AI can check a contractor's interim application against the contract: arithmetic, percentage claims, variations without instructions, retention and previous payments. It can read the clauses you paste and test each line against them. It cannot see the site, so the quantity surveyor still verifies the work done and makes and signs the valuation.
The test below treats the chatbot as a junior who has read the contract clauses you give them and nothing else. Run it clause by clause and it behaves well. Ask it "is this application right?" and it behaves like a junior who wants to please you. The difference is entirely in how you frame the job.
Start with the clauses, not the application
Before the application goes anywhere near a chatbot, extract the five or six provisions that govern an interim payment. They are short, they are the whole basis of the check, and writing them out forces you to read them again. Paste them as a block and ask the chatbot to restate each in one plain sentence. If its restatement differs from your understanding, find out why before going further.
These are the clauses that govern interim payments under this
contract. Restate each in one plain sentence, and list for each
clause the figures or dates a valuation would need.
1. Payment: [paste clause]
2. Valuation of work: [paste clause]
3. Variations: [paste clause]
4. Materials on site: [paste clause]
5. Retention: [paste clause]
6. Preliminaries basis: [paste clause]
That gives you a checklist that is specific to this contract, rather than a generic one from a blank template. It is also the thing the generic templates that rank for this topic do not give you: they supply columns, and the contract supplies the rules that decide what goes in them.
Valuation six on a ten-month contract, clause by clause
Take a main contract with a sum of $1,250,000 over ten months, five per cent retention, monthly valuations, and valuation six due. The figures are illustrative.
| Item | Contract value | Claimed % | Claimed value |
|---|---|---|---|
| Preliminaries | 150,000 | 60% | 90,000 |
| Substructure | 200,000 | 100% | 200,000 |
| Frame | 300,000 | 80% | 240,000 |
| Envelope | 250,000 | 40% | 100,000 |
| Services | 200,000 | 25% | 50,000 |
| Finishes | 150,000 | 0% | 0 |
| Work total | 1,250,000 | 680,000 |
The application also claims three variations (VO1 $12,000 and VO2 $8,500, both with written instructions, and VO3 $15,000 marked "instruction awaited"), $22,000 for materials on site, retention of $36,875 on a gross of $737,500, and previous payments of $550,000. That gives $700,625 net and an amount now due of $150,625.
The prompt, run after the clauses are stated:
Check this contractor's interim application, clause by clause, against
the clauses above. For each of the six clauses, list: what the
application claims, whether it appears to comply, and what evidence
I should ask for. Quote the clause wording you rely on. Then recompute
the gross valuation, retention, net and amount due and list any
difference. Previous payments per my records: 570,000. Time elapsed
on the programme: 5 of 10 months. Do not value the work. If a point
depends on facts you do not have, say so.
[paste application figures]
An illustrative response flags five things. The preliminaries basis clause says preliminaries are valued on a time-elapsed basis, so at five months of ten the claim should be $75,000, not $90,000, a difference of $15,000. VO3 has no instruction and the variations clause requires one before a variation is included, so $15,000 should not be in this valuation unless an instruction arrives. The materials-on-site clause requires proof of ownership and insurance, and none is mentioned, so ask for it before allowing the $22,000. Previous payments are shown as $550,000 against your records of $570,000, which overstates the amount due by $20,000. Finally, the arithmetic otherwise agrees. It also notes it cannot check whether the frame is really 80% complete.
What you would fix: take out the excess preliminaries and VO3, correct previous payments, and hold the materials on site until the evidence arrives. If the ownership and insurance evidence comes in before you issue the valuation, the gross is $707,500, retention $35,375, net $672,125 and the amount due $102,125, against $150,625 claimed, a difference of $48,500. If it does not, the $22,000 comes out as well: gross $685,500, retention $34,275, net $651,225 and the amount due $81,225. The chatbot found the contract-driven issues in minutes. It did not touch the QS's real work: measuring on site whether the frame, the envelope and the services stand at the claimed percentages. That stays with you, and in practice is where the largest sums sit.

The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowThe clauses where the chatbot helps most
Preliminaries
Preliminaries (site set-up, management, welfare, insurances) may be valued as a fixed proportion, on time elapsed, or against stated milestones. The contract says which. The chatbot spots mismatches between the claimed percentage and the stated basis, as in the example. Where the basis is "as the works progress", you have more discretion, and the chatbot cannot decide it for you.
Variations
The checks are mechanical: is there an instruction number, a date, a described scope, and a basis of pricing (agreed lump sum, daywork, measured at contract rates)? Ask the chatbot to build a small table of the claimed variations with those columns and a "missing" flag. It will mark empty cells quickly, which is where the argument starts. For how variations are built in the first place, see how small builders price variations and change orders.
Materials on site
Typical conditions are that the materials are for the works, properly stored, insured and, in some forms, ownership has passed to the employer on payment. The chatbot can list which conditions apply and which evidence the application lacks. It cannot tell you whether the materials are actually on site, and a site walk with a phone camera is the check.
Retention and previous payments
Retention rate, the base it applies to, and any reduction at practical completion are contract terms. Previous payments should match the certificates actually issued, not the contractor's recollection. Give the chatbot your payment log and the application's figures and ask for a reconciliation, line by line, with differences listed.
Frame claims, trends, wrong clauses and subcontract claims
Turning 'frame looks high' into a specific query
Before: "Frame looks high." After: "The frame is claimed at 80% ($240,000). The programme shows steelwork erection complete on three of five zones and decking complete on one, which on the pricing document is worth around 55%, so ask the contractor for the stage breakdown behind 80%." Ask the chatbot to turn your site notes into the second form. The judgement is yours; the drafting is quick.
Comparing two valuations in a row
Paste valuations five and six side by side and ask for the biggest movements by item, as points and as dollars. A chatbot shows at once that envelope jumped 25 points and services only five, which prompts the question whether services are being under-claimed because a subcontractor is late, or the envelope over-claimed to catch up on cash flow. It is a conversation starter with the contractor, not a conclusion.
The variations clause pasted from the wrong contract
An illustrative slip: a surveyor pastes the variations clause from the previous project's contract, which had a different procedure for pricing daywork. The chatbot dutifully tests the application against it and reports everything compliant. The error shows only when the contract administrator asks why daywork was accepted without sheets. The fix is a one-line habit: put the project reference and clause numbers at the top of every paste, and ask the chatbot to quote the clause number in each finding so a mismatch is visible.
Subcontractor applications under a main contract
The same method applies one level down. The subcontract usually mirrors the main contract with different periods, rates and set-offs. Paste the subcontract clauses, not the main ones, and add a column for backcharges and any set-off agreed in writing. For comparing subcontract prices before award, comparing subcontractor quotes with AI covers the earlier stage.
Releasing half the retention at practical completion
Say practical completion is certified with the gross valuation at $1,320,000 and five per cent retention held, $66,000. If the contract releases half at practical completion and the rest at the end of the defects period, then $33,000 is released now. Ask the chatbot to set out the release schedule from the clause you paste, then verify it against the actual wording. A sum like this is easy to get wrong when retention was reduced earlier on a changed rate, which is why you reconcile it to the certificates rather than recompute from the contract sum.
Build your own valuation sheet instead of downloading a template
A blank interim valuation template is a set of columns. The columns are not the hard part. The hard part is knowing which columns this contract needs and what each must contain, and a template written for no contract in particular cannot tell you. So build your own, once per project, from the clauses you extracted at the start. Ask the chatbot to propose a layout and then edit it.
From the six clauses above, propose a valuation worksheet layout for
this contract. List the columns, what each should contain, which are
inputs and which are formulas, and the order of the final summary
(gross, less retention, less previous payments, amount due). Then
list the evidence I should attach to each valuation.
An illustrative layout for the example contract: one row per item of the contract sum; columns for contract value, cumulative percentage, cumulative value, previous cumulative value and this period; a separate block for variations with instruction number and basis of pricing; a block for materials on site with the evidence status; a summary that takes the gross, applies retention to the gross, subtracts previous payments actually certified, and shows the amount due. You then build it in your spreadsheet with formulas, so the arithmetic is mechanical and the chatbot's job is to audit it. If you do many projects on the same contract form, keep the finished sheet as your own template, which is more useful than any blank download because it already encodes the rules.
Working through the contractor's supporting documents
An application rarely arrives alone. There is usually a progress report, a variations log, delivery notes or invoices for materials, subcontractor applications and sometimes photographs. The chatbot can help you read these quickly if you give it one job at a time. A few examples follow.
For the progress report, paste the narrative and ask for a list of every activity it says is complete, in progress or not started, then compare that list against the percentages in the application. Where the narrative says "steelwork erection ongoing" and the application claims 80% for the frame, you have your first site question. For the variations log, ask for a table of every variation with its status (instructed, priced, agreed, pending), then check each claimed variation against the status column. For delivery notes, ask for a total by item and compare with the materials-on-site claim, flagging items delivered before the valuation date and those delivered after. A claim for goods whose delivery note is dated the day after the valuation date is a common, innocent error and a common point for correction.
Photographs are the one place where you should be careful. A chatbot can describe what a photograph appears to show, but it cannot confirm location, date or quantity, and it will sound sure while doing so. Use photographs as prompts for your own site visit, never as evidence the chatbot has verified.
A different sector: a fit-out with a front-loaded schedule
Consider a commercial fit-out where the contractor's schedule has a large "design and procurement" line of $180,000 out of a $900,000 contract, claimed at 90% by month two because long-lead joinery has been ordered. The question is whether that line represents work actually done or money drawn forward. Paste the pricing document and the relevant clause and ask the chatbot to compare the line against the stated basis and list what evidence would support 90%: purchase orders, supplier acknowledgments, shop drawings approved. It will produce the list, and you ask for the documents. Where the contract allows payment on order only against a bond or ownership transfer, the chatbot can tell you the clause says so; whether the joinery supplier has actually received the deposit is something only the paperwork will show. This is the same pattern as the main example: the chatbot reads the rules, you gather the facts.
Timing and the clock on your response
Most contracts set a period for you to issue the valuation or certificate after the application date and prescribe what happens if you are silent or late. The detail varies and is a legal matter, so check your contract and, where there is doubt, take advice. What the chatbot can do is build a simple timeline: application received date, last date for your valuation, last date for payment, and the notice dates in between, all drawn from the clause text you paste. Diarise those dates yourself, because a date a chatbot calculated is a date you have not yet verified. For notices, see drafting payment notices and checking the deadlines, and for the contractor's side of the same document, checking a payment application before it is sent covers how builders prepare the claim you are now reviewing.
Feedback that improves the next application
A valuation that returns a smaller figure than claimed invites a call. Use the chatbot to turn your findings into a short, neutral letter and into a list of what the contractor should include next month. Over a few valuations, the repeated causes of error (percentages typed by hand, variations without instructions, missing stored-goods evidence) usually point to a gap in how the contractor builds the application. A short checklist sent with the first valuation saves everyone time on later ones. Ask the chatbot to draft it from your own findings, then edit it so it matches the contract and your voice.
A written-up check you can file
The record of the check is part of the valuation file. Ask for a one-page summary with the headings Clause, Application claims, Finding, Evidence requested, and Action, then edit it yourself. The summary becomes your note to the contract administrator and to the contractor. A short worked version for the example reads:
Preliminaries - claimed 90,000 - basis is time elapsed (5/10 of
150,000 = 75,000) - request none, adjust by 15,000.
VO3 - claimed 15,000 - no written instruction - exclude pending
instruction.
Materials on site - claimed 22,000 - no proof of ownership or
insurance - exclude pending evidence, or request it now.
Previous payments - shown 550,000 - certificates show 570,000 -
correct to 570,000.
Notice the list stays factual. It states what the contract says and what the application shows. It does not accuse, and it leaves the door open for the contractor to supply the missing evidence. The chatbot helps you keep that tone if you ask it to rewrite a blunt finding neutrally.
What stays with the quantity surveyor
- The measurement or assessment of work properly executed, which needs eyes on site or reliable progress evidence.
- The professional judgement on percentages, defects and incomplete work.
- The valuation itself and the certificate or notice that follows, with your name and the contract's deadlines.
- Any dispute about entitlement, extension of time or loss and expense.
The practical sign of getting this right is that your valuation reads as yours: you can defend every figure in it without referring to a chat window. If you cannot, you have let the tool lead. For a wider view of where AI sits in a surveyor's week, how AI can help quantity surveyors covers takeoffs, cost reports and checks of this kind, and the closing step of the job is walked through in the final account worked example.
One last habit is worth building in. After each valuation, note which of the chatbot's flags you agreed with, which you rejected and why. A few months of those notes show you where the prompt is too sensitive and where it misses things, and you can tighten the wording accordingly. A flag you rejected twice for the same reason is a prompt line to rewrite; a point you found yourself that the chatbot missed is a prompt line to add.
Confidential valuations, privacy settings and the RICS standard
Remove the employer's name, the project address, commercial terms you do not need and any bank details before pasting. A table of figures and a handful of clauses is enough for this check. Consumer plans have a privacy switch that stops your chats being used for model training, and business plans generally exclude business content from training by default. Check the setting you are on, since valuations are confidential commercial documents, and write the practice down in your office procedures so every colleague works the same way.
If you are regulated by RICS, read its professional standard on the responsible use of AI in surveying, in force since 9 March 2026. Among other things, it expects a written register of the AI systems you use (what each is for, when it was first used and when it is next reviewed), a risk register reviewed at least quarterly, and written notice to clients in advance where AI is used on their work. A chatbot that checks valuations belongs on that register, and the notice belongs in your terms of engagement. Read the standard itself rather than a summary, this one included.
Interim valuation checks: common questions
Can AI produce the valuation itself?
It can draft a worksheet from your inputs, but the valuation is a professional judgement about work properly executed, and it carries your name. Use AI to check the contractor's application and your arithmetic, then make and sign the valuation yourself.
Should I paste the full contract into the chatbot?
Paste only the clauses that govern the valuation: payment, retention, variations, materials on site, and the prelims basis. That gives a tighter check, keeps confidential terms out, and makes it easy to confirm each flagged point against the actual wording.
What if the chatbot disagrees with my reading of a clause?
Reread the clause and the definitions it relies on. If you still think you are right, you probably are, since it works from a paste with no knowledge of amendments or the employer's requirements. If the disagreement exposes ambiguity, that is worth raising with the contract administrator.
Is a blank template enough to start with?
A template gives layout, not judgement. Build your own sheet from the contract's structure, with a column for each item the contract lets the contractor claim, and use the chatbot to check the contractor's figures against it.
Further reads
- Best AI for Quantity Surveyors: Five Tool Types Compared by Job — Which tool types suit which QS job.
- RICS AI Standard for Small Firms: Register, Assessment, Client Notice — Set up the AI register and client notice the standard expects.
- Bill of Quantities Template: A Worked Example Checked With AI — Check the priced document your valuation rests on.
- Construction Retainage: Track It and Chase Release With AI Reminders — Track retention held and due for release.
- AI Tools for Quantity Surveying: Take-Offs, Cost Plans and Checks — Take-off and cost-plan tools for the same practice.
- AI Tools for Contracting Businesses: Subcontracts and Change Orders — How the contractor side manages subcontracts and changes.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: general quantity surveying practice for interim valuations (percent complete, variations, materials on site, retention, previous payments); ChatGPT and Claude help pages on document upload and data controls; RICS professional standard on responsible use of AI in surveying (1st edition, effective 9 March 2026).