AI governance

Your AI pilot works. Nobody will sign it off.

Pilots stall when risk, compliance and IT ask who owns the process the AI sits in, and nobody has written it down. We map that process so the answers exist before the meeting.

Take the five-minute scorecard Book a 30-minute pilot review
18 questions6 areasYour score appears as you answer, no sign-up

Where pilots stop

The meeting after the pilot works

The pilot hits its numbers and the business case is signed off. Then risk, compliance, legal and IT each arrive with their own questions, and every one of them is about the process the AI sits in.

Klarus surveyed 500 senior leaders at UK and Irish mid-market firms in 2026. Only 10% had scaled all their AI past pilot, and 48% named governance, ethics, security or privacy as a reason projects stalled. Read the Klarus findings

  • What customer data does it use, and where does that data go?
  • Which decisions can the AI make on its own?
  • Who signs off an answer before a customer sees it?
  • What happens when it gets something wrong?
  • Who owns the risk, by name?

One pilot that got through

From "the AI works" to a process someone will approve

A client had built a generative AI assistant to help service agents answer customer queries. Response times fell and the agents liked it. Its answers were accurate. When the business tried to move it into production, risk, compliance, legal and IT each asked different questions, and none of the answers had been designed into the pilot.

We modelled the whole process around the assistant in BusinessOptix, the process modelling platform we partner with: where data enters, which systems it touches, where a person has to judge, which controls exist and who owns each decision. We then laid risks, controls, roles and evidence over the model. That gave the client a target operating model, a description of how the process will run in production.

The pilot became a governed production process, with named owners and evidence that the controls work.

Pilot Readiness Scorecard

Score your pilot against the questions it will face

Answer for the process your pilot will change. Each Yes scores 2 and each Partly scores 1. Your score and the two areas to fix first appear as you go.

 

The AI Governance Assessment

Three weeks, one pilot, a fixed price

We map the process your pilot will change, test it against the six areas in the scorecard, and hand back what your sign-off meeting needs.

  1. Week 1, map. Workshops with the people who do the work. We build the current process in BusinessOptix and mark where the AI acts, where people decide, what data moves and who owns each step.
  2. Week 2, test. We take the map through the six areas with your process owner, data owner and risk lead, and write up each gap with its evidence and an owner.
  3. Week 3, plan. We agree the order to close the gaps, draft the controls against the process steps, and present to whoever approves the pilot.
  • Process mapThe process as it runs today, with AI touchpoints, decisions, data and owners marked.
  • Readiness scoreA score for each of the six areas, with the evidence behind it, on one page.
  • Gap registerEach gap with its consequence, owner and fix, in the order to tackle them.
  • ControlsDraft controls and human approval points tied to the steps they protect.
  • 90-day planWhat has to happen before go-live, who does it and how long it takes.
  • Sign-off packA short readout written for whoever approves the pilot.

The process map stays yours, as the first entry in a process library you can extend to the next use case.

Three ways in

Pilot Readiness Scorecard

Eighteen questions across the six areas, scored as you go, with the two weakest areas explained.

Time
5 minutes
Cost
Free
Start the scorecard

Pilot review call

A call with a Cocoon consultant to go through your scorecard and the pilot you have in mind.

Time
30 minutes
Cost
Free
Book a call

AI Governance Assessment

The three-week review above, for one pilot or planned use case.

Time
3 weeks
Cost
Fixed, agreed on the call
Talk to us about it

Seven AI governance engagements in the last 18 months.

We have also connected the systems behind an iconic British motorcycle brand, automated supplier payments across a national pub estate, rebuilt patient intake for a private hospital operator and untangled KYC and AML compliance across several jurisdictions. Each one had to answer the same sign-off questions an AI pilot faces now.

Tell us where the pilot is stuck.

We will tell you honestly whether we can help.

Book a 30-minute pilot review Take the scorecard first