Blog: Workforce
Is your AI business case really a headcount plan?
Published by Scott Seaborn, Founder
An AI pilot in the order office saves each person an hour a day. By the time the business case reaches the board, that hour has become two fewer roles.
Those are different claims. The first is a measurement. The second is a forecast about work nobody has yet checked.
A quarter of employers using AI expect fewer staff
On 17 November 2026, the CIPD published its Autumn 2026 Labour Market Outlook: a YouGov survey of 2,051 senior HR professionals and decision-makers in the UK, carried out between 17 September and 11 October 2026.
Among employers using AI, 25% expect it to reduce headcount over the next 12 months, up from 21% a year earlier. Among those with deeply integrated AI, the figure is 44%, although 17% of that group also expect AI to increase headcount. Meanwhile 40% of employers describe their use as ad hoc or experimental, and 10% of AI users say it has improved job performance a lot.
This is evidence of what employers expect, not proof that AI has removed jobs or that planned reductions will deliver the savings assumed. Read the research.
The FOFO™ view
A headcount number is the easiest benefit to write into a business case and one of the hardest to verify in advance. Time saved on a task does not automatically become a role the business no longer needs, as we set out in what happens to the time AI saves.
Imagine a manufacturer that uses AI to draft quotations. Each estimator now produces a first draft in minutes. Someone still checks material prices, clarifies non-standard specifications and handles the customer who changes the order twice. Remove one estimator and that work lands on the others, or on sales.
The people who know where the work really goes are rarely in the room when the case is written: the estimators themselves, the team leaders who cover absence and the customer service staff who pick up the exceptions.
Test the saving before you bank it
Before a headcount reduction goes into an AI business case, check that you can answer:
- Which tasks stop, and how many hours did they take last quarter?
- What new work appears: checking, correction, exceptions and queries?
- Who covers peak demand, absence and the difficult cases?
- What knowledge leaves with the role, and where is it written down?
Then measure output per person, rework, cycle time, overtime and agency spend for three months before and after the change. If those figures do not move, the saving is not yet real.
AI Compass™ gathers that insight from the people doing the work through AI-powered surveys, branded for your organisation and sent by you. Slid3rs™ helps your people rank the resulting opportunities by business value and delivery complexity, and FlowRoom™ brings leadership together to agree which to fund.
Decide what the work needs first, and the headcount question becomes much easier to answer.
Related reading
Next