Where should your council start with AI? Six takeaways from our SMART: AI Assessment webinar
By Fiona Watson, Chief Marketing Officer, ICS.AI
Every council we speak to is asking the same question. Not whether to use AI - but where to start, what it will save, and how to be sure the savings are real.
On 30 September we spent an hour webinar on exactly that, with Martin Neale (CEO and Founder), Aneeq Star (Chief Revenue Officer), Dwayne Johnson (Chief Local Government Officer), Nick Horne (AI Enablement Consultant) and me as host.
Here are six takeaways from the session. Over the next six weeks we'll go deeper on each one, and the full recording is available to watch on demand.
The easy savings are gone - and savings have to land in year
Martin opened with the pressure most of our audience live with every day. In February, the government agreed around £1.5bn of exceptional financial support for a record 35 councils for 2026-27. The LGA estimated councils face a £7bn funding gap by 2028-29 just to keep services at today’s level, and half of council’s finance chiefs (51%) lack confidence they’ll have enough funding to meet their statutory duties by then.
After more than a decade of cost cutting, the traditional levers - shared services, headcount freezes, maximum council tax rises - have largely been pulled. The next savings won’t come from cutting budgets. They’ll come from changing how the work gets done. That is where AI comes in: it changes the economics of the work, absorbing demand and giving staff time back.
Timing matters as much as scale. Many of the councils we work with need their savings in-year. A saving identified for 2028 doesn't help a council that runs out of road in 2027.
Done as a joined-up programme, the value is significant: typically £8m-£19m a year for a unitary council, on lower-bound assumptions.
Pilots don't add up to transformation
AI is everywhere. 95% of councils are exploring or using it. Yet 56% don't know whether it has saved them any money, and only 22% report realised cost savings (1).
Martin's point was that the difference isn't the technology - it's the operating model. Scattered pilots each bring a different vendor, different guardrails and a separate bill. They never compound, governance gets duplicated, and the council is left to turn each one into savings.
The alternative is a whole-council transformation programme: hundreds of use cases, one governance framework they all inherit, and savings tracked from day one. And because new generations of AI arrive every 18 months to two years, it has to keep moving - as Martin put it, transformation is "a photograph of a moving train".
Derby City Council: where it started
Derby City Council has shown what that looks like in practice. Derby first proved AI working for residents through its AI front door, then used the SMART: AI Assessment to build a council-wide plan and business case. The assessment itself produced:
261 AI opportunities identified, narrowed to a portfolio of 54
44 planning, workshop and feedback sessions, involving 100+ council staff
That plan then led into a wider AI transformation programme, governed by an AI Ethics Board with an independent chair and a DPIA for every use case. Across that programme, Derby has handled 1.4m+ resident enquiries through its AI front door, resolving up to 56% against 20% expected, and across the wider AI transformation programme identified £12M+ in savings.
Every council starts from a different place
Aneeq and Dwayne spend much of their week with council leadership teams across England, Scotland and Wales. Four themes come up again and again when councils come to us:
Aneeq shared a question a chief executive once put to him: if you know all of these tools work, why do we need an assessment? It's a fair challenge. The answer is that no two councils are starting from the same place. One council had a clear focus on cashable benefits within 18 months because of its financial position. Another council that had already stabilised its finances wanted something different - to hear from its own staff where the pain points were, and where outcomes for residents could improve.
Dwayne, a former director of social services and chief executive, added that what trips councils up is rarely the tools. It's the vision, the governance, and how new AI works alongside legacy systems.
What 12 assessments have taught us
We have now delivered 12 SMART: AI Assessments across UK local government and higher education. Across them all, five things make the difference:
Leadership: a clear vision, a whole-council mandate, and honesty about where you are and what you're trying to achieve. Buy-in has to come from senior leaders, elected Members and staff alike.
The right people: quality over quantity. Leaders, managers and the people who do the work, with finance and a technical lead in the room.
Data: the more you share, the better the outcome - and quality matters as much as volume.
Approach: real use cases brought to life in workshops, benefits realisation from day one, and working alongside your other technology partners.
Governance: ethics built in from the start, a risk-based approach, and a ready-made risk register that can flex as AI changes.
Aneeq also addressed a question many people raise early on. None of our assessment reports has resulted in compulsory or voluntary redundancies at any authority, and none of our reports have recommended them - despite identifying millions of pounds of savings.
Inside an assessment: every service, every level, every voice
Nick has been on site for all 12 assessments. Across them, we have identified 1,450+ AI opportunities and £63-98m in total savings.
On the ground, that means 5-7 days on site and 90-120 minute workshops across the whole council, not just IT. Each one brings together senior leaders, managers, the people who do the work day to day, finance and technical leads.
Every voice counts. Sceptics' concerns - "Is our data really safe?" - become design rules: governance, safeguards and a human in the loop. Enthusiasts become the early wins and AI Champions. Bring both with you, and the people in the middle follow.
We now also dig into the systems that aren't systems: spreadsheets, SharePoint lists and mailbox rules that together keep a service running, even though nobody ever bought them as a system.
A typical assessment surfaces up to 200 opportunities, drawn from a catalogue of 250+ proven use cases and from your own teams' ideas. A few that come up again and again:
Use case | What it does | Example saving a year* |
Inbox management | Shared mailboxes triaged, routed and first replies drafted | £1.1m |
Debt and benefits | Separates can't pay from won't pay, and flags unclaimed benefits | £708k |
AI contact centre | Phone and web enquiries answered 24/7, with a warm handover to staff | £563k |
Safeguarding triage | Referrals summarised and prioritised for the front door team | £442k |
Certificate checks | Gas and safety certificates read, checked and chased in seconds | £293k |
FOI and SAR automation | New requests matched to past disclosures, with first drafts written | £122k |
*Medium-scenario annual estimates from one recent council assessment.
And one nobody saw coming: whose sofa is it anyway? A resident snaps a photo of a dumped sofa. AI pins the location, works out who owns the land - council, highways or private - and sends it to the right team, first time. Today, every report is checked by hand.
Nick finished with two live demos: an illustrative assessment report as an interactive website, where leaders can flex adoption and watch the business case change, and a working prototype that reads a batch of gas safety certificates, flags exceptions and drafts the follow-up letters.
Every number checked - and the choice to proceed is yours
Aneeq closed on how councils usually buy transformation. Consultancies advise but aren't there for delivery, or deploy but aren't accountable for the outcome. Product vendors sell technology but not results. Either way, the council carries the risk. ICS.AI provides the plan, the technology and full deployment, backed by guaranteed transformation savings.
The assessment itself follows seven steps and typically takes 8-16 weeks, depending on pace:
Meet your SLT: a one-hour, in-person walkthrough
Procure: usually via G-Cloud 15, with other frameworks available
Plan: workshops booked with our AI consultancy team
On site: an intensive workshop week and a structured data request
Build: the report and a detailed business case
Validate: your teams verify every number
Present: the validated business case goes back to your SLT
Aneeq singled out step six as the most important. Every figure is checked by service teams and finance before anything goes to SLT or Members, so the result isn't our predictions - it's a business case your council can stand behind.
The assessment is a fixed fee, agreed upfront, and a separate decision from any transformation. There is no obligation to go further. For councils that do, our Transformation Savings Guarantee of up to £5m underwrites year-one savings, sized conservatively for each council (subject to terms).
Watch the full session
The full recording, with chapters for each part of the story, is available on demand – watch the webinar.
Where should your council start?
If any of this sounds familiar - pilots that don't join up, a vision without a costed plan, or savings that need to land this year - let's talk. We'll come and meet your SLT in person, at no cost, and walk through what an assessment would look like for your council.
Sources:
Comments