Faster than we can explain

What surveying is actually doing with AI

Independent intelligence on AI for the UK surveying profession.

What we found

Five key areas we discussed:

Foreword

The noise arrived before the evidence. Litmus exists because of what I kept hearing, and what I kept failing to find.

Across surveying, construction and the wider built environment, artificial intelligence arrived accompanied by a great deal of noise. Vendors with something to sell. Confident opinions that travelled further than the evidence behind them. Misinformation that took hold because there was nothing independent to check it against. Some of it well meant. Very little of it tested.

Then the RICS Professional Standard on the responsible use of AI took effect on 9 March 2026, and firms across the profession began working out what it meant for them.

What was missing was anywhere that looked across the whole profession at once. There is excellent work happening in individual disciplines. There is nothing that applies a litmus test across all of them, from residential valuation to quantity surveying to geospatial, and reports back without an agenda. That is what this group is for.

This is the first Litmus report. It is a reading of a conversation, not a dataset. It captures what senior practitioners are seeing from very different positions in the profession. It carries no survey figures, because the survey has not yet run. That work comes next, and it is described at the end of this report.

I convene and chair Litmus. The findings in this report belong to the group.

Nina Young FCA

Chair, Litmus

About this report

This report draws on the first Litmus session, held on 21 July 2026, attended by five contributors and the chair. The session ran under the Chatham House rule. The recording is not shared outside the group.

Contributors speak in a personal capacity. Where a contributor is quoted by name, the quotation reflects their professional view. Observations concerning specific firms, clients or institutions are reported without attribution, in the group’s own voice, in keeping with the Chatham House rule. Every named quotation was approved by the contributor before publication.

This is qualitative work. No claim is made that any of it is representative of the profession as a whole. Establishing that is the purpose of the research described in section six.

Where the profession is

The RICS AI Standard took effect in March, then something closer to silence.

Jony Milner, Technical Manager at Legal & General Surveying Services, described the shape of the reaction to the new RICS AI Standard.

Contributors described a common first difficulty. Firms could not establish where AI was already in use. Staff who reported no AI use were, on questioning, using assistants embedded in office software, or writing tools, without recognising either as AI. The initial structural response in several firms was to appoint internal champions or departmental leads, largely in order to find out what was already happening.

One contributor described the argument that shifted their organisation from discussion to action. Rather than presenting the standard as an obligation to be met in due course, it was presented as a failure already under way. If a firm cannot evidence a mandatory requirement, it is already non-compliant. Framed that way, the matter moved.

The pattern reported after that initial burst was consistent across several firms. Visible compliance work was completed. Momentum then slowed, with limited direction on what a next stage would involve.

James Garner, member of the RICS construction professional group panel, Head of AI and Data at Gleeds and founder of Project Flux, and a contributor to the standard itself, has observed the same trajectory more widely, and takes a positive view of what the standard achieved.

Contributors raised the question of enforcement independently of one another. The broadly shared expectation was that a publicly visible case, in which a firm is found not to have complied, will do more to concentrate attention than the standard itself has done. Nobody welcomed that prospect. Several regarded it as likely.

The standard is also being read beyond the body that wrote it. Contributors reported that other professional bodies are adapting it for their own members, and that it is being referenced outside the UK.

A UK professional standard becoming a reference point elsewhere is worth noting. It raises the stakes for how well it is implemented here.

A pattern that emerged repeatedly is that adoption does not track firm size. Contributors described firms avoiding AI entirely, firms experimenting without direction, and firms moving with real intent, at every scale from sole practitioner to international consultancy. The spread is visible from outside a firm as well as within it.

The risks that are not being priced

Exposure accumulates whether or not anyone has authorised it.

Suzanne Hill is a business strategist, founder of Constructive Intelligence, and an AI consultant working with construction firms. Her framing removes the question of whether to adopt.

She calls the underlying condition organisational drift. Firms defer the decision. Exposure accumulates in the meantime, and it accumulates whether or not anyone has authorised it. Her two working principles are worth reproducing as she states them: context before capability, rules before tools.

Chris de Gruben is senior director of AI strategy and governance at Artefact, where he works with property firms on deploying AI. He set out the economics that sit behind that.

The point extends beyond the model developers. Contributors distinguished between three commercial incentives now shaping the market: the frontier developers, who need training data; the platform providers, who do not own the models and instead seek to own the environment the models are used in; and the hardware manufacturers, competing for the consumer layer. The incentives differ. None of them are aligned with a firm retaining full control of its data.

A specific and under-examined exposure emerged around existing software. Vendors are adding AI functionality to products firms already use, waved through under processes designed for something else entirely.

Hill was clear that size offers no protection. Contributors described firms with turnover in the hundreds of millions instructing individual departments to trial AI tools independently, with no central governance, no register, and no vendor assessment. Contributors also raised concentration and lock-in risk, and contractual exposure — confidentiality obligations, non-disclosure agreements and client data restrictions being breached by ordinary use of ordinary tools.

AI pointed back at the surveyor

Weaponised against the surveyor by complainants and claimants.

The most original material in the session concerned exposure running in the opposite direction. Nik Carle is a litigation partner at Browne Jacobson LLP, specialising in professional liability for property professionals and acting on both defence and claimant sides. He set out two distinct tracks.

The second track is already visible. Contributors reported a marked change in the character of complaints from the start of the year. Submissions that would previously have been informal and loosely argued are now well structured and articulate. They cite specific RICS standards. They set defined response deadlines. The change was sufficiently abrupt to be noticed as a step rather than a trend.

Contributors also described a more serious development, tested internally. A photograph of an ordinary domestic interior was altered to introduce a structural defect. Those reviewing it could not reliably determine whether the defect was genuine.

The consequences are operational and commercial at once. If a photographed defect cannot be trusted, re-inspection becomes the default position, at a cost measured in a surveyor’s day plus travel for every contested case. Where re-inspection does not happen, the alternative is settling claims for defects that were never there. For firms handling volume work, neither outcome is sustainable at scale.

Carle noted an asymmetry that follows. A regulated professional is constrained in how they may respond, by conduct rules that do not constrain the person making the complaint. The obligations that define professional standing also limit the ability to meet an AI-assisted complaint on its own terms.

From automation to reinvention

Three stages, arriving in quick succession.

De Gruben’s account is the clearest description in the session of where the technology is heading, because it is drawn from deploying it. He described three stages, arriving in quick succession.

The third stage is different in kind.

He was equally clear about what has not been solved.

Hill pressed the practical objection. The capability is impressive in principle. A highly regulated industry has different conditions.

De Gruben accepted that neither is resolved. An important distinction emerged here: compliance with written rules is the tractable part.

He put a figure on that, about his own discipline: “I can automate eighty per cent of the valuation process with the tools available to me today.”

Contributors also raised the prospect of a discipline being absorbed rather than automated.

That view was shared by more than one contributor. It is contested elsewhere in the profession. It is recorded here because two contributors reached it independently.

Two problems arrive with scale. The first is cost. Contributors described firms encountering unbudgeted five and six figure monthly charges once staff began building agents freely. The second is sprawl.

The response now emerging is consolidation: templated agents, restricted deployment, and an orchestration layer to manage what has already been built.

Garner offered the counter-argument, and it deserves equal weight.

Hill’s warning about the wrong kind of progress is the one to hold on to:

Milner raised a question that follows from the same logic, applied to residential work. The underlying observation is that the profession is largely engaged in performing existing tasks faster, while the question of which tasks survive goes unasked.

The clearest example raised in the session concerned procurement rather than productivity. A large consultancy acquired an AI platform business, and is embedding it across its own systems. Contributors described the acquisition as defensive. The platform was regarded inside the business as a threat to it, and the response was to buy it. The consequence runs into procurement.

The mechanism is sequencing. If design optimisation and performance guarantees move upstream to the design stage, the order in which work is procured changes, and firms competing against that are competing on a different basis. Contributors noted that acquisitions running the other way may matter more.

The judgement gap

A product of having done the work.

The most consistent concern across the session had nothing to do with technology. Contributors described an experience paradox. Surveyors with long practical experience carry an instinct for when something does not look right, and that instinct is what catches an error before it becomes a claim. Newer entrants, contributors reported, are markedly less likely to challenge a plausible but incorrect statement. The capacity to be sceptical about an output is itself a product of having done the work.

Contributors also described a pattern of departure rather than resistance.

Contributors were less troubled by refusal to adapt than by quiet departure. Some experienced surveyors will simply go, and the judgement goes with them. Hill argued that the profession has the value equation backwards.

Carle brought evidence from the legal sector, where this has already been tested in a decided case involving a junior lawyer, a senior colleague and a supervising partner. All three were found wanting — the core failure was a complete trust in what the AI was saying to her. The junior lawyer’s conduct was the most striking element.

Garner was sharply critical of the response from higher education.

Contributors described a technology firm that renamed its software engineers as AI engineers and instructed them to stop writing code, recruiting instead for creativity, judgement and communication. The same shift is beginning to appear in surveying job specifications.

Professional qualification is starting to respond. A new RICS pathway in data analytics and intelligence is in pilot, developed with input from contributors to this group.

Garner closed with a longer view, pointing out that Plato had the same worry about writing eroding memory. “We will adapt, because we will have to.”

Who decides what competent looks like

Where a claim turns on the use of AI, or on the failure to use it, someone has to say whether the surveyor fell short. Courts rely on expert evidence to answer it. The test is whether the work sat inside the broad band of what a reasonably competent practitioner would have done in the same circumstances, or fell so far outside it that no competent practitioner would have made the same error. Perfection is not the standard.

Establishing where that band sits requires independent expert evidence. And the expert has to hold two things at once.

There is a sequencing problem beneath this. The band cannot be established until there is a body of real cases to establish it from, and there is no such body yet. The first claims will be argued in the absence of a settled view of what competent AI use looks like.

There is a second problem, and it is circular. Expert witness work is a sideline for many surveyors. On this evidence it becomes a considerably more significant one, and the people who can do it credibly do not yet exist in numbers.

What Litmus will test next

Ask how far AI has moved into UK surveying and you will get a different answer depending on who you speak to. That was true in this session, among people who spend their working lives close to the question.

No independent, full scale survey of AI adoption across UK surveying exists. Individual disciplines hold partial views. Vendors hold commercial ones. Nobody holds a picture of the profession as a whole.

Litmus is building it. The State of AI in Surveying survey will run across every surveying discipline, at every level of seniority, from those still qualifying through to firm owners.

Responses will distinguish between those answering for themselves and those answering on behalf of a firm. Both matter, and they are different pictures. A firm’s stated position and the daily experience of the people inside it are frequently not the same thing.

We need to hear from you. Whether your firm has a full governance framework in place or has never discussed AI at all, the picture is incomplete without you. Particularly if you think you have nothing useful to contribute, because that view is itself part of what needs measuring.

Contributors

About Litmus

Litmus is an independent intelligence group on AI for the UK surveying profession.

How it works

Litmus brings together senior practitioners from across surveying, construction and the professions that sit alongside them. The group meets quarterly for two hours, under the Chatham House rule. Contributors take part in a personal capacity and give their time voluntarily. They are not paid, and their firms have no standing in the group. Litmus is convened and chaired by Nina Young, founder of Surveyors UK.

What it publishes

– A report after each quarterly session

– Occasional briefings where something warrants it

Everything Litmus publishes is free. Nothing is published until every contributor has approved it.

Declaration of interest

Litmus is convened by Surveyors UK, an independent media and community business which sells products and services to the surveying profession. Litmus is editorially independent of that business. No Litmus publication recommends, endorses or promotes any product or service, including those of Surveyors UK or any business connected to it. We publish this so readers can weigh it for themselves.

Cite as

Litmus (2026), Faster than we can explain: what surveying is actually doing with AI, 1st Edition, September 2026.

Contact

Contributor enquiries and media: nina@surveyors-uk.com

Published by Surveyors UK (Advantage) Ltd, company number 12836014, registered in England and Wales. Registered office: Popeshead Court Offices, Peter Lane, York, England, YO1 8SU.

© Surveyors UK (Advantage) Ltd 2026. This report may be shared and quoted freely with attribution to Litmus. It may not be reproduced in whole or resold.

This report records views expressed by contributors in a personal capacity. It is not legal, regulatory or professional advice, and no reliance should be placed on it. Firms should take their own advice on their obligations under the RICS Professional Standard and any other applicable requirement.