TL;DR
The head of the civil service has put AI at the centre of her plans for reform. Giving the annual Cabinet Secretary lecture at the LSE on 10 November, Dame Antonia Romeo said AI “will radically rewrite all of the domains in which the Civil Service operates”. She also warned that an AI-enabled state could leave the UK more exposed to cyber attack.
The case she made
Romeo described AI as “the defining opportunities and challenges of this generation”, with capabilities improving “at an exponential rate”. She went further: “AI potentially represents the most profound paradigm shift in modern governance.”
Her argument was that public services no longer have to be limited by how much human staff can handle. She pointed to work already under way, such as HMRC using machine learning to protect and recover £10 billion in tax in the last tax year, and MHCLG using AI on planning applications to help cut routine householder decision times by half.
She cautioned that automating single tasks without redesigning whole workflows “will not fully capture” the gains. She called ethical and transparent deployment, with guardrails, the “overriding principle”, and said civil servants should be given “the most relevant frontier models for their work”. On jobs, she said “jobs will change”, while arguing that AI will not replace human creativity, interaction or judgement.
Structures and the review
Romeo said the new OPMC, combining teams from No 10, the Cabinet Office and No 10 North that serve the Prime Minister, will coordinate a government-wide AI strategy. A review of the civil service’s organisation, performance and transformation, started in July and due with the Prime Minister before summer 2027, will ask what AI means for the civil service’s boundaries and accountability. The enlarged Fast Stream graduate scheme is to be “AI-focussed”, she added.
What it means
The lecture treats AI as a question of how government is organised, not just which tools departments buy, and it comes from the most senior civil servant. It follows permanent secretaries’ four priorities for a government AI model and a study finding that AI drafts bland policy.
In our view, the cyber warning matters most for suppliers. If departments rebuild workflows around AI rather than adding tools at the edges, the security of those systems becomes a delivery risk in its own right. The review’s findings on accountability will show who carries that risk when an AI-assisted decision goes wrong.