Interim Chief AI Officer — Integration (Energy and Utilities)

Three questions every AI use case must answer

Governing AI in an operating business comes down to three questions asked of every model that influences a decision. What evidence shows it works for this purpose, here, now? Who can override it, and do they actually do so when they should? And what would cause it to be switched off? Most organisations can answer the first in part, the second on paper and the third not at all. The Interim Chief AI Officer is appointed to make all three answerable, and holds executive authority to act on the answers.

The judgement calls are rarely about technology. One is how much evidence is enough to let a model influence a field or customer decision, given that waiting for perfect evidence means operating on intuition instead. The best practice is to make an early decision reversible and then tighten it as the facts improve. Another is when to retire a model that people like using but cannot justify. A third is how to treat local teams that built useful tools outside any governance at all: some deserve adoption, others closure, and the difference has to be explained in terms those teams accept.

The authority to make those calls is real. You will write the rules for when AI may be used, including the limits within which local managers can approve a use case, and choose the measures that steer AI spending, capacity and control. Each AI commitment is yours to continue, pause or retire. No plan goes forward while it leaves a team holding cost or risk it never accepted, and no investment proceeds that ignores its users, its data dependencies or how it could fail. The structure of the team beneath you and the way problems are raised need your approval; disputes with operating managers are yours to settle before money is committed, and you will decide the cadence of the board's reviews of progress, exceptions and unresolved questions.

Eighteen months from register to routine

By day ninety, a register of AI use cases, the evidence standard each must meet and the rules for human override should be agreed with the Regional President, with an owner for every model. By around month nine, retirement controls should have been tested on live use cases, and those lacking evidence or ownership paused or retired. By month 18, AI decisions should be governed through the agreed evidence, override and retirement controls as normal practice, with exceptions reaching the board on cadence.

Who can carry this

Through a regulatory remediation you kept the board's account of events tied to what operations could prove; the organisations where you did it ran across three countries, with 600 or more people and no less than $270 million in revenue. You accept that your first achievement here may be switching a model off rather than switching one on, and you can back that up with evidence. Selection will test what you have done in regulatory remediation, digital transformation and turnaround, and look for level 4 risk judgement alongside level 3 in data and AI governance, product delivery and architecture at scale. It is the wrong post for anyone who would keep an AI commitment running until each national head consented to end it, or who wants influence without 18 months of accountability.

A short, direct route

Only three stages. During the first week the Regional President talks with you by private video call; in week two the same Regional President sets you a practical case to work through; and in week three you sit down in person with the directors. Each stage is answered within four working days.