Local government is adopting AI, but careful thoughtis needed

New Statesman​

13-06-2025

Photo byArtificial intelligence is already being widely used across local government to deliver services. These involve everything from customer service chatbots to writing paperwork to preventing damp. Central government is currently piloting a set of AI tools for local government. Other councils are working with commercial AI providers.
Somerset Council was one of the first local governments in the UK to join Microsoft's Early Access Program for AI. Somerset started with Copilot, a generative AI chatbot, and has now moved on to Magic Notes, which turns meeting recordings into a first draft of paperwork and forms and is now used by around half of social care teams in the UK as a way to improve the quality of paperwork while reducing the time spent on it by 50 per cent. Somerset Council says this has allowed frontline staff to free up time to be spent on 'person-centred work' and improved their work-life balance.
AI has great potential to improve democratic processes and decision-making and from there flows improved services for citizens, according to Alexander Iosad, director, government innovation policy at the Tony Blair Institute. AI can be used to engage with citizens and get data from them to accelerate the back-office process and to improve decision making, he explains.
One problem of the traditional process for policymaking is having 'more data than you can turn into actionable insights', Iosad says. AI is not just able to process all of that information, and to do it faster so the information is still current or in real time, but is able to cross-reference it across services.
However, there are still significant hurdles to integrating AI tools into council services and realising its potential. A survey of county councils and unitary councils by the County Councils Network (CCN) found that while 85 per cent were using AI (and the remaining 15 per cent were considering using it), local governments felt that a lack of staff capacity, funding and training were holding them back. The Tony Blair Institute report says councils 'lack the confidence, capabilities and infrastructure required to unleash this innovation'.
The Tony Blair Institute worked with a local council in the UK and estimated that AI could be applied to 26 per cent of tasks, resulting in a saving of one million work hours, or £30m in financial costs, each year. Scaled up nationally, this would mean around £8bn in cost savings.
Organisations of all types, but particularly government and public services, are looking to AI to save them money, but less than half of those surveyed by the CCN said AI had resulted in savings so far. This compared to over 90 per cent reporting improved staff productivity and three quarters that said it improved services. Transformation of services usually requires significant investment in order to realise improvements to the quality of services and their savings. Big digital projects have historically been difficult to deliver and to realise their benefits, particularly in the public sector. In 2021, Birmingham City Council invested £19m in a new IT system with the aim of reducing costs, but spending on getting the system up and running has since spiralled to £90m.
Subscribe to The New Statesman today from only £8.99 per month Subscribe
There are also questions over some of the ways AI has already been deployed. Several councils have bought and use AI systems to identify families in need of support and young people who may be at risk of becoming involved in gangs, while others use them to highlight potential council tax fraud. These have been mired in allegations of bias, consent and lack of transparency.
In Rotterdam, an algorithm was used to assign a fraud risk score to residents based on data from 12,707 previous investigations. But in practice, women, young people, parents and migrants got higher scores and were more often flagged for investigation. Subjective assessments by caseworkers, such as whether a person was 'flexible' enough to deal with the challenges they were facing, were incorporated into the data, giving it the sheen of neutrality. The city suspended its use in 2021 following an external ethics review. In 2015, Hackney Council paid for an Early Help Profiling System provided by the company Xantura that used data collected by the council to identify families that were 'at risk' – the families were never informed their data would be used in this way.
These challenges mean the governance, transparency and privacy policies around AI in local government need to be robust. The Local Government Association says 'councils must also be mindful of the risks and challenges associated with safe and responsible AI adoption', listing strong data foundations, data protection, bias and privacy as key issues to address. The Tony Blair Institute advises councillors to champion AI innovation, but to de-risk it by looking to tools that are already in use, collaborating, using existing standards and training the workforce to improve AI literacy.
'The biggest challenge is data,' says Iosad. 'It's getting the right data and then setting up the systems allowing you to work with data effectively.' Where the UK is arguably weakest, he says, is having data sets collected by government that can be linked and used together in order to really personalise services and to make effective decisions at local, regional and national levels.
'The blueprint for modern digital government has a very encouraging line on this, not just recognising that there is a challenge, but mostly setting out the ambition that every data set that is used in the public sector has an API,' Iosad says. He adds this will also put in place the safeguards required for data.
What is clear is that AI tools and technologies will be taken and used by local government in the way that new technologies have been in the past. Some of the risks and challenges around technical literacy, training, choosing good suppliers and managing those contracts are familiar. However, the challenges of transparency, scrutiny and identifying bias in complex systems are more novel and require local government not just to look for an easy answer but the best solution to create a better place and life for residents.
Related