Executive Summary
Time saved becomes valuable when we decide what it is for.
HK Strategy Advisory, 2026What this paper is
Whenever I hear a confident prediction that AI will allow an organisation to remove a certain percentage of its workforce, I find myself asking a rather ordinary question: what could those people achieve with the time they would get back? This paper argues that the time AI releases is an opening, not an outcome. What an organisation does with it is a leadership decision, and it needs an owner.
After more than four years of using AI, I have considerable confidence in its practical value. It has increased my productivity, extended the range of work I can undertake and saved me money. Those results reflect sustained practice, careful checking and an understanding of when specialist help is still needed.
That experience has made me enthusiastic about AI's potential. It has also made me sceptical of proposals that move too quickly from time saved to people removed. There is often a great deal of valuable work waiting to be done.
Why this matters now
Three recent pieces of evidence bear on this, each on a different question. One looks at what happens when individuals are given an AI tool. One looks at how firms that adopt AI recruit. One looks at UK entry-level hiring. They do not add up to a single finding, but each raises a leadership question before it raises a technology one.
What you will find inside
The difference between a saving and released capacity, the case for growing without recruiting at the previous rate, and why junior roles need protecting. Then four questions that turn released capacity into a result, and three questions every board should answer before declaring an AI initiative a success.
The Evidence at a Glance
| Figure | What it shows | Source |
|---|---|---|
| 7,137 knowledge workers, 66 firms | AI tool users spent less time on email. Researchers did not detect broader changes in the quantity or composition of tasks from individual access alone. | Dillon, Jaffe, Immorlica & Stanton, NBER Working Paper 33795, rev. November 2025 |
| 5,172 customer-support agents | AI assistance raised productivity by 15% on average, with larger gains for less experienced workers, and evidence that it supported learning. | Brynjolfsson, Li & Raymond, Quarterly Journal of Economics, May 2025 |
| Firms adopting AI | Hired fewer people than comparable firms that did not use AI. No clear effect on overall Danish employment at that stage. | Danmarks Nationalbank, 16 September 2026 |
| UK entry-level hiring | Sharp declines in several entry-level professional occupations, alongside a weaker wider market. Causation by AI not established. | DSIT and LinkedIn, 8 June 2026 |
The Work We Have Never Had Time to Do
Most organisations contain capable people whose days are crowded with administration: transferring information between systems, preparing routine reports, reconciling records and documenting work they have already completed.
Some of this is essential. Some reflects processes that have accumulated layers over the years, each individually defensible and collectively exhausting.
People can be working very hard while their abilities are poorly used.
When AI releases some of that capacity, the opportunity extends beyond completing the existing workload more cheaply. Organisations can consider work they have repeatedly postponed, undertaken inadequately or never felt able to afford. Customer relationships need attention. Services could be improved. Opportunities remain unexplored. Managers need time to think, and younger colleagues need someone available to teach them.
A business case that counts only the cost of today's tasks may overlook much of tomorrow's value.
Where released capacity could go
Consider a GP. If technology can safely reduce the administrative burden, the benefit might appear in more appointments, longer consultations or greater attention to patients with complex needs. The capacity released has somewhere useful to go.
I see a similar opportunity through my work as a trustee of a charity helping people into employment. Understanding a client's circumstances, building confidence and developing relationships with employers all take time. Administration competes for the same hours. If AI can ease that burden, I want us to examine how much more our people could accomplish: more clients supported, more sustained attention where the barriers are greatest, and stronger relationships with employers.
Productivity acquires a human purpose when we can explain whom it will help.
A Saving Is Only Part of the Dividend
There is a useful distinction between reducing expenditure and releasing capacity.
| Measure | Reducing expenditure | Releasing capacity |
|---|---|---|
| What happens | AI allows work that was previously bought in to be done internally. A saving may follow directly. | An employee completes a task in less time. The organisation gains capacity. |
| What it needs | A comparison of cost before and after, including implementation, checking and correction. | A leadership decision about how to use it, based on demand, priorities and the constraints on growth. |
| Main risk | Counting the saving before the checking and correction costs are known. | Time released from one administrative burden quietly disappearing into another. |
Saving five hours means little if nobody has considered what should happen in those hours.
What the research shows
A randomised study across 66 firms and 7,137 knowledge workers found that users of an AI tool spent less time on email, but the researchers did not detect broader changes in the quantity or composition of tasks from individual access alone.[1]
My reading is that providing the tool leaves a substantial management job unfinished. Work needs to be reconsidered, priorities clarified and responsibility assigned. Otherwise, time released from one administrative burden can quietly disappear into another.
Growing Without Recruiting at the Same Rate
Greater capacity can allow an organisation to grow without adding staff as quickly as before. That can produce a meaningful financial benefit while retaining the knowledge held by existing employees. Experienced people understand customers, relationships, recurring problems and the exceptions that rarely find their way into a manual.
An organisation's knowledge is larger than its documentation.
A September 2026 analysis from Denmark's central bank found that firms adopting AI subsequently hired fewer people than comparable firms that did not use it. The bank reported no clear effect on overall Danish employment at that stage.[2]
This is consistent with an adjustment through slower recruitment. It does not settle the wider employment question or demonstrate that every existing role should remain unchanged. It does, however, reinforce the importance of examining hiring patterns alongside redundancies.
An organisation may be able to serve more customers, improve its margins and retain valuable colleagues without expanding its workforce at the previous rate. That is a credible ambition. It also carries a responsibility.
Keep a Way In
If recruitment slows, organisations still need to consider who is being brought in and how future expertise will develop.
Junior roles serve an immediate purpose and a longer one. They produce work today while giving people somewhere to acquire experience. Through that experience, people learn to recognise an implausible answer, trace a mistake, handle an exception and judge when to ask for help. These abilities become particularly valuable when AI can produce fluent, convincing output at speed.
UK hiring data shows sharp declines in several entry-level professional occupations. The evidence requires care: junior hiring has broadly weakened alongside the wider market, and the figures do not establish that AI caused the falls.[3]
The leadership question remains valid without claiming that causation has been proved. If we recruit fewer beginners, where will our future experts come from?
The next generation of expertise needs a place on today's payroll.
There is a tension here that deserves attention. AI has saved me money and reduced my reliance on some external services. An organisation that ceases to develop its own people could, over time, become more dependent on its technology and less able to assess its work. The capacity to challenge a system has to be developed somewhere. Retaining experienced colleagues matters; creating opportunities for others to learn from them matters too.
Give Apprenticeship More Room
AI could help organisations improve the way people learn. A study of 5,172 customer-support agents found that AI assistance increased productivity by 15% on average, with larger gains for less experienced workers. The researchers also found evidence that the assistance supported learning.[4]
Those findings come from a particular setting. They nevertheless offer a constructive possibility: AI can help newcomers become capable more quickly. Employers can build on that possibility through supervised work, deliberate checking of answers, exposure to clients and responsibility that grows with competence.
Experienced colleagues may also have more time to teach if some of their administrative burden is removed. A portion of the time saved could become time spent explaining a difficult decision, reviewing an error or helping a junior colleague understand why an apparently sensible answer is wrong.
An organisation might therefore slow its overall recruitment while maintaining a deliberate intake of junior talent and giving those recruits a stronger start. Some of the AI dividend should help build the people on whom the organisation's future depends.
Governing the Dividend
Released capacity does not turn itself into a result. Where nobody has decided what it is for, it tends to be absorbed by whatever is most urgent. Four questions turn it into a decision, and each needs an answer in writing.
| Question | What a good answer looks like |
|---|---|
| Who decides? | A named senior executive owns delivery: the choice of how released capacity is used and the work to achieve it. |
| What benefit is intended? | A specific outcome, stated in advance: more clients supported, faster growth with fewer hires, a junior intake protected. Not simply "efficiency". |
| How will it be measured? | Net of implementation, checking and correction, and set against the intended benefit as well as the cost saved. |
| When will it be reviewed? | A date, set before the decision is taken, at which the board sees the result and can change course. |
The responsibilities are different. The executive owns delivery. The board oversees whether the intended benefit is the right one, what it means for the workforce, and what was actually achieved. This is the Accountable and Responsible thinking of the HK Strategy Advisory CLEAR Framework applied to the benefits of AI as well as its risks. Where the outcome affects roles, hiring or how staff are treated, the Lawful and Ethical dimensions apply too: take appropriate employment and equality advice, and be open with staff about how released time will be used.
Decide What the Dividend Is For
There will be circumstances in which AI reduces the number of roles an organisation requires. That deserves an honest assessment, informed by its prospects and obligations. My objection is to treating workforce reduction as the natural first measure of success.
The assessment should include direct savings, unmet demand, service quality, opportunities for growth, retained knowledge and the development of future capability. Before declaring an AI initiative successful, I would want leaders to answer three questions:
What time and expenditure have we actually saved, after allowing for implementation, checking and correction?
What valuable work will the capacity released now allow us to undertake?
How will we continue developing the people whose judgement we will need in future?
My own experience has persuaded me that AI can deliver substantial practical and financial benefits. The opportunity is considerable. The quality of the outcome will depend on the choices we make with it.
Time saved becomes valuable when we decide what it is for.
Three Things to Do This Quarter
Separate ownership from oversight
Name a senior executive to own delivery of the AI dividend, and put its intended benefit, workforce implications and results on the board's agenda.
Set the measure and the review date
Require any AI business case to show the intended benefit and the net saving after implementation, checking and correction, with a date for the board to review it.
Protect a route in
Decide deliberately how many junior colleagues you will bring in, and how much experienced time will be set aside to teach them.
Work With HK Strategy Advisory
If this paper has raised questions about how your organisation governs AI, including what it does with the benefits, we can help. HK Strategy Advisory works with SME boards, charities, founder-led businesses and leadership teams to build practical, credible AI governance frameworks, from initial audit to full board-level implementation. Contact us to arrange an AI Governance Review.
This paper is provided for general informational and educational purposes. It does not constitute legal, regulatory, or professional advice.
Sources
- 1Eleanor W. Dillon, Sonia Jaffe, Nicole Immorlica and Christopher T. Stanton, Shifting Work Patterns with Generative AI, NBER Working Paper 33795, revised November 2025. ↩
- 2Danmarks Nationalbank, Firms hire fewer people when they start using AI, 16 September 2026. ↩
- 3Department for Science, Innovation and Technology and LinkedIn, A snapshot of entry-level hiring in the UK, 8 June 2026. ↩
- 4Erik Brynjolfsson, Danielle Li and Lindsey R. Raymond, Generative AI at Work, The Quarterly Journal of Economics, May 2025. ↩