What happened when a six-person NHS system workforce team was given nine months to build AI into the way it actually works.
Six people|Nine months, Sept 2025 – June 2026|Three days plus 1:1 coaching|Two surveys, the same questions before and after
The System Workforce PMO serves a partnership running from care homes of twenty staff to NHS organisations of four thousand. At the baseline it rated its belief that AI could improve its work at 4.6 out of 5 and its confidence in using AI at 2.4 — and costed 105 hours* a month of friction it wanted gone.
What it delivered
Hours handed back every month
57–60*
At an eight-hour day: 7 working days a month, 85 days a year — about 0.4 of a full-time post back to a six-person team.
Team AI confidence, self-rated out of 10
3.75→6.75
Every needle moved. The share of the working week AI supports rose 9% to 45%.
“More face-to-face work with the team and partners without having to worry about admin. The team is building credibility.”
The System Workforce PMO — endpoint survey, June 2026
The team's own verdict
4.5/5
Would recommend it to other NHS teams
4.5/5
Supports the NHS system's long-term goals
~40%*
The team's estimate of its productivity gain
Growing Beyond — August 2026 · * the team's own estimates
1 · What it was worth
The team costed its own friction. This is what AI paid back.
105h*
Monthly friction, as the team costed it
57–60h*
Handed back every month at the endpoint
~680h*
The same saving across a year
£783*
Of that saving, the part given a £ value
Where the hours came back
14h
Meetings — minutes, agendas, actions
34h
Reporting & data — amalgamating, summarising
9–12h
Comms & docs — 1:1s, PDPs, frameworks
85 days
a year — roughly 0.4 of a full-time post
How much more productive is the team?
Each team member's own estimate of the team's overall gain. Separately, every session attendee rated their own productivity two points higher than at the close of Day 1 — a team average of 6.3 → 8.3.
What this means
A small team serving a whole partnership got back the equivalent of four months of one person's working year, and put it into partner engagement and face-to-face work.
Every hours and money figure is a team member's own estimate — baseline five people, July–August 2025; savings four people, June–July 2026, so this is not a like-for-like subtraction. “More than twice as productive” is open-ended, so the average counts it at its floor of 100%.
A fourth baseline theme — workflow and planning, 20 hrs a month — has no matched endpoint figure and adds nothing to the saving. It is the project the team has taken on itself.
2 · What changed
Every confidence speedometer moved — and the lower the start, the higher it climbed
Self-rated AI confidence, 1–10, one speedometer per team member. Team average 3.75 → 6.75. The largest shift, 1 → 7, belongs to the person who arrived with the most reservations.
What made the adoption stick
Fourteen questions tracked how the team experiences AI. Every one rose or held at its ceiling, and the three largest jumps were all about AI becoming visible — which is how capability spreads on its own.
2.75→5.00
Seeing leaders use AI
Now unanimous across the team.
2.25→4.25
Using a colleague's workflow
One person's solution, picked up by others.
3.00→4.75
Sharing tools and tips
Traded as routine, in the flow of the work.
The person who began most stretched
Control of time and workload
2/5→5/5
Alongside the largest rise in AI-supported workload, 10% to 75%.
Stress managing it — lower is better
4/5→1/5
Capacity released, not capacity consumed.
Where the whole team landed
14 of 14
Culture and capability measures that rose or held at their ceiling
1 → 5
Team members running four or five AI-supported workflows
3 of 4
Have demonstrated an AI workflow to the team, up from one
3 · What it was
Two immersion days, eight months of live practice, one day to consolidate
Days 1 & 2 · Sept–Oct 2025
Build awareness, capture opportunities
Keep / Stop / Start — where the work hurts
Benefits and risks, on the evidence
Utley's 2×2: how well AI is used against how valuable the work is
Live experiments on the tools the team could reach
Oct 2025 – May 2026
Team-led practice and 1:1 coaching
Practice on real workflows, run by the team
“Show, don't tell” — leaders demonstrating
Three team members took a 1:1 coaching session
Workflows passed between colleagues
Day 3 · 8 June 2026
Review, embed, plan forward
Held as a Thinking Environment — equal turns, individual AI stories, a shared vision
Lumina appreciation: what each person brings
Team WOOP: direction and action planning
Six team-voted AI Empowerment Awards
4 · What the team said
Faster and better, together
“It has enabled me to become more thorough and focused in my role — allowing my output pace to increase whilst also increasing quality.”
From “cheating” to complementing
“I feel more comfortable using Copilot — not viewing it as cheating but rather complementing my work and making me more effective with my time.”
It travelled beyond the team
“They borrowed my survey and review model… they use that now across the West Yorkshire compassionate leadership review.”
“We all got so much out of the programme. You made it relevant to our world of work.”
Dominic Blaydon — System Workforce PMO
Programme outcome
Nine months on, the System Workforce PMO runs its meetings, reports, plans and communications with AI woven into the work — visibly, safely, and in its own voice. It is left with less admin between the team and the partners it serves, and an AI adoption journey it now carries forward on its own.
