Aipedia: Quantity surveyors

AI order-of-cost estimates and benchmarking: the QS risks

Aipedia, by the Apex broking team · Last reviewed 2026-10-30

A quantity surveyor can use AI to produce order-of-cost estimates and benchmarks, but the figure is still the QS's advice and must be checked before a client relies on it. Early estimates drive land purchases and funding, so an AI figure that proves too low can cause losses far larger than the fee.

Part of: AI in quantity surveying

In short

AI estimating tools build feasibility and order-of-cost estimates from historic cost data, applying location, inflation and specification adjustments automatically. The risks are factors that have not been properly adjusted, benchmark datasets the QS cannot see or verify, and a client buying land or agreeing funding on a figure that turns out too low. Under the RICS AI standard, the QS needs written supplier due diligence (4.1) and a named surveyor's written reliability decision (4.2). An estimate claim is usually treated as negligence under RICS-compliant PI, subject to the wording and the scope of the duty.

How QSs use AI for early estimates

At feasibility stage, design information is thin. AI estimating tools fill the gap by matching a proposed scheme to similar historic projects, applying cost per square metre or elemental benchmarks, then adjusting for location, inflation, building type and specification. Some tools produce a full elemental order-of-cost estimate in minutes from a short description and a floor area. Others are used to benchmark a QS's own estimate against a wider dataset.

The speed is useful. The difficulty is that the QS often cannot see which projects sit behind a benchmark, how old they are or how the adjustments were made. Measured take-off from drawings is a separate task, covered at AI take-off and cost plans.

The duties that apply

These duties apply whether or not AI is involved:

Material impact (1.2): an order-of-cost estimate is advice a client relies on, so record the material-impact determination for the tool.
Supplier due diligence (4.1): ask the vendor in writing about the source, age and make-up of its cost data and how location and inflation factors are derived. Record the answers.
Reliability (4.2): a named, appropriately qualified surveyor applies professional scepticism and records whether the estimate is reliable enough to issue.
Client information (4.3): tell the client in advance, in writing, that AI benchmarking is used.
Rules 2 and 3: the RICS Rules of Conduct require competence and a diligent service, whatever tool is used.

See the RICS AI standard applied to QS work for how each section of the RICS AI standard applies to cost tools.

What can go wrong

Common ways this goes wrong include:

Unadjusted factors: a tool applies a national average where a regional factor was needed, or uses an inflation index that stops at the base date.
Opaque benchmarks: the dataset mixes building types, procurement routes or specifications the QS would never have chosen as comparators.
Missing scope: abnormals such as demolition, contamination, utilities or site constraints are not in the benchmark at all.
False precision: an estimate produced to the nearest pound looks more certain than the information supports, and risk allowances are trimmed.

Illustrative: a developer asks a QS for an order-of-cost estimate to support a land bid. The QS runs the scheme through an AI benchmarking tool and issues the result with a short covering note. The benchmark draws on older projects in a cheaper region and leaves out significant abnormals. The developer buys the site and agrees funding on that figure. When the scheme is tendered, costs are well above the estimate and the scheme is no longer viable. The developer claims its losses from the QS.

How PI usually responds

A claim that an estimate was negligently low is usually handled as professional negligence under the full civil liability cover of the RICS Approved Minimum Wording, subject to the policy terms. The use of AI is not, on its own, a reason to decline: the endorsement in the wording says cover is not restricted solely because a computer system was used.

The bigger questions are often liability and quantum: whether a reasonably competent QS would have produced the same figure, what the client was told about the estimate's limits, and which losses fall within the scope of the duty. Our guide to QS cost overrun disputes covers these arguments. RICS-regulated firms must also hold at least the turnover-based minimum limit in the RICS PII Requirements, which may be low against a land purchase claim.

Practical controls

Put these controls in place:

Record the tool, its data sources and vendor answers on the AI register.
Check the comparator projects behind any benchmark where the tool allows it, and note where it does not.
Adjust location, inflation and specification factors yourself, or check the tool's adjustments.
List abnormals and exclusions separately, outside the AI output.
State the basis, date and limits of the estimate in the covering report, including that AI benchmarking was used.
Have a named QS sign off the estimate before issue.

How Apex helps

Apex arranges PI for quantity surveyors and cost consultants. A broker can check whether your limit fits the decisions clients make on your early estimates, and how your policy treats AI-assisted cost advice. See also AI in funder monitoring reports, AI whole-life costing and the quantity surveying AI hub.

Sources

Frequently asked

Can I rely on an AI cost benchmark if the vendor will not show the data?

You can use it, but the RICS AI standard expects written due diligence (4.1) and a named surveyor's reliability decision (4.2). If the data cannot be seen, record that limitation, test the output against your own knowledge and tell the client how the figure was produced.

Is an order-of-cost estimate held to the same standard as a pre-tender estimate?

The test is what a reasonably competent QS would have done with the information available at that stage. Early estimates are expected to carry more uncertainty, which is why stating the basis, exclusions and limits clearly matters, whether or not AI was used.

Will my PI insurer ask about AI estimating tools?

Some insurers may begin asking at renewal, and RICS terms of engagement now need to mention PI cover for AI use where available. Answer proposal questions accurately and describe how outputs are checked, because non-disclosure can affect cover.

Check your PI covers how you use AI

Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.

Start a PI quote Request a callback

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.