A task can shrink while the job around it gets bigger.
My job has recently got bigger.
It changed from Digital Strategy Director at one NHS Trust to Group Associate Director for Innovation and Partnerships, working across two NHS Trusts. I retained responsibility for digital maturity and skills, while my reach widened into innovation, partnerships and commercial management. The territory grew, the governance certainly grew, and so did the potential consequences of a good or bad judgement.
AI did not cause that change. But it has made me notice something about the way jobs change. They do not always get bigger in one direction. A role can retain some of what it was while absorbing new territory. Responsibility, influence, capability, governance and reward do not necessarily move together, or at the same speed.
That becomes particularly interesting when technology promises to give us some of our time back.
Who gets the saved time?
The appeal is obvious. Something that took an hour now takes twenty minutes. Forty minutes have been released. But released for what?
A large Danish study surveyed around 25,000 workers across 11 occupations after generative AI tools entered their work and linked their answers to payroll records. The workers who used them did save time, yet eight in ten said that time went back into other work.
Some did more of the same work. Some did different tasks. New work appeared too, and much of it involved making the technology itself work inside the organisation: checking outputs, integrating systems, reviewing quality, dealing with compliance and deciding how the tool should be used.
A task can disappear without the job getting smaller.
That is not necessarily bad. If I can spend less time producing something routine and more time thinking about something difficult, that may be a very good trade. An employer investing in technology can reasonably expect some benefit from it. A public service under pressure may use released capacity to help more people.
But saving time and deciding what the time becomes are different events. Often there is no meeting at which somebody thoughtfully allocates the extra forty minutes. The queue does it. The inbox does it. The next patient, deadline, request or problem arrives and fills the space.
Sometimes more than time moves
The strongest example I found is not generative AI at all.
In a cluster-randomised trial across 103 rural villages in Lesotho, researchers studied what happened when lay community health workers were given mobile clinical decision support alongside training, protocols and supervision. Rather than referring every patient with uncontrolled hypertension onwards to a health facility, the workers could initiate and adjust a defined first-line treatment themselves.
After 12 months, 58 per cent of patients in the intervention group had their blood pressure controlled, compared with 48 per cent under the usual referral pathway. There was no relevant difference in safety outcomes.
For someone living a long way from a clinic, this is not mainly a story about productivity. More capability had moved towards where the patient already was.
I have started thinking about this as reach: how much of the problem, process or consequence you can touch from where you stand. Reach does not have to mean promotion or seniority. A technician may gain access to expertise that previously sat with a specialist. A frontline professional may safely resolve something that used to require escalation. A corporate role may stretch across functions, services or organisations that were previously separate.
Technology can change where capability is available.
But the Lesotho example contains an important warning too. The workers were not simply handed a tablet and told to do more. The system around the work changed with them. The treatment they could provide was bounded; there was training, clinical rules sat inside the decision-support process, supervision remained, and more complicated cases still had somewhere to go.
The reach expanded because capability, authority and safeguards moved with it. That matters far beyond healthcare. If technology enables someone to reach further into a process, their knowledge may need to grow. Their confidence may need to. Responsibility and governance often will. Sometimes authority, support or recognition may need to change as well. Otherwise greater reach can become something less attractive: greater load.
The same technology can become a different condition
Large organisations may have one strategy, one technology and one set of policies, but they contain many local working environments.
Official UK evaluations of the same Copilot technology have already produced different measured answers. A cross-government experiment reported 26 minutes a day of self-reported time saving; DWP estimated 19 minutes a day; HMRC’s more conservative trial put the gain closer to an hour a week; and the Department for Business and Trade found no robust evidence that reported time savings had yet translated into higher productivity.
That is less a contradiction than a clue. The tool may be the same. The work, people, measures and conditions around it are not. The same licence can arrive in two teams and become two very different conditions: in one, people have a problem worth solving, confidence to experiment and colleagues who share what works; in another, uncertainty about permission, relevance or risk can leave the same capability largely untouched.
The organisation creates conditions. Teams make many of those conditions real.
James Freed FBCS, Deputy Director at the NHS Digital Academy within NHS England, makes a related argument through BCS’s Insight Exchange: technological capability can move faster than the organisational and cultural changes needed to absorb it.
That gap matters because technology adoption is also a human-development problem. People have different starting skills, confidence, professional responsibilities and reasons to use a new tool.
In our own work, we are beginning to use the SFIA Foundation’s skills framework beyond traditional digital roles to understand some of the digital capability increasingly required inside clinical and other roles, starting with Allied Health Professionals. The intention is not to turn an AHP into a digital professional or replace the capability frameworks of their profession. SFIA can sit alongside them, helping make the digital dimension of a changing role more visible.
Someone can be confident without yet being capable. They can be capable but work in a role that gives them little opportunity to use it. Or the role itself can begin to require capabilities that its old description barely recognised. Those are different problems.
What changes after the licence arrives?
Because our team is small, we are approaching some of this through pull rather than trying to impose the same capability programme everywhere at once. Teams can bring forward problems that matter to them, and support can form around real work.
There is a weakness in pure pull, though. The people already confident enough to ask may arrive first, while those who feel least capable may never knock on the door. So the answer is probably not push or pull. Make access, permission, support and safe boundaries widely available, then let meaningful use grow around problems people actually recognise, while paying attention to the teams that never come forward.
Our RADAR research lab is also supporting the orchestration and benefit-realisation approach around Copilot across the two Trusts, from corporate to frontline work. I am increasingly interested in what we will be able to see over time: not simply how many licences were used or how many minutes somebody says they saved, but whether confidence changed, what people can now do that they could not do before, which tasks disappeared or appeared, whether something previously handed elsewhere can now be resolved locally, and whether checking and governance grew with it.
Eventually an even more interesting question appears: has the work changed enough that our description of the role no longer quite fits?
A capability model may be more useful as a moving picture than a photograph.
This problem is much older than AI. Early factories were organised around the physical demands of steam power: shafts, belts and machinery had to sit in arrangements that made sense for the source of power available at the time. Electric motors eventually made different layouts possible.
The old factory was not foolish. It had been organised sensibly around yesterday’s capability. Changing the source of power did not automatically redesign everything around it. We may be living through another version of that problem now: a tool can change remarkably quickly while roles, workflows, confidence, professional boundaries, governance, measures and organisations move at different speeds.
Where can the work reach now?
Which brings me back to my own job. Moving from Digital Strategy Director to Group Associate Director for Innovation and Partnerships did not simply give me a bigger version of the same role. Some capabilities travelled with me. Others became newly important. The territory widened across organisations and into innovation, partnerships and commercial work. The reach of my judgement widened with it.
AI may make some tasks smaller. It may release time. It may bring expertise closer to the person who needs it. It may allow someone at the frontline to resolve more, or enable one person to work across territory that previously required several hand-offs. Those can be enormous gains.
But if our work can reach further, the more important question may be what needs to develop with it.
Where can your work reach now that it could not before — and are you, your role and the conditions around you developing with it?
Sources
• Government Digital Service — Microsoft 365 Copilot Experiment: Cross-Government Findings Report
• Department for Work and Pensions — An evaluation of DWP’s Microsoft 365 Copilot trial
• HM Revenue & Customs — Evaluating the impact of Microsoft Copilot in HMRC: Phase 3 trial
• Department for Business and Trade — Microsoft 365 Copilot evaluation
• SFIA Foundation — Getting started with SFIA skills profiling across teams
• SFIA Foundation — Extending SFIA support to clinical roles in digital healthcare