Research note · September 2026 · about 6 minutes · this is the thinking behind what Cutting Hedge builds
AI can write a quote that looks brilliant. That’s exactly the problem. A quote isn’t a piece of writing — it’s a price you’re committing to, and a nicely worded wrong number is still a wrong number.
The short version
- Construction professionals see real potential for AI in cost and document work — but about 45% of organisations use none, and under 1% have it embedded.
- AI is good at organising messy notes and spotting what’s missing. It shouldn’t be setting your price.
- The safe first step is quote preparation, not automatic quoting. You still price it and you still send it.
Why quoting looks like an obvious fit
A good quote has to account for what the customer asked for, the site, materials, labour, travel, waste, risk, VAT, exclusions and when you can actually do it. And it has to be clear enough that the customer knows what they’re getting.
Parts of that suit AI well: tidying up notes, summarising an enquiry, spotting what’s missing, pulling up similar past jobs, producing a first draft. But the price itself is professional judgement. So the real question is:
Can AI get a quote out the door faster without making the price less reliable? Possibly — but only if it’s working for you, not instead of you.
What construction research says
The Royal Institution of Chartered Surveyors gathered more than 2,200 responses for its 2025 report on AI in construction. Asked where AI could make a big difference, respondents pointed to progress monitoring and project scheduling (36% each), resource optimisation and reviewing contracts and documents (30% each), risk management (29%) and cost management (25%).
Actually using it is another matter.
| Stage of AI use (RICS, 2025) | Share |
|---|---|
| No AI implementation | ~45% |
| Early pilots | 34% |
| Regular use in specific processes | just under 12% |
| Use across multiple processes | 1.5% |
| Embedded across the organisation | less than 1% |
The barriers RICS found go straight to the heart of quoting: lack of skilled people (46%), fitting AI in with existing systems (37%), data quality (30%) and high implementation costs (29%). A price is only as good as the information behind it.
A January 2026 report from the AI4QS initiative at Birmingham City University, looking at AI in quantity surveying, reaches a principle worth borrowing. It flags bias, transparency, accountability and professional integrity as real challenges, and says the goal is technology that enhances human expertise and ethical judgement rather than replacing it.
What AI can realistically do in a quote
This is very different from typing “price this job” into a chatbot. A useful setup shows its assumptions and tells you what it doesn’t know.
| Stage | Where AI helps | What stays with you |
|---|---|---|
| The enquiry | Summarises the request and lists what’s missing | Confirm the job, location, urgency and scope |
| Site information | Organises voice notes, photos, measurements, documents | Check it’s accurate, complete and current |
| Scope of work | Turns notes into a structured list | Confirm inclusions, exclusions and assumptions |
| Costs | Pulls up your own rate card or similar past jobs | Check quantities, labour, materials, waste, VAT, margin |
| Risk | Flags unusual conditions or gaps | Decide if you need a site visit or a specialist |
| The customer document | Drafts a readable quote and cover note | Approve every price, promise, technical statement and date |
| Learning | Compares what you quoted with what the job cost | Update how you quote — without copying old mistakes |
Why a quoting mistake costs more than a typo
A typo in a Facebook post is embarrassing. A wrong quantity or a missing item in a quote eats your margin, starts an argument, delays the job, or creates a safety problem. The usual ways it goes wrong:
- misreading a photo or drawing
- leaving out prep, access, protection, skip hire or finishing
- old material prices or labour rates
- mixing up a provisional sum with a fixed price
- getting the VAT wrong
- an unrealistic finish date
- missing a site condition that needed a proper look
- carrying an assumption over from an old job that doesn’t apply to this one
AI writes fluently even when the assumption underneath is wrong. A polished-looking quote can make everyone more confident in a number nobody checked. That makes your review more important, not less.
Simple rules, borrowed from the big frameworks
The US National Institute of Standards and Technology publishes a voluntary framework for managing the risks of generative AI. You don’t need to run a compliance department. But the core of it boils down to a handful of questions any trades business can answer.
| Rule | The question |
|---|---|
| A clear job | Which parts of quoting is it allowed to help with? |
| Approved inputs | Which rate cards, photos, documents and notes can it use? |
| Checking | Who checks quantities, prices, scope, VAT and assumptions? |
| Sign-off | Who approves the quote before the customer sees it? |
| A record | Can you see the original notes and what was changed? |
| Follow-through | Are you comparing what you quoted with what the job actually cost? |
| An escape hatch | What happens when information is missing or the job is unusual? |
The point is simple: a draft should never quietly turn into a commitment.
A safer place to start
Not automatic quoting. Quote preparation. Say you’ve just walked round a job and you record a two-minute voice note in the van. From that, a supervised setup could give you:
- a tidy summary of what the customer wants
- a list of the rooms, materials and tasks you mentioned
- a checklist of what you still need to find out
- possible exclusions to confirm
- a plain-English description of the work for the customer
You still decide whether it needs a proper survey, what materials to use, how long it’ll take, and what to charge.
How to know if it’s helping
“The quote looks more professional” isn’t the test. These are.
| Measure | The question |
|---|---|
| Speed to customer | How long from site visit to the customer having a quote? |
| Jobs won | Are clearer or faster quotes getting accepted more often? |
| Margin | Are your margins holding up more consistently? |
| Accuracy | How close is the quoted cost to the actual cost? |
| Return visits | How often do you have to go back or ring for missing details? |
| Corrections | How many AI mistakes did you have to fix before sending? |
| Disputes | Are there more or fewer arguments about scope or price? |
If it gets a quote out ten minutes faster but leads to more disputes or thinner margins, it isn’t working. Judge the whole job, including the time you spend checking.
What should never be automated
AI shouldn’t diagnose a safety-critical problem, promise compliance, guarantee a completion date, approve a price, change a contract, or give technical advice outside your competence. Customer photos and documents can also contain personal information, so find out how any tool stores and uses what you upload before you upload it.
The rule: AI can prepare, organise, compare and flag. You decide the price, approve the quote, and send it.
What this research can’t tell you
The RICS survey went to construction professionals worldwide, mostly at larger firms. The AI4QS report is about quantity surveyors, not sole traders. The NIST framework is general guidance, not evidence that any quoting tool works. There’s no published study I can find measuring whether AI-assisted quoting wins more work or protects margins for small Irish trades businesses. Anyone telling you it saves a set percentage of time is making it up.
Tell me how you quote
I’m running a short survey asking local tradespeople which parts of the admin cause the most trouble — including quotes, site visits, and following up with people who ask for a price but don’t book. It takes 6–8 minutes, and leaving your contact details is optional.
Keep reading
- Can AI reduce the burden of construction documentation?
- What AI is actually worth paying for in a trades business
- The human side of AI in the trades: how trust is built
- 10 practical ways AI could help Irish trades businesses
Sources
- RICS — Artificial intelligence in construction report (2025)
- Saka et al. — AI for Quantity Surveying Report: Exploring Impact, Building Competence, and Advancing Responsible Use (Birmingham City University / AI4QS, January 2026)
- NIST — Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (NIST AI 600-1, July 2024)