Make the governance you already pay for capable of governing AI

Your existing governance does not become obsolete because a system uses AI. Start with the authority the organisation already has, then make it capable of dealing with the additional work.

An industrial company already decides who can commit capital, accept risk, change a production schedule and sign a supplier contract. I would extend those mandates and name the owners of the AI-specific responsibilities.

Put the work with the people who own the decision

Investment committees should assess value after evaluation and support costs. Risk teams need visibility of shared models and concentrated supplier exposure. Engineering and process safety need evidence about output variability, material model changes, uptime and safe fallbacks. Procurement and data owners need explicit terms for information use, model updates and exit.

Fund the specialist competence and capacity to do this work. Adding AI to a policy heading achieves very little.

The business council I advocate connects those owners, settles cross-business trade-offs and coordinates common standards. Business leaders retain responsibility for outcomes. Independent assurance must remain able to challenge delivery.

Ask for one integrated operating design

The executive board should require a design covering investment, deployment, daily operation, change and retirement. For each stage, management should show who decides, what evidence they need, where they escalate, and which gaps require additional resources.

Test that design against a real workflow. If an AI-assisted process gives an unreliable answer, who stops it? If a supplier changes the model, who decides whether the existing evaluation still holds? If two business units depend on the same service, who sees their combined exposure?

These questions make governance concrete. A tidy organisation chart leaves them unanswered.

Repair weak structures, build missing AI capabilities and train the people who hold the mandates. Keep useful shared expertise and coordination. Make the responsibilities work together across the life of the deployment.

Your board already expects this discipline when investing in an industrial asset. Apply the same seriousness to AI.