Interim Group Chief Information Officer — Integration

Who may overrule a model, and who may switch it off

AI governance is often discussed in terms of principles. In practice it comes down to three direct questions. What evidence must a model produce before its outputs are trusted in a live decision? Who has the right to override it when it is wrong, and how is that override recorded? And who decides when a model should be retired, on what criteria? A business that can answer all three for every system in production governs its AI. One that cannot is relying on goodwill.

The Interim Group Chief Information Officer, accountable to the Technology Committee Chair, holds executive authority to answer those questions across the group for 12 months. You will agree the group technology thresholds kept by local leadership, and set what a request for technology money, people or a control exception must demonstrate. Inherited technology commitments survive only on your confirmation. You can reject a proposal that hands a model's risk to a team that never agreed to hold it, and a value case that assumes people will trust and use an AI system, ignores the data it depends on, or says nothing about how it could go wrong. You approve how group technology leadership is organised and how issues rise, settle disputes between the committee and those running operations, and decide the cadence of the board's reviews of progress, exceptions and choices still open.

What twelve months of control should look like

Settling the rules is the work of the first ninety days: evidence standards, override rights and retirement criteria for AI decisions, agreed with the Technology Committee Chair and assigned to accountable people. Around month nine the controls should be tested on live systems rather than on paper, with the findings reported to the Audit Committee and each gap given to someone to close. By month 12 AI governance should be part of the group technology cadence, models retired when they meet the agreed criteria, and every override decision supported by evidence.

The leader, and how we will meet

Through professionalisation you kept what the board heard consistent with what operations showed, and you have found out who truly pays for an exception before allowing it to run on. The businesses where you gained that experience were in technology or digital platforms, each reaching $750 million or more in revenue, at least 1,300 people and four or more countries. Alongside professionalisation, the board weighs digital transformation and succession. Risk judgement at level 4 is decisive, because override and retirement are risk calls first; data and AI governance, product delivery and architecture that scales are each expected at level 3. You can tell where local discretion should stay and where it should go, and you write the boundary down.

First comes a remote conversation, held in confidence, with the Technology Committee Chair, early in week one. A working session with the Audit Committee Chair in week three tests how you would evidence override and retirement controls. Week five is a face-to-face panel of executives who lead the group's functions; week seven, a discussion with the directors. A response in writing follows each stage within 3 working days.