A demonstration can show an AI system producing a report, measuring a drawing or extracting information from a specification.
That is useful.
But before introducing it into a surveying practice, there are other questions to answer.
What happens when it misses something? Who checks the result? Where does the client’s information go? And what evidence supports the decision to use it?
On 23 October 2026, RICS published a practical AI procurement guide to help members and regulated firms work through these decisions. It explains how to apply the responsible-AI professional standard that came into effect on 9 March 2026.
For practice leaders, the practical task is to connect the proposed improvement with evidence, appropriate controls and clear responsibility.
01 / SCOPE
Understand where the standard applies
The standard applies to AI outputs that materially affect the delivery of surveying services. The assessment depends on how the output influences the service, and a determination of material impact must be recorded with its reasoning.
Consider two examples.
An AI tool helping arrange an internal meeting has a different relationship with professional work from one preparing quantities that feed into a cost plan.
The same distinction applies to reporting. Correcting the formatting of a document is different from generating descriptions or conclusions that a surveyor will rely on.
Our recommendation is to map the proposed workflow before choosing a tool.
Identify the information going in, the output coming back and the decisions that output could influence. That gives the firm a much clearer basis for assessing the system.
02 / PURPOSE
Start with a clearly defined problem
The procurement guide asks firms to establish the business problem and consider whether AI is appropriate, including available alternatives.
For a practice leader, a useful starting point might be:
“Our surveyors spend too much time assembling the first draft of a condition report.”
Or:
“Our quantity surveyors repeatedly transfer measured quantities into our billing structure.”
Those statements give you something specific to investigate.
You can establish the current effort, understand the bottleneck and agree what an improvement would need to achieve.
Then consider the options.
Could better templates solve the problem? Does existing software already provide the capability? Would a small integration help? Is there a genuine reason to develop a tailored system?
The reasoning should be clear enough for someone outside the project to understand.
Our guide to calculating AI ROI explains how to connect that decision to a measurable business case.
03 / SUPPLIER EVIDENCE
Ask for answers you can keep
The guide emphasises written supplier evidence covering data handling, training information, limitations, liability, environmental impact and development stakeholders. It also recommends understanding the complete data flow before signing.
We suggest turning those topics into a short supplier questionnaire.
Start with the information your team will upload.
For a drawing-based workflow, that might include project names, layouts, specifications and commercially sensitive cost information. For condition reporting, it could include photographs, addresses and client documents.
Ask the provider to explain the journey in plain English:
- Where is the information stored and processed?
- Who can access it?
- Is it used to train or improve models?
- What is retained, and for how long?
- Which other providers are involved?
Then ask about the system itself.
What is it designed to do? Where does it perform poorly? What happens when the input is incomplete or unclear?
Keep unanswered questions visible. They should inform the decision about whether, and under what conditions, the system is suitable.
For more detail, read where your data goes when you use AI.
04 / TESTING
Test representative work before relying on the result
RICS requires firms to retain a record of the extent of practical fitness-for-purpose testing. The procurement guide suggests parallel testing against recently completed work.
For a BoQ workflow, we would use completed drawings and specifications, with the necessary permissions and information safeguards in place.
Agree the test before running it.
Which quantities will be compared? What discrepancies need investigating? How will omitted rooms, ambiguous finishes or incomplete information be handled?
Measure the whole process, including professional review and corrections.
A system that produces quantities in minutes may still require substantial checking. Understanding that review effort is essential to assessing the benefit.
An illustrative test record could include:
| Measure | What to record |
|---|---|
| Time | Preparation, processing, checking and correction |
| Quality | Omissions, discrepancies and unsupported assumptions |
| Exceptions | Cases requiring further information or professional judgement |
| Usability | Whether the checked output fits the firm’s normal workflow |
| Decision | What can proceed, what needs changing and what remains unsuitable |
That gives leadership a useful basis for deciding the next step.
05 / PROFESSIONAL REVIEW
Make the surveyor’s decision visible
Material outputs require documented reliability decisions under a named, appropriately qualified surveyor. Automated or high-volume outputs require regular randomised sampling.
For a technical workflow, we recommend designing the review process alongside the automation.
In a room-measurement system, the QS should be able to inspect the detected boundaries, correct quantities and resolve specification questions before accepting the bill.
In a condition-reporting workflow, the surveyor should be able to compare the draft with the inspection evidence and correct unsupported descriptions.
The interface should make that work straightforward.
Show the relevant source. Make corrections easy. Keep a record of the review and its outcome.
This also helps identify recurring weaknesses. If users repeatedly correct the same type of error, the team has something concrete to improve.
06 / OPERATIONS
Plan for everyday use
The standard requires governance policies, an AI register, relevant training and a risk register reviewed at least quarterly. It also requires advance client disclosure of material AI use and appropriate engagement terms.
We recommend agreeing the practical ownership before moving into production.
Who supports users? Who investigates an error? Who decides whether a software update requires further testing? Who keeps the supplier information current?
Continuity deserves attention too.
The procurement guide asks firms to consider what happens if the supplier disappears.
Ask how you would retrieve project records and continue working if the service became unavailable.
Agree export arrangements, support responsibilities and the information needed to transfer or replace the system.
These questions are much easier to resolve before the workflow becomes essential to delivery.
07 / NEXT STEPS
Build the evidence alongside the solution
A supplier cannot make a surveying firm compliant on its behalf. The firm retains responsibility for applying the standard.
A useful provider should help make that responsibility manageable.
Our recommended approach is to gather the evidence as the project progresses: define the problem, explain the technology, test representative work and agree the professional controls.
That avoids reaching the end of a successful proof of concept only to discover that nobody has resolved the questions needed to approve its use.
At Go Left, our approach connects these decisions through the BUILD™ framework, from understanding the current process to implementation, learning and ongoing development.
The objective is to release professional capacity while giving the firm a clear basis for trusting and managing the new workflow.
This article provides a practical introduction. Read it alongside the RICS responsible-AI professional standard and AI procurement guide.