A tale of more than two cities: risk, reward and AI governance

This think piece explores the risks and rewards of AI adoption in local government, highlighting varying approaches from overly cautious to secretive. It advocates for a ‘middle way’ of managing AI risk through transparency, collaboration, and continuous learning and emphasises that, above all, AI should be viewed as a powerful tool requiring careful handling.

It was the best of times, it was the worst of times; it was the age of automation, it was the age of algorithmic anxiety; it was the epoch of opportunity, it was the epoch of overpromising; it was the season of data-driven hope, it was the season of reputational dread.

In the world of local government AI adoption, we find ourselves not in Dickens’s London or revolutionary Paris, but perhaps somewhere closer to Brighton or Brisbane, Stirling or Sligo. Councils today face a revolution of their own: the rise of artificial intelligence in public services.

Some are marching boldly into this new era, AI tools in hand and chatbot pilots in motion. Others are peeking nervously over the parapets of policy and procurement, wondering if they should even open the email from that eager vendor promising to solve all their problems with “cutting-edge machine learning”.

Let’s start with just two cities featured in our new LGIU case study on AI governance.

AI generated image

San Francisco, where AI-powered surveillance is being deployed with minimal oversight, and where city councillors have been shocked to discover there is no full list of what AI tools are actually in use. A city of innovation, yes, but also a cautionary tale of the risks of ungoverned experimentation.

Then there is Amsterdam, where AI governance is practically a civic art form. Ethical impact assessments, public registers of algorithms, multi-stakeholder consultations. It’s all very structured, very Dutch, and very impressive. But it’s also the product of significant national investment, an EU regulatory environment that industry insiders have labelled stifling (that’s what they would say) and a high baseline of public trust and engagement.

Most councils, it’s fair to say, are neither San Francisco nor Amsterdam. They’re not rolling out mass surveillance without scrutiny, but they’re also not quite ready to launch their own ethical AI registries. More often, they’re in a muddle in the middle: under-resourced, under pressure, and unsure whether they’re being too cautious or not cautious enough. So what’s a sensible council to do?

Let’s talk about risk

Recent research by LGIU and Browne Jacobson found that risk management in local government has undergone a quiet revolution of its own. Once the domain of box-ticking and back-office reporting, risk is now becoming embedded in how councils think, plan, and act. The best councils don’t see risk as something to be avoided at all costs, but as something to be understood, shared, and, crucially, managed.

AI deserves the same treatment. It is neither saviour nor saboteur. But it is powerful. And powerful tools require careful handling.

But harmful and old fashioned attitudes to risk can be hard to shake out. We can sometimes see four types of council AI behaviour emerging:

1. The AI-averse council – unwilling to dip a toe in the algorithmic waters for fear of reputational harm or ethical pitfalls. The result? Missed opportunities to improve services, ease staff workloads or save money. The risks of not providing better services or value for money are underestimated, the risk of seeming out-of-touch and unable to handle the impact of AI on local workforces is ignored and the potential implications of innovative staff leaving or experimenting are swept under the rug.
2. The invisible AI council – where different departments quietly trial AI tools without central oversight. Risky, not because the tools are necessarily bad, but because no one is keeping track. It isn’t clear where AI is impacting decision-making or services to senior staff or to the public.
3. The lock box council – where use of AI is not just monitored but heavily scrutinised, without the opportunity to experiment and learn under lower-risk conditions. Red tape and the perception that failure has a high cost limit staff willingness to consider doing things in a new way.
4. The black box buyer – dazzled by vendor promises, buying tools they don’t fully understand. Outputs become harder to question, and accountability gets murky.

Unhelpful attitudes to AI can expose councils to risk or miss out on opportunities.

Using a balanced risk approach

In the LGIU report Navigating risk in local government: Challenge, change and capacity we outline how some of the characteristics of councils that manage risk well. This more mature approach to risk can also be used to govern AI well. Adapting key lessons from that report, we can describe how local government can use AI to support innovation, efficiency and improved outcomes while managing the risks of an emerging and potentially revolutionary technology.

Great Expectations: Leading with Vision and Purpose

Strong leadership sets the tone. Councils that manage AI well have clear, strategic ambitions, actively support responsible experimentation, and ensure oversight is in place. They treat AI as both an opportunity and a responsibility.

• Visible leadership:  Leaders don’t just support risk management – they also champion responsible AI use. They set clear expectations around transparency, accountability, and citizen benefit, ensuring AI aligns with council values and priorities.
• Watching the public mood: Public trust in council AI use is hard-won and easily lost. Local governments actively monitor reputational risk, involve communities in decisions, and demonstrate how they will ensure that AI use is fair and effective.

Hard Times: Embedding AI in the Real World

Rather than seeing AI as hype or hazard, these councils embed it into real service delivery and risk frameworks. They anticipate consequences, consider resource realities, and plan for long-term sustainability—not just quick wins.

• Embedding risk in reality AI is treated as a real-world risk and opportunity – not a sci-fi fantasy or a tech toy. Councils embed AI risk considerations into service planning, workforce strategy, and financial oversight from the outset.
• Cross-departmental collaboration: AI governance isn’t left to IT alone. Procurement, legal, comms, HR, and service delivery teams are involved in shaping how AI is adopted, reviewed, and governed across the organisation.

A Tale of Two Cultures: From Fear to Trust

Councils move away from either paralysis or unchecked enthusiasm. Instead, they foster a culture where people feel safe to question, test, and learn. They engage residents early, talk openly about risks, and ensure decisions are transparent and fair.

• From fear to learning: A healthy culture welcomes exploration and experimentation with AI – within clear ethical boundaries. Councils allow space to trial and learn without fear of reputational ruin from early missteps.
• Relationship building matters: Strong relationships – with staff, suppliers, and citizens – support ethical AI. Councils that listen well, engage early, and build trust are better placed to deploy AI that works for everyone.
• The courage to question: Councils encourage challenge: “How does this tool work?” “What data is it trained on?” “What are the risks?” “What if we stick with existing processes?” This builds resilience against black-box technologies and vendor overclaim.

Bleak House No More: Making AI Understandable

Good governance demands clarity. These councils communicate in plain English—about what AI does, how it’s used, and who is accountable. Internally and externally, they demystify algorithms and explain impacts clearly and accessibly.

• Clear and connected communication: AI tools and their impact are communicated in plain language, both internally and to the public. Councils provide accessible information about how algorithms affect decisions and outcomes.
• Investing in tomorrow’s talent: Councils invest in developing AI literacy across the workforce – not just hiring data scientists. Staff are supported to understand AI risks, ask the right questions, and lead innovation responsibly.

Using better approaches from risk management can support getting the best from AI

Shared risk, shared learning

In our report, we make some specific recommendations for the English political context, but these can be extrapolated for anywhere. For example, we call for a centre-local shared learning platform for risk management. AI risk doesn’t stop at borough boundaries or national borders. Local governments must work together – through networks, shared registers, and open forums – to pool expertise, share what works, and avoid common pitfalls.

Whether you’re cautiously curious or already deep into deployment, we invite councils everywhere to get involved in the Future Local Lab as we explore what AI (and other issues) mean for the future of local democracy and public services.

Because in this tale of many councils, the best of times won’t come from perfection. It’ll come from collaboration, curiosity, and the courage to manage risk openly and intelligently.

And if we get it right?

It will be a far, far better thing that we do, than we have ever done.

Looking for more?

Explore LGIU’s new Future Local Lab – a holistic programme of resources aimed at helping you to prepare for and respond to the emerging challenges facing local government everywhere.