Interim Chief AI Officer — Recovery Programme

An AI officer whose job is to govern, not to evangelise

Banks have used statistical models for decades in credit, fraud and pricing, and model risk management is a mature discipline. Newer AI systems stretch it. They are adopted faster, by more teams, often through vendors, and their behaviour can shift as the data shifts. The Interim Chief AI Officer is not appointed to increase the number of AI use cases. The task is to make sure each one exists for a reason, can be overridden when it is wrong and is retired when it no longer earns its place.

Three controls carry most of the weight. Evidence: what a model must show about performance, fairness, explainability and data lineage before it is used in a decision that touches customers or capital. Override: who may set a model's output aside, on what grounds, and how that is recorded so that patterns of override become information. Retirement: the conditions under which a model is switched off, and the discipline to do it when they are met.

Where the authority sits

This is an executive appointment reporting to the Technology Committee Chair. You will set how AI decisions are allocated, including the thresholds local leaders keep, and define the evidence required before any model is adopted, overridden or retired. You will rule on which AI commitments continue, turn back any proposal that offloads cost or risk with no one accountable, and decline investment where adoption, dependencies or the downside are unaddressed. You will approve how AI issues climb, and give the committee a fixed rhythm of reporting on progress, exceptions and decisions still open.

Twelve months from inventory to audited control

In the first ninety days, an inventory of AI use cases should be agreed, with evidence standards for adoption, override and retirement set for each. By month nine, model risk controls should be operating, with override and retirement decisions logged and reviewed by the committee, so that directors are looking at actual decisions rather than policies. By month 12, AI governance should belong to accountable business leaders, and audit should have tested the evidence, override and retirement controls.

A technologist who will say switch it off

You have operated in businesses that reached $840 million in revenue, 1,500 people and seven countries, and you can tell a model exception that a market really needs from one that persists out of habit. The committee will weigh your record in professionalisation, digital transformation and succession. It wants risk judgement at level 4, and at level 3 architecture at scale, product delivery, and governance of data and AI. You protect controls that work, put a visible price on each exception and leave the AI team better led than you found it. You give the committee a straight account of what is unresolved and do not let it become a pause. A technology commitment that should stop will stop without a tour of country leaders for their agreement, and you will answer for all 12 months.

Meeting the committee and its auditors

The Technology Committee Chair opens, in the first week, with a private conversation by video. The Audit Committee Chair takes week three, looking with you at how AI controls could be tested and evidenced. The fifth week brings you together in person with executive leaders from each function, who own the use cases, and in week 7 you meet the full board. Our comments on each stage reach you in writing within three working days.