How AI Is Changing Commercial Real Estate in the UK

AI’s effect on UK commercial real estate so far is narrower and more specific than either the hype or the February 2026 stock market panic suggested. It is automating structured, well-defined tasks - document review, valuation modelling, brief matching, inventory scoring - while relationship-driven work, negotiation, and judgement calls stay firmly with people. Adoption is real and accelerating. A single, sweeping “AI takeover” of the industry is not what the data shows.

The panic is worth understanding on its own terms first. On 11 February 2026, shares in CBRE, JLL, and Cushman & Wakefield fell 12%, 12%, and 14% respectively in a single trading session - the sharpest one-day drop for two of the three since the pandemic sell-off of 2020. The trigger, according to Bloomberg’s report on the sell-off, was a product demo: Anthropic had shown agentic tools capable of auditing complex commercial leases and generating valuations in seconds. Investors read it as evidence that the fee-generating administrative work underpinning the big real estate services firms was more exposed than the market had priced.

I build software for the exchange between brokers and workspace operators, not for the parts of commercial real estate that just took a same-day, double-digit correction. But the panic and what’s happening underneath it point in different directions, and the gap between the two is worth separating out properly.

The panic, and what triggered it

A single demonstration moved a combined tens of billions of pounds in market value in one afternoon. That’s a strong reaction to a capability that, as far as the public record shows, hadn’t yet been rolled out inside any of the affected firms.

What the demo showed was document review and valuation modelling done faster - a real capability, applied to real work that currently sits with junior analysts and associates at the big services firms. That’s worth taking seriously. It is not the same as those firms’ entire commission and fee-based model being replaced overnight, which is closer to what a 12-14% single-day move implies the market was pricing.

A demo that reads a lease faster is not the same claim as a firm that no longer needs the people who build client relationships around what’s in it.

Markets overreact to demonstrations of capability more often than they underreact, particularly when the underlying technology has spent two years generating headlines it hasn’t yet had to fully justify. The more useful question isn’t whether the demo was impressive (by most accounts, it was). It’s what the actual adoption data across the industry says is happening, separate from what a single afternoon’s share price implies.

What the adoption numbers say

Adoption is real and accelerating, but “transformative” is mostly not the right word for what’s landing.

Deloitte’s 2026 Commercial Real Estate Outlook, based on a survey of more than 850 C-suite executives across 13 countries in North America, Europe, and Asia-Pacific, found that 92% of firms have now piloted AI in some form. Only 5% say they’ve achieved all of their AI goals. More tellingly, the share of executives describing AI’s impact as transformative fell to around 1%, down from roughly 12% the year before - while the share reporting merely incremental operational improvement rose from 7% to 24%.

Read together, that’s a market moving from inflated expectation to a more accurate one. AI is improving specific, structured tasks inside real estate firms. It isn’t restructuring the industry in the way both the optimistic pitch decks and the pessimistic stock reaction of February implied. In the 2026 PwC and Urban Land Institute Emerging Trends in Real Estate Europe report, 75% of European real estate leaders said they were now using AI in their operations, up from 51% the year before - a steep adoption curve, running alongside a much flatter curve of firms who’d call the results transformative.

The incremental 24% is a more useful place to look than the 1%. In practice, that kind of improvement tends to be narrow and specific rather than sweeping: document review that used to take an analyst a full pass now needing a lighter check, structured extraction from leases and briefs that used to be manual now happening automatically, drafting work that used to start from a blank page now starting from a usable first version. None of this is nothing. All of it is bounded. It’s a slow erosion of the hours spent on work that never needed a human’s full attention, not a single dramatic handover of judgement.

Of course, “mostly incremental” doesn’t mean “safe to ignore.” A 17-point jump in real, if modest, operational improvement across a single year is a fast-moving baseline, and the firms treating this as noise from a single overreacting Tuesday are making a different bet to the ones reading the underlying trend line.

Where the activity concentrates

The clearest UK evidence sits in flexible and managed workspace, not in the valuation and lease-audit work that spooked the market in February.

CBRE’s analysis of AI-driven London office demand found that AI companies accounted for 34% of London tech office take-up in 2025, up from just 4% a decade ago. Of the 100 leading AI startups in the city CBRE studied, 75 operate from flexible workspace rather than conventional offices. These are fast-growing, headcount-uncertain businesses choosing flexibility over long-term commitment, and they’re doing it at a pace that’s reshaping demand in the segment. Savills’ global survey of the forces shaping real estate in 2026 put technology in second place among the year’s most influential themes, driven specifically by this kind of AI-linked occupier activity.

This is the segment I work in, and it’s a useful corrective to the February narrative. The companies generating the most AI-driven office demand aren’t automating away the broker relationship that sources their space - they’re moving fast enough that the broker’s speed and market knowledge matter more, not less. I’ve written in more detail about how AI is changing the flex workspace broker workflow - the short version is that AI is compressing the administrative steps (brief parsing, matching, shortlist collation) while the judgement calls stay exactly where they were. The same shift viewed from the supply side, where the brief lands rather than where it starts, is covered in how AI workspace matching changes what operators receive.

Managed workspace tells a similar story from a different angle. A managed workspace deal is bespoke by definition - fit-out, term structure, and branding negotiated between broker, client, and operator rather than picked off a shelf. That’s precisely the kind of context-heavy negotiation the adoption data says AI isn’t touching yet. What’s changed is upstream of the negotiation: sourcing and shortlisting, which follow the same brief-in-and-match pattern as flex, whatever the client eventually signs. I’ve set out what sourcing managed workspace looks like in UK CRE in more detail elsewhere; the pattern holds across both segments even where the deals themselves look nothing alike.

The AI panic story and the AI adoption story are both true. They’re just describing different layers of the same industry.

The governance catch-up

Professional bodies moving to regulate a technology is usually a lagging signal that the technology has moved past the pilot stage, not a leading one.

RICS, the UK-headquartered chartered body for the surveying profession, published its first global professional standard for responsible AI use in surveying, effective from 9 March 2026. It’s mandatory for RICS-regulated firms using AI in valuation, construction, infrastructure, and land services, and it requires documented governance: risk registers, due diligence records, and clear policies on how AI outputs get reviewed before they’re relied on.

That RICS felt the need to mandate this now, rather than three years ago, tells you something the share price move on 11 February doesn’t: AI use in valuation and surveying work has become common enough across the profession that “no policy” was itself becoming the risk.

What Great Space does differently

On the Great Space platform, AI is applied to exactly the kind of structured, well-defined task the adoption data says AI is good at right now: matching a broker’s brief against operator inventory. Location, desk count, budget, term, and start date are required fields before a brief can be submitted. The platform scores operator fit against those fields before the brief reaches the network, rather than distributing it broadly and leaving assessment to whoever happens to open the email first.

The result: median response time on the platform is under two hours from brief submission¹. With 150+ verified UK workspace providers in the network, covering both flex and managed workspace, the brief reaches operators the broker may not have a direct relationship with, scored by genuine fit rather than distributed as a blind broadcast. What it doesn’t do is replace the broker’s read on which client needs flex versus managed workspace, or which operator’s space photographs better than it shows in person. That judgement is exactly the layer the adoption data says isn’t going anywhere. For the fuller picture of what a platform built around this distinction looks like, see what a workspace deal platform is.

The future of commercial real estate software isn’t a single audited lease making headlines for an afternoon. It’s the unglamorous exchange running underneath thousands of briefs, quietly, every day - which is a much less exciting story than a 14% share price move, and a much more accurate one.

If you’re a UK workspace broker spending more of your week on brief distribution than client conversations, see what Great Space does for brokers or start a free 30-day trial. No credit card required, and providers always receive and respond to requirements for free.

Footnotes

1. Based on Great Space platform data, Q2 2026. ↩