Solihull Council has begun a 24-week programme to examine where artificial intelligence could improve public services, choosing to build a portfolio of evidence and business cases before committing to wider technology deployment.

The West Midlands authority has appointed public-sector AI supplier ICS.AI to assess opportunities across adult social care, children’s services, economy and infrastructure, and public health. The work is expected to identify about 200 possible applications, narrow these to 50 finance-validated cases, and produce a roadmap for future investment.

Rather than beginning with a single chatbot or automation platform, the programme will examine the council’s readiness across five dimensions using ICS.AI’s target operating model. The supplier says privacy, ethics, and safeguarding will be applied during the assessment, with council-owned baseline data used to validate the eventual business cases.

Councillor Dave Pinwell, Solihull Council’s cabinet portfolio holder for resources, said the authority wanted to identify improvements that would deliver the greatest benefit for residents while providing “full value for the Council”. His statement places financial discipline alongside service quality, although the programme’s usefulness will depend on how those objectives are measured when they pull in different directions.

A catalogue is not an investment case

Local authorities have many processes that appear suitable for automation, from triaging correspondence and summarising case notes to identifying missing information, scheduling appointments, and helping staff search policy documents. Listing possible applications is comparatively straightforward; demonstrating that a system can operate safely, affordably, and reliably within a live service is more difficult.

Adult social care and children’s services present particularly demanding conditions because decisions can affect care packages, safeguarding interventions, family relationships, and access to statutory support. Even where an AI system does not make the final decision, its summaries, risk indicators, or prioritisation can influence the evidence placed before a professional.

The discovery programme will therefore need to distinguish administrative support from systems that materially shape decisions. A transcription tool, for example, carries different risks from a model that recommends whether a case should be escalated, while automated resident communications require controls to ensure that vulnerable people receive accurate information and can reach a human service.

Public health and infrastructure services create different questions around data sharing, forecasting, and accountability. Predictive tools may help identify maintenance demand or target interventions, but historical council data can reflect earlier service gaps and uneven patterns of access. A technically accurate model may still direct resources poorly if its underlying measures do not represent the outcomes the council is trying to improve.

Finance validation needs operational evidence

Solihull’s target of 50 finance-validated use cases suggests that the council wants more than a broad innovation catalogue. Credible financial models must account for software licences, integration, information governance, staff training, procurement, model monitoring, security, and the cost of maintaining a human route for cases that cannot safely be automated.

Claims about time savings also require a plan for converting released hours into useful capacity. Removing several minutes from an administrative task does not automatically reduce expenditure or improve services, particularly where workloads remain fragmented across teams and the saved time cannot be consolidated into fewer posts or additional appointments.

Techopia has previously examined the gap between public-sector AI productivity claims and measurable service improvement. Solihull’s discovery-first approach could help close that gap if each case identifies the operational change required alongside the technology, rather than treating model access as the intervention.

The council will also need to decide how supplier independence is preserved during the assessment. ICS.AI is being paid to identify and prioritise opportunities in a market where it also provides AI transformation products, creating a potential incentive for the roadmap to favour applications suited to its own technology. Transparent evaluation criteria, open procurement, and access to the underlying analysis would reduce that risk.

Existing council protocols provide another layer of governance, but policies must eventually become practical controls. Staff need to know which systems they may use, what information can be entered, when outputs require verification, how errors should be recorded, and who remains accountable when an AI-assisted process causes harm.

The roadmap will be judged by what it excludes

A strong discovery programme should reject unsuitable ideas as clearly as it recommends promising ones. Some processes will lack usable data, while others will be too sensitive, too infrequent, or too dependent on professional judgement to justify automation. In several cases, simpler workflow redesign or better integration between existing systems may produce greater value than an AI model.

The planned shortlist of 200 opportunities therefore creates a risk of false precision. Unless each proposal begins with an identifiable service problem, the exercise could produce a large inventory whose scale looks impressive but offers little guidance on which investments should proceed.

Conversely, a smaller set of well-supported cases could give the council a defensible route from experimentation to procurement. Useful outputs would include baseline service costs, expected benefits, data requirements, equality and safeguarding assessments, technical dependencies, and criteria for stopping a deployment that fails to deliver.

Solihull has started by delaying implementation until it has examined readiness and value, although discovery work is not itself transformation. By the end of the 24 weeks, the council should be able to show not merely how many possible uses it found, but which proposals survived scrutiny, which were rejected, and what evidence residents will be able to see before public money moves into deployment.