AI is already shaping decisions in councils across England — from social care to planning. Getting the governance right turns out to be less about values and more about plumbing. New research from ai@cam’s Decision-Making with AI in Connected Places and Cities project suggests the main obstacle facing councils isn’t a lack of ethical intent, but the institutional capacity to translate principles into routine administrative work.
The research highlights three realities of governing AI in local government and what it takes to address them.
1. The principles that get acted on are the ones already built into everyday processes
Not all ethical guidelines are acted on. And the pattern of which ones stick is revealing: the principles that get embedded are the ones that already have a procedural home.
Principles like transparency and data protection tend to integrate relatively easily into local government practice. They closely align with long-standing administrative routines – traditions of record-keeping, established lines of responsibility, and mature legal frameworks that set clear requirements. Where practical tools already exist, ethical commitments can be incorporated into everyday work and sustained over time.
Other principles, such as fairness, follow a different trajectory. Fairness often requires contextual judgement about how algorithmic systems affect different groups. These judgements are harder to translate into standardised administrative procedures. Without a practical anchor in existing routines, they tend to stay as intentions rather than actions, creating real reputational and legal risk for councils.
The implication is that governance tools are not neutral: decisions about what gets documented, by whom, and when shape which values get acted on. The most practical path forward is to build on what already exists. Councils can adapt existing procedures – such as equality impact assessments, data protection reviews, and procurement processes – to ask the right questions about AI, rather than building entirely new frameworks.
2. The ethics problem is a capacity problem
Local authorities are already adopting AI, and do not operate in a governance vacuum. The values to inform AI development are there — transparency, accountability, fairness, privacy. The difficulty is elsewhere: translating those values into stable administrative practices that can survive contact with the daily pressures of local government.
Councils operate in highly structured governance environments. Decision-making processes are already shaped by legal duties and established administrative routines. Introducing AI into this landscape raises a practical question: where do ethical expectations actually fit within existing procedures?
In this light, responsible AI becomes a question of organisational design. Ethical principles can shape behaviour if they are connected to concrete administrative tools - forms, templates, review procedures, procurement processes, defined roles, and lines of responsibility. Without these anchors, ethical guidance risks remaining aspirational rather than operational.
3. Good governance needs the right tools
When central government responds to concerns about AI by producing more principles and frameworks, it can inadvertently add to the burden on local authorities rather than easing it. Guidance that sets out ethical expectations without providing usable tools simply widens the gap between what councils are supposed to do and what they can realistically manage. What local government needs most is practical, adaptable administrative tools, not more statements of intent.
But better processes are not enough on their own. The research also flags a subtler risk: the very standardisation that makes governance workable at scale can, over time, hollow out ethical practice and reduce genuine reflection to a routine box-ticking exercise.
Ethical AI in local government can not be achieved through frameworks alone. It requires institutional commitment and ongoing review – a willingness to revisit routines, ask honest questions, and adapt when needed. The councils getting this right aren’t waiting for a new framework. They’re asking better questions within the processes they already have.
This piece draws on the paper “From Virtue to Practice: Operationalising AI Ethics in Local Government in England” by Kristina Khutsishvili, Kwadwo Oti-Sarpong, Viviana Bastidas, and Jennifer Schooling, developed as part of the ai@cam project, Decision-Making with AI in Connected Places and Cities.