AI Quote Generators for UK Tradesmen: Built-In or ChatGPT?

Use the AI inside your job software if you want it to draft priced lines from your own job history; use ChatGPT or Claude if the problem is wording, exclusions and follow-up messages, because a general assistant has never seen your prices. Most tradesmen end up using both, and a three-job test shows which tool earns which job.

"AI quote" means four different things across the main job tools. ServiceM8 drafts a description and billing items from the job card and similar past jobs. Tradify generates a description or tidies text you have typed. Powered Now turns a few words into a quote description or a customer message. Fergus has an assistant that can build and review quotes and turn a supplier's PDF into one. Only some of those touch the numbers, and that decides what is left for a general assistant to do.

Follow me on Instagram@sagnikteaches

What the four job-software tools do with a quote

I read each vendor's own pricing and help pages on 29 October 2026. The names below are the vendors' own labels, and the right-hand column records what each page does not say, which matters as much as what it does.

Connect on LinkedInSagnik Bhattacharya

The ServiceM8 threshold deserves a sum. A firm completing 25 jobs a month reaches 100 jobs in four months, so a new account gets a weaker tool at first, and a firm that has used the software for years gets a better one. ServiceM8's marketing page says 30% of invoices drafted with Auto-Invoice are "accepted for use". The page doesn't define that phrase, but it is a fair warning to expect edits on most drafts. The detail on how it works is in ServiceM8's Auto-Quote help article.

Subscribe on YouTube@sagnikteaches

That article's own tips are the practical lever, because the draft is only as good as what the job card holds. Add job notes as you go, track your time on site, send supplier invoices into ServiceM8, and record a short video with a voiceover, which it transcribes into the job diary. None of that is clever. It is the same habit of writing down what happened that makes a human estimator accurate, and the tool simply rewards it.

What a history-based draft gets wrong is predictable, because it borrows from the nearest past jobs. An illustration: the job card says "swap two kitchen sockets for USB sockets; customer has bought them; about an hour". A draft built from similar past jobs could easily read:

Supply USB double socket             2      $38     $76
Labour                               1 h    $55     $55
Testing and electrical certificate   1      $85     $85
Total                                               $216

Two lines came from the history, not from this job. The customer bought the sockets, so the supply line goes. The certificate line sat on the past jobs because those added circuits, and whether a like-for-like swap needs one is your call, not the tool's. Edited, the quote is $55 of labour plus whatever paperwork you normally issue for a swap, and the draft has still saved you typing the description. Expect that pattern on most drafts: right shape, a line or two inherited from a different job.

Tradify's pricing page sums SmartWrite up as generating descriptions and improving writing with AI, and that fits the job it does: it makes your quote read better and doesn't decide what goes in it, so a wrong line stays wrong, just better worded. Read every draft before saving, as with any of these tools. If you want a side-by-side of the two products for electricians and plumbers specifically, Tradify against Fergus for plumbers and electricians goes through their AI features, and the ServiceM8 piece for a small heating firm asks whether its helpers pay back.

What a general assistant adds, and what it cannot know

ChatGPT, Claude and Gemini are better at the writing side than most built-in tools, because you can tell them exactly what you want: a scope in plain words, three options at different prices, a list of exclusions, a polite note explaining why the quote is higher than the neighbour's, a follow-up after a week of silence. They also handle messy input, such as a voice-note transcript, without needing a form.

They know nothing about your business unless you tell them. That means three gaps: your rates, your past jobs and your customer record. The first is fixed by pasting or uploading a rate sheet. ChatGPT and Claude both have projects that keep files and instructions across chats (ChatGPT Projects hold 25 files on Go and Plus, five on Free), so you set the rate sheet up once. The other two gaps stay open, which is why a general assistant will never spot that you quoted the same customer last spring. ServiceM8 lists a ChatGPT and Claude connector on its pricing page that lets those assistants look up job details (ChatGPT's own name for connectors is now apps), which narrows the gap if you already use that software. Arithmetic is the other weak point, and why AI is bad at maths and what to check in quotes explains how it shows up.

Here is a prompt that keeps an assistant to your numbers, followed by what an assistant might realistically return. The job and figures are illustrative.

You write quotes for a small electrical firm. Use ONLY the rates below. If something isn't in the list, write "price needed" instead of guessing.

RATES
Labour: $55 an hour
10-way consumer unit, supplied: $210
Testing and electrical certificate: $85 fixed
Making good plaster and decorating prep: $60 fixed

JOB
Replace the consumer unit in a three-bed semi. Estimated 5 hours. Customer wants the old unit removed and the cupboard left tidy.

Output: a table of line items with quantity, rate and line total, then the total, then four lines of assumptions and exclusions in plain language.

Item                                  Qty    Rate    Total
Labour                                6 h    $55     $330
10-way consumer unit, supplied        1      $210    $210
Testing and electrical certificate    1      $85     $85
Making good, plaster and prep         1      $60     $60
Waste removal                         1      price needed
Total                                                $685

Assumptions and exclusions
- Existing cables are in a condition to be reused.
- Redecoration is excluded.
- Any additional circuits are quoted separately.
- Work to be completed in one day.

Four problems, none of them dramatic. The assistant changed 5 hours to 6 and added $55 you never agreed; at 5 hours the labour is $275 and the total is $630, not $685. Its "price needed" for waste removal is good behaviour, because it didn't invent a rate. The first assumption is an engineering judgement about the existing cables that only you can make on site, and "completed in one day" is a promise you didn't give it. So the fix list is: restore the hours, answer the waste question yourself, and turn those two assumptions into questions for the site visit.

Where an assistant is at its best is explaining a price. A customer holds a cheaper quote from another firm, and you want a calm reply that sticks to what you are actually offering. The prompt gives it only the facts it may use.

A customer has two quotes: ours at $1,240 for a consumer unit replacement and another at $980. Write a short, calm paragraph explaining what our price includes, using only these facts: 10-way unit supplied, testing and certificate, making good, one-year workmanship cover. Don't criticise the other firm and don't claim anything not listed.

Our quote of $1,240 covers the unit itself, fitting it, full testing with the electrical certificate, and making good around the cupboard afterwards. It also includes a year of cover on our workmanship, so if anything we've done needs attention, we'll put it right. Quotes can look different when one leaves out the testing paperwork or the making good, so it's worth checking the other one line by line. We're happy to talk through ours.

Two edits. "We'll put it right" is stronger than "one-year workmanship cover", so match it to your real terms. The sentence about quotes that leave things out hints that the other firm cut corners, which you told it not to do, so delete it. What remains is a reply you can send in two minutes, and it never needed access to your price list.

Test both on three jobs you have already priced

Don't decide from a demo. Take three finished quotes where you know the real price and the real hours: a small job, a typical one and an awkward one with unknowns. Strip the customer details. Run each through the built-in tool and through a general assistant that has your rate sheet, and fill in a score sheet for each run.

1. Lines right: how many of the lines on your real quote appeared, and how many extra lines did it add?
2. Price against your real quote: the percentage difference on the total.
3. Minutes to a sendable quote: from opening the tool to a version you would email, including your edits.
4. Invented claims: promises, guarantees, compliance statements or time-scales you didn't supply.
5. Exclusions and assumptions: did it flag what the price doesn't cover?
6. Would you send it as it stands? Yes or no, with the reason.

Use the same three jobs for every tool so the comparison is fair, and time yourself, because the minutes column is where built-in tools usually win or lose. Keep the original quote beside the draft: you are testing against what you actually charged, not against what the tool thinks is reasonable.

One imagined firm's three-job test, scored

These figures are invented to show how to score, not results for any product. A three-person electrical firm tests a built-in tool trained on 140 of its past jobs against a general assistant given its rate sheet. The jobs are a socket swap and an added double socket (small), a consumer unit replacement (typical) and a part-rewire of an occupied house with an unknown ceiling void (awkward).

The pattern is the typical one. The built-in tool is fastest and closest on jobs that resemble work it has seen, because it knows your own item names and prices. On the awkward job it undershoots by 22% and drops six lines, because rare work has no close match in the history. The assistant is slower on the simple jobs and adds the odd stray line, but it handles the unusual scope better because you can talk it through the unknowns. Total time was 23 minutes against 25, which is a tie, so the deciding factors were accuracy on typical work and the quality of the scope notes on awkward work. Both needed a human to add up the totals.

The firm's decision: use the built-in tool for line items on routine jobs, and bring the assistant in for awkward jobs, written scopes and chasing messages. That is a rule you can write on one line, which is what a test should give you. Here is an illustrative chaser prompt for the second job, because follow-ups are where an assistant earns its subscription fastest.

Write a two-sentence text to [first name] following up a $1,240 consumer unit quote sent last Tuesday. Friendly, no pressure, offer two possible days next week, and mention the quote is valid for 30 days. No emojis.

A reasonable result reads: "Hi [first name], just checking you received the quote for the new consumer unit last Tuesday. I've got Wednesday or Friday free next week if you'd like to book it in, and the price holds for 30 days." Check the validity period against your own terms before sending, as the assistant simply repeated the figure you gave it.

What the extra AI costs on each side

Where the AI is included in a plan you already run, the question is whether it earns its place in the test. Where it costs extra, the sums are small enough to check against minutes saved.

A single login shouldn't be shared between staff: Anthropic's consumer terms forbid sharing login details, so three people means three subscriptions or a team plan. If only the owner writes quotes, the sum is $240 a year at most, and the free tier may already be enough. The more useful figure is minutes. If the test shows the tool saving you 10 minutes on each of 20 quotes a month, that is more than three hours, which at $50 an hour is over $150 of time against a $10 upgrade.

A third route: standalone quote-generator sites

The pages that rank for this search are mostly standalone generators, so they belong in the comparison. I checked two. TradeQuoteAI charges $12, $24 or $49 a month after a 14-day trial, lets you type or speak line items and sends PDFs, and its page doesn't mention a saved price list. Jotquote has a free tier of five quotes, paid plans from $15 a month billed annually, builds an itemised quote from a plain-language description or a photographed handwritten note, and keeps a price book on its paid plans.

Neither page described how it stores or shares what you upload, and one says photos go straight to an AI model. Before using any of them with real customers, check four things:

• Does it use your own price book, or its own idea of market rates?
• Where do your quotes and photos go, and can you delete them?
• Can the quote move into your invoicing and accounting without retyping?
• What happens to your saved quotes if you cancel?

A generator that can't answer the first and third questions is a typing aid. It may still suit a one-person firm with no job software, but it won't replace the test above.

Before an AI-written quote reaches a customer

Whatever tool drafts it, run the same short check. An approval step for AI-written quotes turns this list into a habit your team follows without thinking.

• Add up the lines yourself. Don't trust the printed total, whoever wrote it.
• Compare hours and quantities with your site notes. Assistants round up quietly.
• Remove promises you didn't make: time-scales, guarantees, "compliant with all regulations". A quote that promises compliance can bind you to it.
• Check the provisional sums. Anything unknown should be labelled as such, with how it will be charged.
• Confirm payment terms and the validity period against your standard wording.
• Keep customer details out of consumer chat tools, or switch off model training in privacy settings first.

The compliance line is the one that catches people. A before-and-after shows it. Your notes say: "Rewire kitchen, new sockets, cooker circuit." An assistant expands that to: "Supply and install a new radial circuit for the cooker and four double sockets above the worktop, all work tested and certified, and complying with all current regulations." The first half is a useful scope. The last clause is a legal promise you didn't write, so delete it and replace it with whatever certificate your trade normally issues.

Which route suits which kind of firm

These are starting points to test, not verdicts, and each assumes you run the three-job test before committing money.

• A solo electrician with no job software should start with a free assistant and a rate sheet, and spend nothing until the quotes prove slow.
• A two-van plumbing firm on ServiceM8 with years of job history has the strongest case for the built-in tool on routine jobs, with an assistant for covering notes and chasers.
• A three-person joinery firm on Tradify's Pro plan pays $10 more per user for SmartWrite. If wording is the problem, a free assistant may do the same job; if it's slow typing on site, the in-app version saves a step.
• A builder with long scopes and provisional sums gets more from an assistant for scope and assumptions, plus Fergus-style conversion of supplier PDFs into lines if the quote starts from a supplier price.
• A heating engineer on Powered Now can try its AI text writing for descriptions on the top tier and use an assistant for chasers. Powered Now against Tradify for a sole-trader heating engineer compares the two on cost and AI.

The common thread is that the built-in tool wins where your own history is the asset, and the assistant wins where the job is new or the words matter. Run the test, keep the score sheet, and repeat it in six months, because every vendor on this page is changing its AI features monthly. If you want help running it on your own quotes, that is part of what my AI implementation consultation does, through the contact page.

Quote-writing questions tradesmen ask about AI

Can I paste customer names and addresses into ChatGPT?

Avoid it unless you have checked your plan's data settings. Consumer plans have a model-training switch in privacy settings, and business plans don't train on your content by default. The simplest habit is to replace the name and address with 'the property' and paste only the job description and your rates.

Does ServiceM8's AI quote work on the free plan?

ServiceM8's pricing page lists AI Quote & Invoice on every plan and caps AI Helper Assists at 10 a day on Free. It doesn't say whether a quote counts as one assist. Its help article also says you need at least 100 jobs in the account before the drafts are reasonable.

Will a general assistant keep my prices consistent between quotes?

Only if you give it the same rate sheet every time and tell it to write 'price needed' for anything missing. Keep the rates in a ChatGPT or Claude project so each new chat starts from the same file, and still add up the totals yourself, because arithmetic slips are the most common error.

Can AI price a job from photos?

Treat it as a rough steer at best. ServiceM8 describes its image tool as helping clients visualise the outcome you are quoting, not pricing the work. One standalone generator reads a photographed handwritten note, which is reading your writing, not judging the job. A site visit still sets the price.

Further reads

• AI App for Tradies: ServiceM8, Tradify or Fergus by Business Size — Pick between ServiceM8, Tradify and Fergus by firm size.
• 10 ChatGPT Prompts for Plumbers: Call-Backs, Quotes, Complaints — Ready prompts for quotes, call-backs and complaints.
• Turning On-Site Voice Notes Into Job Sheets and Quotes With AI — Turn on-site voice notes into job sheets and quotes.
• Following Up Silent Quotes With AI: A Decorator's Playbook — Chase quotes nobody answered, with wording that works.
• Best AI Receipt Scanning App for UK Tradesmen: Xero, Tradify, Powered Now — The receipt-scanning side of the same job software.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ServiceM8 pricing page, AI job-management page and Auto-Quote help article; Tradify pricing page; Fergus AI page; Powered Now pricing page; two standalone quote-generator sites; OpenAI and Anthropic plan pages. All checked 29 October 2026.