Inside an AI assessment: five days on site, every service, every level
By Nick Horne, AI Enablement Consultant, ICS.AI
"What actually happens in an AI assessment?". It's one of the questions I get asked most. Usually by someone who has sat through a few too many AI presentations and wants to know what they’d be signing their teams up for.
The short answer: a week on site with people who run your services and the people who do the work. I've been on site for every SMART: AI Assessment we’ve delivered, so here’s what that looks like.
The numbers so far
Across 12 assessments in UK local government and higher education, we’ve found more than 1,450 AI opportunities and £63m to £98m in identified savings. Identified, not banked. More on that later.
Each organisation typically gets up to 200 opportunities. Some come from our catalogue of 250+ proven use cases. Plenty come from staff who bring their own ideas. It takes 40+ hours of workshops, and that’s where the real work happens.
A week on site
We’re usually with you for five to seven days, running 90 to 120 minute workshops across customer services, finance and resources, HR and people, housing, planning, social care, environmental services, and corporate and leadership.
An AI assessment is a whole-council exercise, not an IT project. What I’ve found is that the best opportunities often turn up where nobody expected, and the biggest savings come from processes that cross several services.
We also send a structured data request, so the business case is built on your numbers, not sector averages.
Who's in the room
We’re deliberate about who comes. We want five perspectives:
Senior leaders, for the strategy and priorities.
Managers, for how the service really runs.
The doers, who do the work every day and know exactly where the time goes.
Finance, so the numbers stand up from the start.
Technical leads, so every idea works within the systems you’ve actually got.
Put those five groups in a room and the conversation changes. A frontline officer mentions a weekly workaround. Their manager works out how many hours it eats. Finance puts a number on it. The technical lead knows which system it touches. One conversation, and the opportunity is found, costed and sense-checked.
A happy side effect
Someone describes a problem they've been stuck with for months. Someone else in the room says they sorted the same thing last year. Same organisation, sometimes the same building, but they've never met, or haven't spoken in ages. Neither had any idea the other could help.
It's not something we plan for, and it won't show up in the business case. But it happens often enough that we've come to look forward to it. Sometimes the fix is already in the building, and it just takes the right people being in the same room to find it.
How ready are you?
We also look at how ready you are to make AI work, across five areas:
Strategy and leadership: a clear direction and someone senior backing it.
Technology and data: your systems, your data and how they join up.
Service and use case: where the most practical places to start are.
People, skills and process: capability and appetite for change.
Ethics and governance: using AI safely and responsibly.
Readiness isn't a pass or fail. It shows what needs to be in place and sets the order of the roadmap. In practice, a big opportunity in a service that isn’t ready goes later. A smaller one in a service that is goes first.
Sceptics and champions
Every workshop has both, and we’ve learned to welcome them.
The sceptics ask the questions you’d want asked. "Will it undermine professional judgement?" "Is our data really safe?" "Isn't this one more thing for stretched teams?" Leave them unanswered and they’ll stop adoption later. So we turn them into design rules: governance, safeguards and a human in the loop where it matters.
The champions turn up with a different energy. "I've already used Copilot to do this." "We came to the session with ideas ready." "Let's scale what's already working." They become your early wins, and the start of an AI Champions network that helps colleagues get going.
The bit that matters is having both in the room. Sceptics tell us what will stop adoption. Champions show what's possible. Both shape the plan. As I said in our recent webinar: “If we can please both of those groups, then we’re definitely going to nail the people in the middle”.
That’s what turns an assessment into savings. It’s not the report. It’s the people who own it afterwards.
Who's with you
We keep the load off your teams. You get one team from kick-off to final readout, including former senior local government leaders and sector specialists where needed. We also run hands-on sessions with real AI tools and help with comms to staff, Members and leadership.
After the week on site
The workshops are the visible bit, not the end. After we leave:
We write the report and build a detailed business case from your data.
Your teams check every number, service by service, before anything goes to SLT. This is the step that matters most - it's how identified savings start to become ones you can bank.
We show a working demo of the kind of use case you've surfaced.
In our recent webinar I showed an illustrative one: a tool that reads gas safety certificates, checks them against the property register, flags exceptions and drafts the follow-up letters. A familiar problem solved on screen does more than any slide.
From first conversation to validated business case usually takes 8 to 16 weeks, depending on pace. You end up with a plan you've built, tested and own, and what happens next is your call.
If you’re planning one
My advice, whoever you work with: get the people who do the work and someone from finance involved from day one. They make the numbers real.
If you'd like to know what a week like this would look like in your council, we'll come and talk it through with your SLT in person.
Blog Series:
Read the previous blog in the series: What 12 assessments have taught us about getting AI right.
Look out for my next blog "The systems that aren't systems - and whose sofa is it anyway?".
Comments