As councils wrestle with how to deliver more and better with less, the recent surge of developments in artificial intelligence has felt like something of a gift.
AI offers to remove grudge work from frontline workers, speed up internal tasks and predict and prevent issues from arising, helping to deliver cost savings and improving workforce productivity.
By virtue of my role at the London Office of Technology and Innovation, where we help boroughs share best practice and experiment together, I have the privilege of seeing what is going on in many local authorities.
To my eyes, councils have been doing exactly the right thing with AI, starting small and exploring lots of different contexts to see what works. Those contexts have been as diverse as analysing visual imagery data to spot unlicensed billboards, to drafting education, health and care plans and from monitoring noise data from building sites to helping social housing tenants book appointments without having to go via their council contact centre.
Understandably, much of the work to date has focused on determining where AI can deliver benefits or savings in the immediate term. But, having got off the starting blocks, I believe it is now vital our sector starts thinking longer term, due to three issues.
While prices for the latest AI models are likely to rise, cheaper alternatives can be found using older, fit-for-purpose models or by running open-source AI locally on council-owned hardware.
The first is that residents are using AI, too. The impact is already being felt by many councils that have seen dramatic rises in the number and complexity of Freedom of Information (FoI) requests, complaints, planning objections and letters to elected members. The hallmarks of AI being involved are notable from how many seem to quote specific areas of policy and law.
The volume will only increase with the rise of agentic AI tools that can carry out a complex series of tasks independently. For example, there are already websites emerging that enable residents to object to all local planning applications with bespoke legalistic letters.
As a sector, we need to figure out how to respond. Temporary solutions may involve limiting the length of certain forms, or requiring a login to ensure a human is involved.
In other cases, AI itself can help. In one London borough, before someone can submit an FoI request, AI attempts to answer it using website data, past FoIs and internal policy documents. If the requester still wants more information, they can then proceed to a formal FoI submission. Longer term, the answer may be more about transparency.
Councils will find it to their advantage to provide as much information as possible on their websites so resident questions can be answered without requiring an officer.
Amazon makes clear on its website that its own customer service agents have access to no additional information than the customer can see in their own account.
The second issue is the pricing of AI tools. Currently, few AI companies are charging the true cost of the computing power it takes to draft our documents, create images or write code.
This is the classic and predictable big tech business model. Step one – get users hooked on the product at a subsided rate to win market share. Step two – hike the prices. Councils should expect this and ensure their business cases do not assume the current pricing stays the same when determining whether it is cost-effective to use AI instead of a person for specific tasks.
While this problem affects all sectors, local government's traditional procurement practices – where it is not unheard of to have five to 10-year IT contracts – could leave us more exposed.
We cannot afford to get locked into single IT vendors if we want to stay nimble. Technological solutions do exist.
While prices for the latest AI models are likely to rise, cheaper alternatives can be found using older, fit-for-purpose models or by running open-source AI locally on council-owned hardware.
Councils should engage early with their digital teams to explore these options.
The third issue is skills. The ideal approach is for AI to free up staff time for high-value human tasks, with staff acting as the ‘human in the loop' to check that what the AI has produced is correct.
However, the way this works may change over time. Consider the common use case whereby AI is used to transcribe social work meetings and draft case notes, with the social worker taking on the role of auditing the note.
We should ask: is their ability to spot errors and weaknesses based on the fact they honed that skill by writing notes manually for many years? Will newer social workers still have that skill and judgement five years from now?
Again, this applies to all sectors, but councils should consider how they maintain these core competencies as part of their workforce development plans.
None of these issues diminishes the huge potential of AI to help councils serve their residents better.
Love it or loathe it, the technology is powerful and here to stay. But if we want real, lasting results, the time to start thinking about these long-term trends is now.
Eddie Copeland is director of LOTI, London's collaborative public sector innovation team
