Use this whenPeople are building their own AI tools, and you need a way to prioritize and sort them without slowing them down.
Why it exists
Once subject matter experts build their own tools, tools multiply. Some overlap and some contradict each other. A council can manage that, but a council that approves everything becomes the thing that shuts building down. The charter is what keeps it useful.
The framework
- Step 1Who sits on itWho needs to be in the room?
- Step 2What it decidesWhich calls are the council’s?
- Step 3How it runsWhat is the rhythm?
- Step 4What keeps it openHow does it avoid becoming a gate?
Who sits on it
Who needs to be in the room?
Business leaders from the functions doing the work, the AI lead, someone for data and security, and a champion voice. The people who build should be heard by the people who decide.
What it decides
Which calls are the council’s?
Priorities, which level a tool sits at, which duplicates to merge, which tools to retire, and when teams should reorganize around new skills.
How it runs
What is the rhythm?
A regular cadence, a review of the tool registry each time, and decisions written down with a reason.
What keeps it open
How does it avoid becoming a gate?
Personal tools need no approval. The default answer is yes, with conditions. The council commits to a response time.
How to use it
- Form the council once building takes off, and not before.
- Seat leaders from the functions doing the work, the AI lead, data and security, and at least one champion.
- Agree on what it decides, and what it never needs to see.
- Meet on a fixed cadence, work from the tool registry, and write every decision down.
- Publish the rules that keep it open, and check them against how it is actually behaving.
Rules that keep it from becoming a gate
- Personal tools need no approval, and the council never asks to review them.
- The default answer is yes with conditions, such as an owner and a registry entry.
- The council commits to a response time, so a request never sits waiting.
- Every no comes with a reason and a way to get to yes.
- Decisions are written down, so the same question isn’t asked twice.
What it looks like
Picture a team that has built a tool for its own work, and two other teams have started using it. At the council’s regular review, the registry shows it has quietly become shared. The council says yes with conditions: name an owner, record who relies on it, and agree on a review date. It also notes that a similar tool exists elsewhere, and asks the two builders to talk.
Where it breaks down
A council has no value if it can’t decide, and it can’t decide if the wrong people are in it. It also only knows what the registry shows, so a stale registry means blind decisions. Done badly, it slows building and teaches people to hide what they make. If you hear that, shrink its remit before adding rules.
Try it with AI
Help me write a charter for an AI business council. Ask me one question at a time about our organization: who is building tools, where tools overlap, and who currently decides. Then draft the charter in four parts: who sits on it, what it decides (priorities, tool level, duplicates, retirements, and team reorganization), how it runs (cadence, registry review, written decisions), and the rules that keep it from becoming a gate. Make sure personal tools need no approval, the default is yes with conditions, and there is a response-time commitment. Flag anything that would slow people down.