AI governance
The RICS AI standard asks for a paper trail. StackProof keeps it.
The RICS professional standard Responsible use of artificial intelligence in surveying practice has applied to all members and regulated firms since 9 March 2026. Here is how StackProof sits inside it, and what it gives you to show it.
Where AI is used, and where it is not
Rule checks: no AI
Most flags come from fixed rules written from real complaints: contradictions, carried-over details, figures, dates, ratings and your firm’s own credentials. The same report gives the same flags every time.
AI review: assistive, and optional
An Anthropic Claude model reads the report text against itself and, if switched on, against public records for the address. Its flags are marked as AI and need the surveyor’s call like any other. A surveyor can switch it off for a report.
What the standard asks, and what you get
The standardA written register of the AI systems the firm uses, with their purpose and review dates.
StackProofReady wording for your register entry, below. Copy it, add your dates and your owner.
The standardDue diligence before using an AI system: whether it is the right tool, the data risks, and the risk of wrong or biased outputs.
StackProofWhat StackProof does, what it sends, who processes it and its known limits, set out on this page. A fuller due diligence pack is available on request.
The standardProfessional judgement on whether an AI output can be relied on, with a named surveyor responsible.
StackProofEvery flag needs the surveyor’s call: Fixed, Not an issue (with a reason) or Unsure. Each call is recorded with who made it and when. StackProof never changes a report by itself.
The standardRegular checks where AI is used at volume.
StackProofA QA record for every report, sent to your admin when the report is marked final. On Super, admins and reviewers see every surveyor’s scans and the firm’s figures.
The standardTelling clients in writing, in advance, when and why AI is used.
StackProofSuggested wording for your terms of engagement, below, including how a client can opt out.
The standardEnough knowledge of AI and its limits to use it responsibly.
StackProofPlain statements of what the AI review can and cannot do, so your team knows what a flag is and what it is not.
Known limits
Worth sharing with your team, so a flag is read as a question for the surveyor and not as an answer.
- It can miss a mistake. A report with no flags is not a report with no errors.
- It can raise a flag that is not a mistake. That is why every flag needs a surveyor’s call.
- Public records it checks against (energy certificates, listings, flood maps) can be out of date or wrong.
- It reads the text of the report, not the photographs, plans or the property itself.
- It does not give advice, set a value or a condition rating, or decide whether a report can be issued.
Wording you can use
Your firm decides, and records, whether a quality check like StackProof has a material impact on your work. Either way, this wording is a starting point. Adapt it to your own policy.
AI system: StackProof AI review (Anthropic Claude, provided through SurveyStack) Purpose: Quality check of draft survey and valuation reports before issue. Flags possible errors and inconsistencies for the surveyor to review. Does not write advice, set values or condition ratings, or change the report. Material impact: [Your firm's assessment and reasoning] Data: Report text only. Not used to train AI models. Held until the draft and final are compared, then deleted. Flags and surveyors' calls kept as the QA record. Human oversight: The surveyor reviews every flag and records Fixed, Not an issue or Unsure. The named surveyor signs the report and remains responsible for it. Date of first use: [date] Next review: [date, no more than 12 months on] Owner: [name and role]
We use software, including artificial intelligence, to check our draft reports for errors and inconsistencies before they are issued. It does not write our advice, set any value or condition rating, or change the report. Every point it raises is reviewed by the surveyor named in the report, who remains responsible for its contents. Your report is not used to train AI models. If you would prefer your report not to be checked in this way, or you have a concern about how AI was used, please contact us before the inspection or at any time afterwards.
This page explains how StackProof works alongside the RICS standard. It is not legal or regulatory advice. Check your own obligations against the standard itself, and see security and data for how report text is handled.