RICS AI standard: a checklist for small firms
The RICS AI standard, often called the RICS AI guidance, as a plain-English checklist of the registers, disclosure and sign-off a small UK firm needs.
If you work in a small surveying practice, keeping up with compliance often falls to you and a small team. There is no head office tracking every change from RICS or government, then turning it into a workable process. That is why we try to bring some perspective and practical guidance. Today we are looking at the RICS AI standard, which many surveyors still call the RICS AI guidance, and what it means for firms like yours. As a technology firm working in UK surveying, we follow this area closely.
RICS describes “a substantial proportion” of regulated firms as sole practitioners or sole-principal micro-businesses. This guide turns the standard into practical work for those firms.
The RICS AI professional standard: what changed and when
On 9 March 2026 the RICS professional standard Responsible use of artificial intelligence in surveying practice came into force. It applies to every RICS member and regulated firm worldwide when AI has a “material impact on the delivery of surveying services”.
The standard does not tell you whether to use AI. It requires you to know which tools you use, understand their risks, stay in control and keep evidence. Many surveyors still call it the RICS AI guidance, but it is a professional standard, so its requirements are mandatory rather than optional.
Who does the RICS AI standard apply to?
If you are a member of RICS, the standard applies to your professional use of AI. If your practice is RICS-regulated, it also has firm-level duties such as maintaining the AI and risk registers. In both cases, the requirements apply where AI materially affects the delivery of surveying services.
Report drafting, structured dictation, AI-assisted quality control and valuation tools may affect the professional output. A spellchecker or suggested email reply is less likely to do so. When you are unsure, record your decision and the reason.
RICS AI standard checklist for small firms
Each item below comes from the standard, although some duties sit with the member and others with the regulated firm. We have translated the requirements into plain English, but the RICS standard is the source to work from.
1. A register of the AI tools you use. Record the tool, supplier, purpose, adoption date, who may use it and its next review date. Note where the supporting evidence is kept. The format is not prescribed. A simple spreadsheet will suffice, or use whatever system you already rely on, provided the record is easy to find, review and update.
2. Check each tool before you use it, and keep the answers. First decide whether AI is suitable for the task and whether a sensible alternative exists. Then record what surveying data the tool receives, where it is processed, who else handles it, how long it is kept and whether it is used to train AI. Client details, property addresses, photographs, site notes, valuations and reports matter more here than the account and card details used to buy the software.
Also ask about the quality and possible bias of the information used to develop the tool, the consequences of errors, environmental and wider stakeholder impacts, and what responsibility the supplier accepts. Record your own practical tests and any answer the supplier cannot provide. The supplier’s trust, security and privacy pages are useful places to start. If client data leaves the UK, extra UK GDPR rules may apply. We explain those in our guide to the 2026 UK data-protection changes.
3. Keep an AI risk register. This is not your firm’s full risk register. It records the risks created by AI tools that materially affect your surveying services. RICS-regulated firms must include the chance of biased or incorrect output, gaps in the information supplied about the tool, and what happens to data put into it. For each risk, record how likely it is, the possible impact, whether the firm will accept it, what you will do to control it, its current status and a red, amber or green rating. Review the register at least quarterly.
RICS requires a red, amber, green rating or a similar method, but it does not prescribe what each colour means. Define the rule for your firm and use it consistently:
- Red: the risk is unacceptable or important facts are missing. Do not use the tool for client work until the problem is resolved. Unknown use of client data would justify red.
- Amber: the risk can be managed, but only with a clear control and regular checks. An AI drafting error controlled by a surveyor checking every output might be amber.
- Green: the risk is understood, low and supported by evidence that the controls work. Keep it under review rather than treating it as permanently safe.
Keep the colour separate from the action status, such as “awaiting supplier response” or “control in place”. Update both when the evidence, the tool or the way you use it changes.
4. A written reliability decision on each output, by a named surveyor. When AI contributes to a piece of work, a named and appropriately qualified surveyor decides whether its output can reasonably be used for its intended purpose. Record the assumptions, main reliability concerns, checks and controls, how those concerns affect reliability, and the final conclusion. Keep the decision on the report or job record, with the surveyor’s name and date.
5. Disclosure to the client in your terms of engagement. For each relevant survey type, explain when AI is used and which parts of the service it affects. Where available, cover the relevant professional indemnity insurance (PII), how a client can question an AI-related result, the complaints or redress route, and whether the client can opt out. Do not promise an opt-out unless the firm can provide the service on that basis.
6. Dip-sampling of high-volume automated output. Where reviewing every individual output is not practical, check random samples at regular intervals and record what you found. Choose a sample and frequency that suit the volume and risk. The firm remains responsible for every output, including those outside the sample.
What insurers may ask
Some practices are already being asked about AI use and internal controls at PII renewal. MFL Insurance’s bulletin for CAAV and RICS members says this is not yet commonplace, but reports questions about how AI is used and what internal guidance is in place.
An AI register, reliability records and client disclosures put the evidence in one place. Document them before the renewal form arrives, not while you are trying to complete it.
How to set it up in an afternoon
- List every tool that touches survey work. Decide whether the standard applies, who may use it, where its records will live and who will keep them updated.
- Start the tool register. Find the supplier’s trust, security and privacy information, record your practical test, then mark unanswered questions as “to confirm”.
- Start the AI risk register. Record each risk, its likelihood and impact, whether you accept it, the control, current status and colour rating. Put a quarterly review in the calendar.
- Add clear AI wording to the terms of engagement for each relevant survey type.
- Decide where the named surveyor will record each reliability decision.
- Choose a random sample and checking frequency that fit the volume and risk of any automated output.
Where SCORE fits
SCORE is practice software for UK surveying and valuation firms. SCORE QC checks a report you have written and flags what to review. A named user makes the final decision and signs off the check on the job.
Our trust centre confirms that SCORE hosts customer records in the UK and carries out AI processing in the UK, with EU failover. It also lists the other companies involved in providing the service, any safeguarded international transfers, and our commitment that customer data is never used to train models. That answers many of the supplier questions.
SCORE is not approved, endorsed or affiliated with RICS, and this article is not legal advice. The standard itself is short and readable, and it is the document to work from.
Worked examples
These examples show what the records might look like. They are not answers to assume about a real tool.
Supplier check
| What to check | Example answer |
|---|---|
| Client and job data used | Client name, property address, photographs, site notes and draft report |
| Where it goes | Stored on UK servers and processed by an AI provider in the EU |
| Who else handles it | The supplier’s named cloud and AI providers |
| Used to train AI | The supplier says customer data is not used for training |
| How long it is kept | Uploaded files are deleted after 30 days |
| Evidence | Supplier trust centre, privacy documents and a recorded sample test |
| What is unclear | Whether backups follow the same deletion period |
The full supplier record should also cover why AI is suitable, alternatives considered, training-data quality and possible bias, environmental and stakeholder impacts, responsibility for errors, and the result of your practical testing.
AI risk register
| Risk | Likelihood and impact | Decision and control | Action status | RAG rating |
|---|---|---|---|---|
| The tool invents or misses a defect | Possible, with a high client impact | Not accepted. Surveyor checks the draft against site notes and photographs | Control in place | Amber |
| Client data is retained or used to train AI | Unknown, with a high data-protection impact | Do not upload client data until the supplier confirms its position | Awaiting supplier | Red |
| Supplier information is incomplete | Likely, with a medium impact | Record the gap, ask the supplier and test sample reports | In progress | Amber |
Reliability decision
Reviewed by A Surveyor on 11 September 2026. I compared the AI-assisted report check with the inspection notes and photographs. The tool may miss defects described in unusual language, so I reviewed each flagged and unflagged section. I corrected one omission. With that correction, I consider the report suitable for issue to the client.
Client disclosure
Adapt this example to the service and have the firm’s wording checked before use:
We use software to check our reports for completeness before issue. It highlights possible omissions but does not inspect the property or make the final professional judgement. A named, appropriately qualified surveyor reviews the results and signs off every report. Our professional indemnity insurance applies to the service [include only if confirmed]. You can question an AI-assisted part of our work through [contact or process], and our usual complaints and redress procedure applies. You may ask us before instruction not to use this tool; we will confirm whether we can provide the service on that basis.
Dip sample
If AI extracts details from 100 reports in a month, the firm might select 10 at random and compare the property address, inspection date and valuation figure with the original reports. Record errors and the action taken. Ten is only an example, not a required sample size.
Prompt to help draft the records
An AI assistant can draft the records if you provide the facts, include no client data and keep the decisions for yourself.
Help a small UK surveying practice prepare records for the RICS professional standard "Responsible use of artificial intelligence in surveying practice". Work only from my facts. Where information is missing, write "to confirm". Use UK English and no client data.
ABOUT THE FIRM
[Team size, RICS status and survey types.]
TOOLS WE USE THAT TOUCH SURVEY WORK
[Tool, purpose, whether it uses AI and whether it receives client data.]
HOW WE CURRENTLY CHECK AND SIGN OFF REPORTS
[Who reviews work and where sign-off is recorded.]
Produce:
1. An applicability decision for each tool, with a reason and any uncertainty.
2. An AI register with tool, supplier, purpose, adoption date, client data, applicability and review date.
3. Supplier questions covering suitability, alternatives, data, model training, bias, environmental impact, errors, liability and practical testing.
4. A risk register with likelihood, impact, risk appetite, control, action status and RAG rating.
5. Plain-English client disclosure for each relevant survey type, including how a client can question the result, seek redress or ask to opt out.
6. A reliability sign-off covering assumptions, concerns, checks, controls and the final conclusion.
7. A dip-sample log and a list of gaps to confirm.
Keep it practical. Present choices for us to make rather than making them for us.