a small group of senior leaders who decide which AI initiatives proceed or stop, set the organisation's risk appetite for AI, own the AI policy, and allocate budget. It governs; an AI lead or centre of excellence does the day-to-day work.
Who should be on an AI steering committee?
An executive chair, the leaders of the functions using AI most, the owner of data and IT, the owner of risk or compliance, and the AI lead. Five to seven people is usually right.
How often should an AI steering committee meet?
Monthly works for most organisations. It is frequent enough to keep pace with AI decisions and infrequent enough to stay focused on decisions rather than operations.
What is the difference between an AI steering committee and an AI centre of excellence?
The steering committee makes decisions about priorities, risk, policy, and budget. The centre of excellence builds shared capability, standards, and tools that teams use to deliver AI work.
Does a mid-sized company need an AI steering committee?
If AI is being used in more than one function, yes. It can be light: five people, one hour a month, and clear decision rights.
If you need someone senior to prepare the decisions and lead the agenda between meetings, that is the role our outsourced Chief AI Officer service fills.
WORK WITH PRAXES
Ready to move from reading about AI to getting your organisation ready for it? Start with the diagnostic.
Take the AI Readiness Diagnostic