Programs · Governance
Governance Network Workshop
Name who governs your AI, and announce it Monday. In 90 minutes, your governance committee and key stakeholders sort the AI decisions piling up, put real people in each seat of the network, and draft the announcement before they leave the room.
Every week without it, more agents ship with nobody's name on them.
Who should be in the room
Anyone on an existing AI or data governance committee, plus the key stakeholders whose teams live with its decisions. Up to 20 people.
Nobody gets blamed
Nobody in the room gets blamed for anything that comes up. Your people already know where decisions get made and who they go to for answers. The facilitator supplies the structure, so the network belongs to the people who'll run it.
The three layers you are naming
A governance network has three layers. Round 4 puts real names into each one. Not job titles, names.
Central AI Council
Purpose, risk appetite, principles. It is the compass, and it is not an approval queue.
Cadence: quarterly.
Functional AI Councils
Tool approval, data-use practice, guardrails for one part of the business. They decide inside the principles the center set.
Cadence: monthly.
Guardrail Keepers
Practitioners, appointed by their peers, answering the day-to-day question in the moment.
Cadence: the speed of the work.
Round 4 ends with a test. The room takes one real decision from Round 2, walks it through the network it just built, and asks who touches it and how long it takes now.
How the 90 minutes run
- 15 min
Where do AI decisions get made here?
Everyone marks privately where decisions happen today and where they should. The gap between answers starts the conversation.
- 15 min
What are we actually arguing about?
The room lists real AI decisions it has faced or can see coming, with dates and names attached.
- 20 min
Which decisions need the Central AI Council?
Each decision gets sorted by stakes and by how hard it is to undo. Most rooms find the council's pile is smaller than they expected.
- 25 min
Build the network.
Names go into three layers: the Central AI Council, Functional AI Councils and Guardrail Keepers. The people others already bring their AI questions to get named first.
- 15 min
What we announce Monday.
The room drafts the announcement, names the overcorrection it won't make, and sets a first checkpoint with a date and an owner.
What the room produces
The one page you walk out with
The network goes on a single page: each seat, the people in it, and what that seat decides. It is short enough to send, which is the point. A governance design nobody can read does not get followed.
The overcorrection you are choosing not to make
Round 5 asks the room to name it and say it out loud in the announcement. The people who built things without asking permission need to hear that they are not the target. A governance announcement that reads as a crackdown gets the experimenting moved somewhere nobody can see it.
A date and a named owner
The last thing on the page is the first checkpoint, with a date on it and one person's name beside it. A session that ends without those has produced a document rather than a move.
By the end of the 90 minutes
- A named governance network on one page: people, seats and decision rights
- Your real AI decisions, sorted by which ones need the Central AI Council
- A drafted announcement
- A first checkpoint with a date and a named owner
Why a policy document does not do this
The usual first move is a written AI policy. It says what is allowed. It does not say who decides the case it did not anticipate, which is most of them, and the model changes underneath it every few months.
The second is a central approval queue. It works until volume arrives, and then it becomes the reason people stop telling you what they are building.
The third is buying a governance platform first. A platform records and enforces decisions. Somebody still has to make them, and the network is who.
A network decides at three speeds, so the day-to-day question does not have to wait for the quarterly meeting.
The longer case, with the research- Format
- 90-minute facilitated working session
- Delivery
- Virtual or in person. In-person delivery adds travel and expenses, billed at cost.
- Group
- Up to 20 people
- Terms
- The proposal sets final scope and terms.
Larger groups, several business areas at once, or a session that has to run around an incident change the shape. Say what you are working with and we will scope it.
Designed by Melissa Reeve, delivered by a Hyperadaptive delivery partner.
Next step: the 90-Day Activation HubQuestions about the Governance Network Workshop
- We just had an AI incident. Is this the right time?
- It can be. The facilitator keeps the room on the design question: what would have had to be different for this not to happen? The incident informs the network, and the session stays out of blame.
- What if we don't have a governance committee yet?
- Start anyway. "Nowhere" is one of the answers in round one, and the network you build gives governance its first home.
- Do we need a governance platform first?
- No. The network makes the decisions, and a platform can record and enforce them later.
- Who facilitates?
- A Hyperadaptive delivery partner trained in the model, running Melissa's design.
Where this comes from
Designed by Melissa, run by a partner
The three-layer network is part of the Hyperadaptive Model, set out in Melissa Reeve's book with IT Revolution. Melissa designed the 90 minutes. A Hyperadaptive delivery partner trained in the model runs it, and Melissa observes deliveries.
How the model works