Why staff need to see the AI business case: lessons from a hospital consultation
· 5 min read
A hospital AI consultation shows why staff must see the business case. Five steps for leaders to test AI plans with the people who do the work.
In brief
A hospital group in Northamptonshire wants to bring in two AI systems after a trial it describes as successful. One would help doctors write up clinic notes and letters, and the other would handle appointment bookings by phone. Local press reports say 50 to 100 administrative and secretarial roles could go, though the numbers are not confirmed, and the trust says it will avoid redundancies by using a mutually agreed resignation scheme. Unison, which says it is not against new technology, has raised a formal dispute because it says the business case behind the plan has not been shared with it.
This article does not take sides. It looks at a pattern that many organisations will recognise: an AI case built and approved near the top, and first seen by the people it affects when their jobs are already in question. Drawing on service management principles from John Sansbury's book The Journey to Service Excellence, Esm argues that frontline staff are the best people to test such a case, because they know where the real process breaks and which exceptions never appear on paper.
The article sets out five practical steps for leaders: share the working behind the decision, map the work with the people who do it, say plainly what changes for each role, name the person who owns the outcome, and plan for what happens when the system gets something wrong. It explains where leadership training helps, and it is clear about the limits, including the need for proper HR, legal and clinical safety advice.
A question every AI programme should expect
"If the case for these changes is so strong, why won't the trust show us the evidence?"
That question comes from Nadeem Samari, a regional organiser for Unison in the East Midlands. It is about University Hospitals of Northamptonshire, the group that runs Northampton General and Kettering General hospitals. But it could be asked inside almost any organisation that is moving from an AI pilot to a decision about people's jobs.
What is happening in Northamptonshire
According to reporting in the Northampton Chronicle and the Northamptonshire Telegraph on 4 and 5 November 2026, the hospital group completed an AI trial earlier this year that it describes as successful. It now wants to introduce two systems. One would help doctors write up consultation notes and send out records and letters. The other would manage appointment bookings, using AI to call patients and take their calls.
The reports say between 50 and 100 administrative and secretarial roles could go, although firm figures have not been confirmed. The trust says it will reduce jobs in a "controlled and compassionate way", avoiding redundancies through a mutually agreed resignation scheme, with a 45-day consultation that began in September and is due to finish at the end of November.
Unison says it is not opposed to technology or change. It raised a formal trade dispute on 28 October because, it says, it has repeatedly asked for the financial, workforce and business case information behind the proposals and has not received it. The trust's chief executive has agreed to meet Unison representatives, and the union has called an urgent members' meeting for today, 9 November. A local branch secretary described members as "very anxious, especially with Christmas on the way."
We only have the press account. We have not seen the trust's case, and this article takes no side in the dispute. The interest for leaders elsewhere is the shape of it.
The pattern behind the dispute
Most AI programmes now have a document like this one somewhere. A trial. Some numbers. A recommendation. It is usually read and approved near the top.
The people whose work it describes are often the last to see it, if they see it at all. By then the question is no longer "is this a good idea?" but "what happens to me?" Scepticism that could have improved the plan turns into resistance to it.
Other research this weekend points the same way. A Censuswide survey for Indeed in Italy, reported by Il Sole 24 Ore on 8 November, found 37% of employees had used AI at work without telling colleagues or their manager. G-P's 2026 World at Work report (vendor research released on 21 October) found 58% of employees across six countries, including the UK, would turn down a job if an employer's AI values did not match their own. People are watching how AI is introduced, not only whether it is.
What service management thinking adds
John Sansbury's The Journey to Service Excellence is a practical guide to service management, not a book about AI. But one of its core principles applies directly. He writes that "it is almost always the frontline staff, i.e., those in a consumer-facing role, that know best how to improve the efficiency and effectiveness of the services that you provide."
He also redefines the job of management: "the primary role of the manager or leader is to use their authority to remove the obstacles to their teams’ high performance."
Esm applies those principles to AI. The people who book appointments and type clinic letters know where the process breaks: the patient who needs a call back in another language, the letter that has to go to two GPs, the booking that only works because someone remembered. A business case that has not been tested against that knowledge is a guess with a spreadsheet attached. The frontline is not an audience for the case. It is the best test of it.
Practical steps for leaders
- Share the working, not just the conclusion. Say what the trial measured, over what period, and who checked the results. If something cannot be shared, say why.
- Map the work before you map the roles. Walk through the real process with the people who do it, including the exceptions and the workarounds that never appear in a job description.
- Say plainly what changes for each role. Which tasks move to the system, which stay with people, and which new tasks appear, such as checking, correcting and handling the cases the system cannot.
- Decide who owns the outcome. Name the person accountable for how the service performs for patients or customers after the change, not just for delivering the project.
- Plan for the failure, not just the success. Agree what happens when the system gets something wrong, how a person can step in, and how you will know.
Where training fits
The gap here is rarely technical. It is leadership capability: knowing how to lead people through a change they did not ask for, how to run a consultation that genuinely tests a plan, and how to apply service management basics such as clear ownership, measuring outcomes rather than activity, and building improvement on the knowledge of the people doing the work. Those are learnable skills, and they matter more as AI moves from pilots into roles.
Honest limits
This is based on local press reporting, and firm job numbers have not been confirmed. Consultation duties, NHS employment terms and resignation schemes are matters for HR and legal advice, which this article does not give. AI used in patient bookings and clinical documentation also needs proper clinical safety and data protection assessment by specialists. Sharing a business case will not remove every concern, and some decisions will still be hard.
A question to take away
If the people whose work your AI case describes read it tomorrow, what would they spot that you missed?
Tags: ai adoption, leadership, change management, business case, frontline staff, public sector
Sources
Join the Exchange free
Join free to unlock the complete knowledge library.
Founder membership is free for life
More from Esm
How should the service desk handle a Microsoft 365 outage it cannot fix?
4 min read
Deploying AI is the easy part: who runs it once it goes live?
5 min read
How do I compare help desk tool pricing before rolling it out across several departments?
4 min read
Turning 'not helpful' into a fix
1 min read
When the author leaves, who owns the article?
2 min read
Test after every change, or find out from users
1 min read
Related articles
How should the service desk handle a Microsoft 365 outage it cannot fix?
Treat a supplier cloud outage as your own major incident: log it once, tell people early on a channel that still works, update on a rhythm and review your dependency afterwards.
Deploying AI is the easy part: who runs it once it goes live?
Only a third of large enterprises have met the business case for their main AI programme. New research points to a gap in who runs AI after launch, and what leaders can do.
How do I compare help desk tool pricing before rolling it out across several departments?
Compare help desk tools on a full year's cost for your real agents, approvers and requesters in every department, not on the headline monthly price.
Turning 'not helpful' into a fix
'Not helpful' is free, honest feedback, and most of it goes nowhere. Turn that click into a fix.
When the author leaves, who owns the article?
Articles keep being read after their authors move on. Make sure the care for them does not leave too.
Test after every change, or find out from users
Knowledge changes fail quietly. A short, repeatable test after every change finds the problem before your users do.