At the RICS global valuation conference at the start of this month, one of our webinar guests did something quietly telling. Oliver Leeming, who runs the valuation division at Lambert Smith Hampton, used AI to count how many times AI was mentioned in the opening keynotes. The answer was about 80 times in 90 minutes. "Valuation" came up around 30.
That ratio says everything about where the profession's head is right now. So we sat down for a candid conversation about what AI actually means for surveying, and tried to separate the genuine shifts from the noise. Here is what came out of it.
The profession is at an inflection point, not a cliff edge
Most of the software our industry runs on was built before this wave of AI. That includes Aprao. We are at our own crossroads: do we build AI features into the product, or do we let the product connect to the AI that firms are already using? Almost every firm now has its own LLM in the building, whether that is Claude or Copilot. The interesting question is no longer "should we use AI", it is "where in the workflow does it actually earn its place".
The RICS standard on the use of AI in valuations, issued in September and effective from March, reflects that shift. Two years ago at the same conference, AI was a secondary, almost theoretical topic. Now it is front and centre, with a mandatory standard requiring a risk register, client notification with an opt-out, and a clear internal policy on how the tools are used.
The counterintuitive bit: smaller firms are more confident
You would assume the big firms with deep pockets and dedicated AI teams are racing ahead. The reality on the ground is more nuanced. The larger firms tend to have the governance structure in place but describe actual application as "hazy". The smaller firms have the structure too, and a noticeably higher level of confidence in using AI inside the real workflow. Being nimble and controllable beats having a big budget more often than you would think.
AI is doing the periphery, not the judgment
Today, AI is largely an auxiliary tool: front-end due diligence, data collection, template management, first-draft writing. Useful, but not yet central to the critical workflow. Where it is starting to creep closer is comparables. We are now seeing tools that do not just aggregate comparable data but interrogate and analyse it, which is the first time AI has really pushed into the judgment piece. That matters, because a comparable is never just a sold price. It is which end of the street it sat on, what condition it was in, what year it was built. Those soft and hard signals are what a good valuer is paid to weigh.
The graduate problem nobody wants to say out loud
Roughly 80% of a year-one graduate's work, collecting, drafting, basic admin, is exactly the kind of task that is ripe for automation. KPMG and Deloitte have already cut graduate intakes by 10 to 20%, and surveying is not immune. But here is the tension: that grunt work is also where graduates built their domain knowledge. If you never trawl through a 50-page lease, you never learn what to pull out of it.
The answer is not to protect the grunt work. It is to redesign the path. Graduates can move up the value curve faster, commit to specialisms earlier, get in front of clients sooner, and become genuine thought leaders in their vertical. But that only works if firms deliberately build their domain expertise through something other than repetition.
The burden didn't disappear, it moved
This is the part that gets missed. AI lifts the administrative load off juniors, but it creates a verification burden on senior, expensive staff. The trouble with an LLM is that it produces a near-infinite volume of output that looks polished and plausible, which means it has to be interrogated more, not less. Use AI to check AI to check AI, and you have simply moved the bottleneck to your costliest people. The fix is to build verification capability into the junior tier, not to dump a ream of plausible text on a partner at the end.
Deploy tactically, not broadly
There is a lot of snake oil out there. You can spend enormous time and money getting excited about tools that do not stick. What worked at LSH was unglamorous: map the entire workflow end to end, find the genuine pain points, and deploy against those specifically. No mass retraining, no disruption for its own sake. The result was a real improvement in work volume against headcount.
The pressure to do this is rising regardless. Loan-security valuation fees have compressed by roughly 15% over a five-year period, while lenders increasingly want scenario analysis, build-to-rent overlays, alternative use classes, that can triple the work on a single instruction. A lender can now feed your report into an AI and generate a barrage of challenging questions in seconds. Responding to that noise professionally is hard if you have not evolved how you work.
What this means in practice
Because the asset is physical, there will always be a human element: the inspection, the site visit, the client conversation. So the biggest near-term wins are not in the valuation itself but around it, billing, invoicing, KYC, onboarding, marketing collateral, event planning. A five or ten person firm starting today, with AI handling the periphery, could carry the overhead of a legacy fifty-person team.
On the judgment itself, nothing fundamental changes. When I ran a development appraisal recently by handing a site brochure to Claude and asking it to build the model in Aprao, the foundational point held: Aprao's formulas are locked down, so the financial calculations are identical every time, unlike an LLM that gives you a slightly different answer on each run. But the judgment, is this build cost realistic, is this a 12, 18 or 36-month build, is this a sloping site with expensive groundworks, is this sales value defensible, is still entirely yours. AI just spares you the manual data entry.
A closing thought
The honest conclusion is an optimistic one. There is real efficiency to be had, and the firms that embrace it will pull ahead. But the goal is not to remove the human, it is to point the human at the highest-value work. Use the technology to turn reports around faster and absorb rising demands, and use people where the judgment, the relationships and the expertise actually live. Be curious, dig in, and stop pretending it is going away. Denial is not a strategy.
Watch the webinar here
Watch and listen to the full conversation with Daniel Norman and Oliver Leeming.
