Derby Council AI helpers resolve only half of cases three years after launch

TL;DR:

• Derby City Council’s AI assistants Darcie and Ali, introduced in 2023 at an initial cost of £168,000, resolve about 50% of cases without human staff stepping in — a rate that has not improved in three years despite what the council called a “major upgrade”.
• The two systems have handled more than 3.2 million routine inquiries since launch and the council says four full-time-equivalent agency roles were lost the year after introduction, saving roughly £200,000 a year.
• The council’s wider AI programme claims more than £12m in identified savings across the authority, half of them in adult care, while opposition leader Steve Hassall warned the rollout risks becoming “a digital barrier” rather than improved customer service.

Derby is one of the most-watched UK local-government AI deployments because it published its plans publicly and committed to a “digital front door” approach earlier than most other councils. The unchanged resolution rate now matters precisely because Derby was the test case other authorities were waiting on. Three years in, the productivity headline survives — fewer agency staff, savings claimed — but the customer-resolution headline does not.

A useful UK data point against productivity-claim inflation

The 50% resolution rate is informative for the rest of UK local government. Councils across the country are evaluating digital-assistant rollouts on the back of supplier productivity claims. Derby’s three-year curve — initial gains followed by a plateau, with an acknowledged failure to handle the Derbyshire dialect — is exactly the kind of mixed-picture data the Ada Lovelace Institute warned this week should sit alongside headline numbers. Resident feedback is 77% positive among those who provide it, but a customer focus group had to be introduced to address an early complaints spike, and the leader of Derby Conservatives Steve Hassall has framed the rollout as cost-cutting “dressed up as innovation”.

The figures also raise an under-discussed question: what happens to the cases the AI does not resolve? Roughly half of contacts still need a human agent — but if agency headcount has been cut and call volumes have shifted to the AI-then-human path, the wait times and complaint paths for the other 50% may be worse, not better. The £12m wider-AI-programme savings figure — half from adult care — has not been broken down in detail, which matters because adult care is the area where service-quality consequences of automation are most acute. Derby’s experience also bears on the live national debate about whether councils should be allowed to keep AI savings as efficiency gains or reinvest them in service.

Looking forward

Other UK councils evaluating digital-front-door rollouts — including those Derby’s commercial partners are pitching to — now have a three-year UK case study to benchmark against rather than supplier-supplied projections. Expect the Local Government Association to be asked for guidance on minimum service-quality metrics, and for the Information Commissioner’s Office to revisit how dialect, accessibility and digital-exclusion considerations apply to AI customer-service tools. For residents and elected councillors elsewhere, the live question is whether to ask for the resolution-rate trajectory before signing the next contract — not just the savings projection.