Quoting Custom Jobs Faster With AI in a Small Workshop

To quote custom jobs faster in a small workshop, let AI turn the customer's enquiry, sketch or photos into a spec sheet with a list of missing details, find your closest past jobs, and draft the quote wording, while a costing spreadsheet does the maths from your own material prices and hourly rates. The slow parts of custom quoting are reading, chasing and writing.

What AI should not do is set the price. It doesn't know your machine time, your supplier's price for oak this month, or that finishing always takes you longer than you think. Asked for a price, it produces a confident, plausible and usually wrong number. Keep it on the words and the organising, and keep the figures in a sheet you control.

Follow me on Instagram@sagnikteaches

Where the time on a custom quote actually goes

Time one quote honestly, start to finish, including the emails. An illustrative breakdown for a bespoke fitted-furniture job in a two-person workshop:

Connect on LinkedInSagnik Bhattacharya

That's about two and a half hours of work, much of it in the evening, and often a week of elapsed time because of the questions. The biggest gain is usually the second row: asking every question once, on day one.

Subscribe on YouTube@codingliquids

Step 1: From enquiry to spec sheet and one question list

Consider an illustrative enquiry to a kitchen fitter who also builds bespoke units in a small workshop:

Hi, we'd like cupboards and shelves built either side of our chimney breast in the living room. Each alcove is about 1.1m wide. Cupboards at the bottom, shelves above. Not sure about painted or natural oak, what would you suggest? Photo attached. Hoping to have it done before Christmas.

Paste the email and photo into an assistant with a prompt that separates what's known from what's guessed:

You are helping a joinery workshop prepare a quote.
From the enquiry and photo, produce a spec sheet with three columns:
KNOWN (stated by the customer), ASSUMED (reasonable default we should
confirm), MISSING (must ask before quoting).
Cover: dimensions, materials, finish, doors and hardware, shelves,
fixing, lighting, skirting and coving, removal of existing items,
access, timescale.
Then write one friendly email asking all MISSING questions and
confirming the ASSUMED ones, numbered, max 8 questions.
Do not suggest prices.

An illustrative extract from the spec sheet:

The question email that comes with it asks for exact widths and ceiling height, the depth, handles or push-to-open, lighting, and whether they'd like both finishes priced. Read it before sending, because the draft usually needs a few edits. An illustrative first version:

Hi [first name], thanks for getting in touch! We'd love to help,
and we can definitely have this done before Christmas.
1. Could you measure the exact width of each alcove?
2. What is the floor-to-ceiling height, and is there coving?
3. How deep would you like the base cupboards?
4. Handles, or push-to-open doors?
5. Would you like LED lighting in the shelves?
6. Shall we price both painted and natural oak?
7. Should we remove the existing shelving?
8. What is your budget for the project?

Three edits. "We can definitely have this done before Christmas" is a promise nobody has checked against the workshop diary; replace it with "we'll confirm dates once we've measured". "Coving" is trade vocabulary many customers won't recognise, so add "(the decorative strip where the wall meets the ceiling)". And question 8 is a judgement call: some workshops find a budget question saves quoting jobs that will never go ahead, others find customers resent it. Decide your policy once and put it in the prompt, rather than leaving the model to decide each time.

Customers rarely answer everything. Suppose the reply says: "Left is 1.08, right is 1.12, ceiling about 2.45 with coving. Push-to-open please, no lights. Both prices would be great." Paste it under the question list and ask which questions are still open. The assistant marks 1, 2, 4, 5 and 6 answered and flags 3 (depth), 7 (existing shelving) and 8 (budget) as open. Let the budget go; they've asked for two prices, which says enough. Rather than another round of emails, put the other two in the quote's assumptions ("base cupboards 350 mm deep; existing shelving removed by the customer") and confirm them at the survey. One email, answered once, replaces four or five rounds. Offering both finishes as priced options also answers the customer's "what would you suggest?" better than an opinion would. If you gather details on site instead, turning on-site voice notes into job sheets and quotes covers the spoken version of this step.

Step 2: Your jobs log becomes the memory

The best guide to what a new job will take is what similar jobs took. Keep a simple jobs log, one row per finished job:

• Job number, date, one-line description
• Main dimensions and materials
• Quoted hours by operation: machining, assembly, finishing, fitting
• Actual hours by operation (from your timesheet or a quick note at the end)
• Material cost, quoted price, and whether it was won

Filled in, one row reads (illustrative):

Keeping the hours split by operation is what makes the log useful. A single total of "31 hours" would say the job ran over; the split says it was finishing, and that's something you can price.

No log yet? Rebuild the last ten or so jobs from your invoices, diary and memory. Rough hours per operation are fine to begin with; being within an hour or two is far more useful than having nothing, and the log gets more accurate with every job you add. Record actual hours at the end of each new job while you still remember them, even if it is a note on the job sheet.

When a new spec is ready, paste it with the log and ask: "Find the three most similar past jobs and show quoted versus actual hours for each operation." An illustrative answer:

Most similar: Job 2025-044, alcove units, two alcoves at 1.0 m, painted MDF, quoted 26 hours, actual 31 (finishing took 9 hours against 5 quoted). Job 2025-061, one alcove unit in oak veneer with a solid oak face frame, quoted 16 hours, actual 17. Job 2024-112, fitted shelving only, not a close match.

Now you know something useful before costing: painted finishes have run over on finishing hours, so this quote should allow more time for painting than the last one did. Check the matches yourself; the assistant will call something "similar" on thin evidence, like the 2024 shelving job here.

It's worth doing the sum on that overrun, because it shows what the log is worth. Four unquoted finishing hours at $45 is $180 on one job. If the workshop does eight painted jobs a year and each runs over by a similar amount, that's about $1,440 of work given away annually, from one line in the estimate that nobody had questioned.

Step 3: The costing sheet does the arithmetic

Your sheet holds your prices, rates and margins. An illustrative costing for the oak-veneer option, two alcoves:

The painted option uses MDF instead of veneered sheet, but the jobs log says to allow about 9 finishing hours instead of 5. Those four extra hours ($180) and the paint roughly cancel out the cheaper board, so the two options may end up closer in price than the customer expects, and that's worth saying in the quote. For why this belongs in a sheet rather than a chat, see whether AI can price a job accurately; other trades have reached the same conclusion, as print shops pricing jobs with AI shows.

Step 4: Drafting the quote the customer reads

With the spec confirmed and the figures from the sheet, ask the assistant to draft the quote text: a description in the customer's own terms, what's included, what isn't, assumptions, timescale, payment stages and validity. Give it the figures to use and tell it not to change them. An illustrative extract:

Option A: natural oak, $3,065 before tax. Built-in cupboards and shelves for both alcoves, in oak-veneered board with solid oak face frames, finished in a clear hardwax oil. Each alcove has a base cupboard with two doors (about 800 mm high and 350 mm deep) and three adjustable shelves above.

Not included: decorating the walls, any electrical work for lighting, removing the existing shelves.

Assumptions: final dimensions confirmed at the survey; walls sound enough to fix to. We'll confirm the price after the survey if anything differs.

Timescale and payment: about four weeks from order to fitting; 40% deposit to order materials, balance on completion. This quote is valid for 30 days.

Clear exclusions do more to prevent disputes than any other part of the quote, and they are easy to forget when you're writing at 10pm. A prompt that always asks for them means they are always there.

The same flow in a fabrication workshop

The steps carry over to metalwork. Imagine a small fabrication shop that receives a phone photo of a pencil sketch: "Steel side gate, 1.2 m wide, 1.8 m high, vertical bars, black." The spec sheet step marks width and height KNOWN, bar spacing and powder coating ASSUMED, and asks the questions that decide the price: is it hung on existing posts or new ones, which way does it open, does the ground slope, is a lock or latch needed, and is it self-closing? The jobs log finds two previous gates and shows that one with new posts took three extra hours on site. The costing sheet adds steel by length, the powder coater's price and fitting time. The quote states that ground conditions for new posts are assumed normal. Same method, different materials.

One detail in that spec sheet needs correcting, and it's typical of how AI reads a customer's words. It listed "gate width 1.2 m" as KNOWN. What the customer measured was almost certainly the opening between the walls or posts, and the gate itself has to be narrower to leave room for the hinges and latch. Built at 1.2 m, it won't fit. The fix is a line in the prompt, "for gates, doors and panels, treat the customer's width as the opening, not the finished size", and a MISSING question asking them to confirm what they measured. The same trap applies to alcove widths, which are rarely square: measure at the top, middle and bottom at the survey.

Checks before any quote goes out

• Every figure in the text matches the sheet. A realistic mistake: asked to "tidy up" the quote wording, an assistant rewrote $3,065 as $3,650. Copy prices from the sheet last, after all rewording, or check them line by line. Catching made-up figures in AI-drafted proposals has a fuller routine.
• Units are consistent. Millimetres throughout; watch for a stray inch or a metre figure read as millimetres.
• Every MISSING item is answered or written into the assumptions.
• Material prices are dated within the last 30 days.
• Hours are sense-checked against the comparable jobs, especially the operation that ran over last time.
• Exclusions and validity are present.

If more than one person quotes, make these checks a formal sign-off; setting up an approval step for AI-written quotes shows how to keep it quick.

What quoting looks like after a month

For the illustrative workshop above, a realistic target is to take a custom quote from about two and a half hours to under an hour of hands-on time: 5 minutes for the spec sheet, 10 to check it and send the questions, 10 for comparables, 20 for costing, 10 to review the drafted text and run the checks. Elapsed time falls further, because the questions go out the same day.

Track three numbers in the jobs log: time spent per quote, win rate, and actual hours against quoted hours by operation. The last one is where the money is. After a dozen jobs you'll know whether your finishing or fitting estimates are consistently light, and every future quote improves because the log does.

An illustrative review after twelve jobs, asked of the assistant with the log pasted in ("For each operation, how often did actual hours exceed quoted, and by how much on average?"), might come back: machining and assembly within an hour either way on most jobs; fitting over on four of twelve, all in older houses with uneven walls; finishing over on seven of the nine painted jobs, by about three hours on average. Check those counts against the log yourself before acting, since the assistant can miscount rows as easily as it misreads a sketch. Then make two changes in the costing sheet, not in the prompt: raise the default painted finishing time by three hours, and add a "walls out of true" fitting allowance that you tick at the survey. The next painted quote carries about $135 more labour before margin, and lands much closer to what the job actually takes.

Workshop quoting questions

Can AI read my customers' sketches and drawings?

It can usually read dimensions and notes from a clear photo of a sketch or a PDF drawing, and describe what it shows. It is less reliable with faint pencil, overlapping dimensions or anything it has to scale. Treat every dimension it reads as ASSUMED until you or the customer confirm it, and never cut from an AI-read dimension.

Should I let AI produce a cut list?

Use proper cutting software or your own method for anything you will actually cut. An assistant can draft a rough parts list to help estimate sheet quantities, but it makes arithmetic and allowance mistakes, such as forgetting saw kerf or edging thickness. For quoting, a rough list plus a waste factor is enough; for making, measure and calculate properly.

How do I keep material prices current in the costing sheet?

Keep one prices tab with a date next to each price, and update it whenever a supplier invoice arrives. Add a check that turns any price older than 30 days red. For volatile materials such as hardwood or steel, ask your supplier for a price before sending a large quote, and state the price validity in the quote.

Further reads

• How to Create Quotes and Estimates in Minutes With AI — Quoting habits that apply to any small business.
• How Small Builders Use AI to Price Variations and Change Orders — Price the changes customers ask for mid-job.
• How to Compare Supplier Quotes Side by Side With AI — Compare material suppliers' prices side by side.
• Where Should a Small Manufacturer Start With AI? — Where else a small making business should start with AI.
• Can ChatGPT Read PDFs, Spreadsheets and Photos? What Breaks — What assistants read reliably from drawings and photos.
• AI Supplier Management: Track Prices, Lead Times, and Risk — Track supplier prices and lead times over time.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.