How to Price Your Trade Services Using AI and Real Market Data (UK 2026 Guide)

ChatGPT

Claude

Google Sheets

Microsoft Excel

Checkatrade

Payaca

Why trades businesses keep underpricing in 2026

Most sole traders still price by feel. That is where the margin leaks.

Ask ten UK plumbers, electricians or builders how they priced their last big job. About four will tell you they used a number they had in their head from a similar job two years ago. Three will tell you they matched what they think their competitor charged. Two will admit they undercut the last quote because they wanted the work. One, maybe, will say they costed it properly.

That is not laziness. It is the maths catching nobody in a sole-trader business. There is no quoting team, no commercial manager and no time at 9pm on a Sunday to model overheads when you have got a 7am start. So you guess. And the guess is almost always low.

The Federation of Master Builders said 98 percent of UK builders absorbed material price increases in 2025 and 81 percent passed them on, which means roughly one in five firms ate the cost themselves. FMB's State of Trade survey shows the same pattern every quarter: workloads up, margins squeezed. The thread running through both: businesses that price defensively lose first. We covered the financial distress numbers in detail in our warning-signs guide for distressed construction firms.

AI changes the time problem. It does not change the maths. You still need the inputs (cost base, region, market rate, job complexity), but the work of comparing them goes from two hours to fifteen minutes. That is the difference between a priced job and a guessed one.

The data you need before you open an AI tool

Bring the receipts. AI cannot price a job you have not measured.

This is where most tradespeople trip up. They open ChatGPT, type "what should I charge for a boiler install?" and get a generic answer pulled from blog posts. The output is only as good as the inputs. Before you ask any AI tool for a price, you need five things on hand.

One. Your true hourly cost. That is your annual overhead (van, fuel, insurance, registrations, accountancy, tools, training, holiday pay, NI, pension) divided by your realistic billable hours. Most sole traders bill 1,200 to 1,400 hours a year out of a possible 1,920. Use the lower number if you are honest.

Two. Material markup baseline. The Screwfix Community and UK Plumbers Forums settle around 15 to 25 percent. Some specialists go higher. Decide yours and stick to it.

Three. Local market rates for your trade. Checkatrade and MyBuilder publish these. Checkatrade's 2026 cost guide shows electricians at £45 to £60 per hour and plumbers at £40 to £70, with London 30 to 45 percent above the national midpoint and the North East and Wales 15 to 20 percent below.

Four. Your last twelve months of actual job data. Quoted price, time taken, materials used, final invoice. If you have not been tracking this, start now. The most useful thing AI can do is spot patterns in jobs you have already completed.

Five. Your target net margin. Most healthy trades businesses run 15 to 25 percent net. If you do not know yours, default to 20 percent and work back.

Using ChatGPT for competitive market analysis

ChatGPT's strength is taking five fragmented data points and turning them into a defensible market position in under a minute.

ChatGPT is best at the part of pricing that involves the outside world: what your competitors charge, what customers in your area expect, what a job like the one in front of you typically costs. It is a research assistant who never sleeps and does not bill you by the hour.

The trick is giving it real data to chew on rather than asking it to guess. You can paste Checkatrade cost guide tables, screenshots from MyBuilder, anonymised quotes you have lost, and ask it to identify pricing patterns in your area. The current model (GPT-5.x at time of writing) handles tables, regional breakdowns and percentile analysis well. It runs about £20 a month for ChatGPT Plus, which is one decent quote on a single job.

What it is good at:

• Breaking down a competitor's quote line by line to expose where they are cutting corners (or cutting their own margin)
• Pulling regional pricing benchmarks from Checkatrade, MyBuilder, Trades Day Rates and Hamuch into a single comparison table
• Drafting customer-facing price justifications that explain why your number is what it is
• Spotting price tiers in your historic job data you did not know you had (typically a "small fix" tier, a "standard install" tier, and a "complex multi-day" tier)
• Suggesting tiered pricing language for proposals (good, better, best) without sounding salesy

What it is bad at:

• Anything specific to your supplier prices, your van running costs, or the patch you actually cover
• Live data that requires looking up today's price at City Plumbing or Edmundson Electrical
• Knowing whether your customer is a friend, a repeat client, or a one-off who wants the cheapest number

The line is clear. Use it to research the market and structure your thinking. Do not use it to set the final number on its own.

Using Claude Cowork to crunch your own job data

Claude Cowork sits inside the spreadsheet so it sees your real numbers, not a generic estimate.

Where ChatGPT shines on market analysis, Claude Cowork is the right tool for analysing your own job data. Anthropic released Cowork in late 2025. It embeds Claude directly into Excel, Google Sheets, Slack, Gmail and Google Drive with a shared context layer, so it can read across all of them at once. For trades, the spreadsheet integration is the part that matters.

If you have twelve months of jobs in a Google Sheet (date, customer, job type, quoted price, time taken, materials, invoiced total, paid/unpaid), Cowork can run the analysis you have never had time to do. Things like: which job types are eating your time without paying for it, which customers consistently take three follow-ups to pay, what your true effective hourly rate looked like for each month, and where your pricing is drifting low. We walked through a full example of this kind of workflow in our Claude Cowork weekly admin guide.

Cowork runs on Anthropic's enterprise tier. For solo traders it is overkill. For firms with a few engineers and a back-office person it pays for itself in saved hours within a month, assuming you have your data in one place. Claude.ai Pro at around £18 a month still gives you the underlying model for ad-hoc analysis if you upload your spreadsheet to a conversation.

What it is good at:

• Reading a workbook with twelve tabs and explaining what each one tells you about pricing health
• Finding the jobs you priced low (compared to the patterns in your own historic data) without you needing to write a single formula
• Variance analysis between quoted and invoiced totals, with cell-level citations so you can verify
• Writing the narrative section of a monthly pricing review for your accountant in your own words

The five-step AI pricing workflow

The same five steps work for a small fix and a multi-week extension. The numbers move; the order does not.

This is the workflow I run every time a new pricing question lands. It takes 20 to 30 minutes for a standard job and an hour for a multi-day project. Compare that to the alternative of pricing in your head while sat in the customer's kitchen.

Step 1: Capture the scope properly. Use your phone to record voice notes on site or take photos of the work. Drop them into ChatGPT (Plus accepts images and audio) and ask it to write a structured scope of works. Five minutes. You now have the same scope description you would have spent 30 minutes typing out.

Step 2: Benchmark the market. Open ChatGPT with your saved pricing context. Paste Checkatrade and MyBuilder rate ranges for your trade and region. Ask for the realistic price band for a job of this scope. You will get a low, mid and high number.

Step 3: Cost the job from your numbers. Take the scope and your true hourly cost into Claude or a spreadsheet. Itemise labour hours, materials with your standard markup, and add your overhead recovery and target margin. This produces your bottom-up price.

Step 4: Compare and decide. Hold your bottom-up price against the market band. If you are sitting at the low end of the band, you are leaving money on the table. If you are above the high end, you need a clear value reason (faster turnaround, higher accreditation, planned maintenance bundle). If you are in the upper-mid range, that is the sweet spot for most trades.

Step 5: Write the quote in plain language. Ask ChatGPT to draft the customer-facing quote using your structured scope, your final price, and a short justification. Three paragraphs, no jargon, no salesy padding. Read it, edit anything that does not sound like you, and send. The whole cycle, scope to send, can be done inside 30 minutes for a standard job.

Prompts to copy and paste

A handful of prompts you can reuse. Save them in your phone notes app and stop typing the same brief twice.

These are the four prompts I would print and laminate if I were starting again. Paste them into your AI tool of choice. Swap in your own numbers where the brackets are.

The benchmark prompt.

The pattern-finder prompt. Paste your last twelve months of jobs as a table.

The competitor-quote breakdown. Paste a competitor quote you lost (anonymised).

The customer-facing quote draft.

For more on how this connects to the wider AI-first business model, our piece on why operating costs decide who survives goes deeper on the maths.

What AI cannot do for your pricing

The price gets agreed in a conversation, not in a chat window. AI gets you to the conversation faster.

This is the bit that nobody selling you AI tools wants to mention. There are four things AI is bad at when you are pricing trades work. Knowing where the line sits matters more than knowing all the clever things it can do.

It cannot judge the customer in front of you. A nervous first-time homeowner needs a different conversation than a landlord with eight rentals. AI does not see body language, does not hear the catch in someone's voice when they hear a number, does not know that the customer's neighbour just paid more for the same work. You do.

It cannot price for relationship value. A 20-year repeat customer with three other properties is worth a different price than a Checkatrade enquiry. AI will give you the technically correct number. You have to decide whether the relationship justifies a discount, a premium or holding the line.

It cannot live-quote at your supplier. Today's copper pipe price at your nearest City Plumbing branch is what it is. Yesterday's training data is not. Always cross-check materials against a live supplier price before sending a quote that depends on volatile stock.

It cannot stand by the quote when the customer pushes back. That is the bit only you can do. The maths gives you the confidence. The conversation closes the work. The numbers are useful precisely because they make the conversation easier, not because they replace it.

What tradespeople and pricing experts are saying

Pricing is the most-discussed topic on UK trades forums. The pattern of advice is consistent across platforms: stop competing on price, build your costs from the ground up, and learn to walk away from work that does not pay. Here is what the most-cited sources actually say.

Recommended videos

A short selection of videos worth your time on AI-supported pricing, quoting and trades business pricing fundamentals.

How to use ChatGPT For Your Trades Business

Profitable Tradie - Business Coaching

The EASIEST way to get more quotes converted

Paul Tinker

Sleepless Tradesman: Quote Jobs in 60 Seconds

Sleepless Tradesman

Why Aaron McLeish Talks About P&H Pricing

TWC

#59 Off The Tools With Aaron McLeish

Off The Tools with Wayne Bettess & Andrew Jones

How to Master a Niche: Aaron McLeish Interview

Value Pricing Academy

Frequently asked questions

My verdict