Every office pipeline, fit-out programme and lease model in the UK rests on one multiplication: people times square feet per person. Antony Slumbers, a writer and consultant on real estate and technology, used his closing keynote at CoreNet Global UK's Future Focus in London on 15 October to argue that the first number in that sum has stopped behaving. His 20 October write-up puts arithmetic behind the argument.
Last week we covered his earlier piece, 'Show me the incentive', on who keeps the gain when AI lifts a landlord's operating income. This one asks a question with sharper consequences for anyone who builds or fits out offices. If output no longer tracks headcount, what does an occupier's space requirement track instead?
One productivity gain, three opposite answers
Slumbers' worked example is simple on purpose. Take a firm of 500 people in 60,000 sq ft, held at a constant 120 sq ft a head. Give it 40% more output per person over five years, and change nothing except how fast the business grows.
If output stands still, the firm needs about 360 people and 29% less space.
If output grows 10%, it needs about 390 people and roughly a fifth less space.
If output doubles, it needs about 710 people and 43% more space.
He is candid about the method: "The napkin maths is deliberately crude; the spread is the point." The next sentence is the one to take into your next feasibility meeting
With a 40% productivity gain, a firm has to grow its output by 40% just to stand still on space.
Growth, in his view, mostly comes from the sector the occupier sits in. Two tenants in the same building with the same AI tools can head in opposite directions, and a single headcount forecast cannot show that.
Why the market data will not warn you
Slumbers expects national take-up and vacancy figures to look reasonably calm for a while, because the change happens inside buildings before it shows up between them. Growth that would once have filled another floor in London lands in a capability centre in Bengaluru or in software instead. He calls this missing absorption.
Vacancy tells you what left. It cannot tell you what never arrived.
The headline numbers bear him out so far. Savills reported on 5 May that 2.2m sq ft of office leases transacted in central London in the first quarter of 2026, up 6% on the same period last year, with active demand at a record 14.6m sq ft and City rents at a new record of £160 per sq ft. None of that says anything about what is happening on floor seven of a particular building, which is Slumbers' point.
Work has a postcode
The second half of the keynote is about where each kind of work settles. Slumbers sorts tasks by whether they are 'plumbing' (moving, checking and reconciling information) or 'cognition' (interpreting and deciding) and by how cheaply the output can be checked. Verifiable processing goes to software. Exception handling goes to hubs. Hard-to-verify strategic work stays in a smaller HQ, used intermittently for clients and for training the next generation.
UK cities start from very different positions. Using ONS workplace data pooled over the three years to March 2026, Slumbers finds the City of London has about six managerial and business-professional jobs for every administrative one, against fewer than two in most big regional cities.
He adds the caveats himself: these are survey estimates, occupation is a proxy, and public sector administrative work is more likely to take AI gains as extra output than as fewer seats. His reading is that the regions are more exposed to processing and London to verifiable professional work such as junior accountancy and routine legal tasks.
The hubs taking that work are already large: Zinnov and Nasscom count 2,117 global capability centres in India as of March 2026, employing 2.36 million people.
What a developer, fit-out contractor or cost consultant does with this
Slumbers is careful about causation. His two building case studies, BT at Three Snowhill in Birmingham and Meta's former floors at Rathbone Square in London, led him to a plain conclusion: "In neither building did AI cause what is happening." Hybrid working and pre-AI cost cutting emptied the floors. AI is the amplifier that makes the next round deeper.
For a developer or asset manager weighing refurbishment capital, his advice is blunt: "Separate a shrinking requirement from a building-quality problem before committing refurbishment money. A better building will not refill floors whose work has moved to Bengaluru."
For a fit-out contractor, the shift he describes is in the brief itself: "Fewer rows of production desks. More rooms where people review, argue and decide."
For a cost consultant or project manager, the lesson is in the three scenarios. A single headcount assumption at Stage 1 is now a hidden bet on the client's sector growth rate, and it should be presented as one.
Takeaway
Slumbers is clear that the strong version of his claim, that work-level understanding of occupiers becomes the decisive input to portfolio strategy, is a hypothesis he expects to be judged on over three to five years. The weaker version is usable now. If your feasibility model or fit-out brief carries one headcount number, you are assuming the occupier's output grows at exactly the rate its productivity does. Nobody knows that. Replace the number with a range, and ask which kind of work the floor is for.
Run occupier space requirements as three scenarios (flat, modest and strong output growth) and show the client the spread.
Before committing to refurbishment or fit-out money, test whether floors are emptying because of building quality or because the work on them has moved.
Ask what share of the occupier's floor is verifiable processing work, since that is the part most likely to leave first.
Office demand is where AI's effect on the built environment shows up in square feet. The Project Flux newsletter follows that evidence every week, caveats included. Subscribe free at projectflux.ai.
Links and Stuff
All content reflects our personal views and is not intended as professional advice or to represent any organisation.
