What the responsible AI standard changes
The RICS professional standard on the responsible use of AI in surveying practice came into effect on 9 March 2026. It applies to AI with a material impact on surveying services and places professional judgement, governance, output reliability and client communication at the centre of responsible use. The current standard and supporting material are available from RICS.
For report preparation, the practical consequence is straightforward: an AI suggestion cannot become professional evidence merely because it reads well. The practice must know why the tool is being used, what information it receives, where it can fail and who decides whether a material output is reliable.
A seven-stage QA and governance workflow
- 1
Define the use and its material impact
Record what the AI does, which surveying service it affects, whether its output may materially affect delivery, and the accountable practice owner. Revisit the assessment when the tool or use changes.
- 2
Approve the system and the conditions of use
Complete proportionate due diligence on capability, limitations, data handling, security, provider terms and human oversight. State which tasks and information are prohibited.
- 3
Control the working context
Provide only information needed for the task and keep the source material identifiable. The ICO’s AI guidance frames this around lawfulness, transparency, minimisation, accuracy, storage, security and accountability.
- 4
Generate into a reviewable draft
Keep the suggestion visibly separate from accepted report content. Preserve enough context to understand the request, model or system used, output, timing and affected job without logging unrestricted client data.
- 5
Verify against inspection evidence and scope
Check the suggestion against notes, photographs, property context, condition ratings, known absences, terms of engagement and the relevant report level. Look deliberately for contradictions and invented specificity.
- 6
Record the reliability decision
For material outputs, record the appropriately qualified named surveyor’s decision and the matters considered. RICS’s summary of the standard’s key elements explains the output assurance and written-decision expectations.
- 7
Communicate, issue and retain proportionate evidence
Give clients the required information about material AI use, complete the normal report QA and retain the decision trail under the practice’s documented policy. A published report remains the surveyor’s professional work.
Plan for predictable failure modes
Responsible use is easier when the review is designed around the ways a system can fail. The following checks should be explicit rather than left to a general instruction to “read it carefully”.
| Failure mode | What it can look like | Required check |
|---|---|---|
| Invented fact | A material, defect, age, inspection result or repair detail that is not in the evidence. | Trace each specific statement to notes, photographs, documents or a clearly identified professional inference. |
| Scope expansion | Commentary beyond the agreed service or inspection limitations. | Compare the draft with the terms, report level and recorded limitations. |
| Contradiction | A risk about an element recorded as absent, or wording that conflicts with a condition rating. | Run element-presence, rating and cross-section consistency checks before approval. |
| Loss of provenance | A polished paragraph whose source or generation history cannot be reconstructed. | Keep the relevant input context, output, edits, model identity and decision record together. |
| Automation bias | A reviewer accepts plausible wording without applying professional scepticism. | Require an active reliability decision and sample completed decisions for quality review. |
| Confidentiality failure | Client or property information reaches an unapproved service or remains longer than expected. | Control approved tools, data categories, retention, subprocessors and incident handling. |
Keep the minimum useful record
A small practice does not need to turn every AI interaction into a large compliance file. It does need a record that can explain the decision after the event. For a materially relevant report output, keep:
- the instruction, report section and purpose of the AI use;
- the approved system and identifiable model or configuration;
- safe metadata describing the source context used;
- the generated suggestion and the material edits made;
- the checks performed, contradictions considered and limitations noted;
- the named surveyor’s decision, outcome and time; and
- the applicable client communication and retention rule.
Logs should help answer what happened, where, to which instruction and why. They should not become an uncontrolled second copy of client files or full provider payloads.
Put the control into ordinary practice
Start with one report type and one clearly bounded use. Train the surveyors on the likely failure modes, review the first completed records together and adjust the workflow where the evidence is difficult to capture. A control that works only during an audit is not yet part of the practice.
SurveyOS is designed around this separation of suggestion and decision: report context can support an AI-generated draft, while the surveyor reviews, amends and accepts or rejects it before publication. The software can support the process, but the practice still owns system approval, professional judgement, client communication and policy.
How this guide was prepared
The SurveyOS Editorial Team prepared this guide using current primary sources and evidence from the product workflow. AI assisted with drafting and structure. The factual claims and product statements were checked against the sources and current application behaviour on 4 August 2026. This is general practice guidance, not legal or professional advice.
Primary sources
Put the workflow to the test
See SurveyOS on one real instruction.
Book a demo and compare the workflow with the way your practice works today.