The RICS AI Standard Is Now in Force: What It Requires and How to Comply

The first global professional standard on the responsible use of AI in surveying became mandatory on 9 March 2026. Here is what it asks of you — and a repeatable way to meet it.

The first global professional standard on the responsible use of AI in surveying became mandatory on 9 March 2026. If you are an RICS member or work in a regulated firm and you use AI anywhere it materially shapes your advice — an automated valuation, a forecast, a cost estimate, a drafted report — you are now expected to be able to show how you decided that output was reliable, and to have written it down.

Most firms have adopted the tools. Far fewer have the discipline the standard now requires sitting behind them. That gap is the whole subject of this guide. Below is what the standard actually asks of you, and a practical, repeatable way to meet it.

What is the RICS responsible-use-of-AI standard?

It is a professional standard, published by RICS in September 2025 and effective from March 2026, governing how surveyors and regulated firms use artificial intelligence. It is not optional guidance. It sets mandatory requirements wherever AI has a material impact on the service you deliver, and it spans valuation, building surveying, construction, infrastructure and land.

The core principle is simple: AI can support your work, but it cannot replace your professional judgement, and a competent surveyor remains accountable for every output relied upon. Everything else in the standard follows from that one idea.

Who does it apply to, and when did it start?

It applies to all RICS members and RICS-regulated firms, anywhere in the world — not just the UK. It has been in force since 9 March 2026, so the compliance question is no longer “when”; it is “can I demonstrate it today”.

It bites specifically where AI use is material. Casual, low-stakes use sits outside the heaviest requirements. Use that shapes the substance of your advice sits squarely inside them.

What does “material impact” actually mean?

An AI output has a material impact when it affects how your professional work is made meaningful — in other words, when relying on it changes the advice the client receives. An AVM figure that feeds a valuation is material. A model forecast that drives an investment recommendation is material. A first-draft email is generally not.

The standard puts the judgement of materiality on you, and asks you to record that judgement. If you decide AI is materially shaping a piece of work, you must write down that determination and your reasoning. That written record is itself a compliance requirement, not an optional nicety.

What does the standard require you to do?

Stripped to its essentials, the standard asks for five things wherever AI use is material:

• Human oversight. A named, appropriately qualified surveyor must take responsibility for the reliability of material AI outputs. Where use is automated or high-volume, that means a written assessment or a defensible sampling-and-assurance approach — not blind reliance.
• A material-impact assessment, in writing. Determine whether AI materially affects the service, and record the determination and the reasoning behind it.
• Client transparency. Tell clients, in writing and before work begins, when and how AI will shape the service — and how they can question or opt out of it.
• Governance and risk management. Hold responsible-use policies, a risk register, and due-diligence procedures covering the AI you rely on, including third-party tools you procure.
• A system governance assessment before deployment. For material uses, record why AI is the appropriate tool and whether it is accurate, reliable and fit for the specific job — accounting for data risks and the risk of erroneous or biased outputs.

There is also a data line worth flagging: confidential or private client information should not be put into AI systems without express written consent and proper safeguards.

How do you comply in practice?

The requirements are clear; the hard part is the daily discipline — a consistent way to interrogate an AI output before you rely on it, that produces the written record the standard expects. That is what turns a policy on a shelf into defensible practice.

I use a seven-point review I call the AI Output Review. Run any material AI output through these checks before it leaves your desk:

1. Fit — is this model the right tool for this asset, market and period?
2. Confidence vs. calibration — does the confidence score actually mean what it implies?
3. Assumption fragility — which single input, if wrong, moves the answer most?
4. Regime check — does it assume conditions that no longer hold?
5. Traceability — for anything AI drafted, is every figure, clause and citation real and checked?
6. Bias and fairness — could the output embed a pattern you cannot justify?
7. Accountability and record — who signs, and is the material-impact note written down?

Each check maps directly onto what the standard demands — professional scepticism, retained accountability, a documented assessment, and client transparency. Done consistently, the review is your compliance evidence.

Does it apply to AVMs and automated valuation?

Yes — and this is where firms are most exposed. An AVM output that feeds a valuation is a textbook material AI output. The standard expects a named, competent surveyor to own the reliability decision, and where valuations are produced at automated or high volume, it expects a written assessment or a documented sampling-and-assurance regime rather than wholesale reliance on the model.

RICS has signalled that dedicated guidance on AI in real estate valuation is coming — a global practice document went to consultation in 2026 — so the expectations around AVMs are likely to sharpen, not soften. Building the review discipline now is the low-cost way to stay ahead of it.

What happens if you don’t comply?

This is enforceable, not aspirational. Compliance is mandatory, and RICS’s Regulatory Tribunal makes final determinations on disputed cases — so a firm that cannot show documented oversight is exposed in a way it was not before. Professional indemnity insurers and brokers are also watching the standard closely, which means weak AI governance can become a question at renewal, not just a regulatory one.

The reframing worth making internally is that the documentation is not bureaucracy. It is the line between AI as an unexamined shortcut and AI as a capability you can stand behind professionally.

How does it relate to the EU AI Act and wider regulation?

The standard is deliberately aligned with the EU AI Act and the broader global direction of travel on AI governance. That alignment is a feature: a firm that builds a genuine review-and-record discipline to meet the RICS standard is also building most of what it needs for the wider regulatory environment, rather than solving the same problem twice.

This guide is a practical overview, not legal or regulatory advice. Check the requirements against the RICS standard itself and your firm’s own obligations. Source: RICS, “Responsible use of AI in surveying practice”, rics.org.