Put your prices and pricing rules in a rate card or spreadsheet, let AI turn each enquiry (an email, drawing, photo or voice note) into structured line items, let the spreadsheet or quoting tool do the maths, and check before sending. QuickBooks Online and Xero can already draft an estimate or quote from a forwarded email.
Minutes is realistic for repeat work with clear prices; custom jobs still need a person to judge the tricky parts, just less of their time. The trap is letting the AI produce the price itself. Language models are good at reading a messy request and poor at applying a rate card consistently, so an AI quote that "looks right" can be wrong on exactly the lines nobody rechecks. Keep the AI on reading and writing, and keep the numbers in something that calculates.
Quote, estimate or proposal: know which one you're writing
The three get used interchangeably, but they commit you to different things, and the automation should know which it's producing.
| Document | What it commits you to | When a spare parts maker uses it |
|---|---|---|
| Quote | A firm price for a defined job, usually for a stated period | Repeat parts, catalogue items, anything already made before |
| Estimate | A best guess that may change once details are confirmed | A new part from a customer's sketch or photo, before drawings are checked |
| Proposal | A recommended approach, options and pricing for a larger piece of work | A year's supply agreement or a stocking arrangement |
Label the document correctly, because customers treat a "quote" as a promise. Quotes and estimates are the focus here; proposals need a different method, with more tailoring and less arithmetic. The example throughout is an illustrative manufacturer of spare parts for food-processing and packaging machines, receiving about 40 requests for quotation (RFQs) a week.
The rate card is the real work
An automation can only be as consistent as the rules it follows, and in most small firms the rules live in the estimator's head. Writing them down takes a day or two, and it's the step that makes everything else fast. Here's the manufacturer's rate card, abridged (illustrative figures):
RATE CARD - prices last updated 1 Sep 2026 (review monthly)
MATERIALS (per kg of blank, incl. cutting)
Stainless 304 bar ......... $6.50 Stainless 316 bar ... $8.90
Aluminium 6082 bar ........ $5.20 Acetal (POM) rod .... $9.40
MACHINE RATES (per hour)
CNC turning ............... $65 CNC milling (3-axis) . $78
5-axis milling ............ $110 Manual/bench work .... $48
CHARGES PER BATCH
Setup: $120 per operation. Inspection report: $45.
First-article inspection (new parts only): $150.
FINISHING (per part, minimum $90 per batch)
Anodise (aluminium) ... $2.40 Passivate (stainless) .. $1.60
RULES
Minimum order value $250. Margin: price = cost / 0.65.
Quantity breaks: 50+ less 5%, 200+ less 10% (on machining only).
Expedite (under 5 working days): +25% on total.
Lead time: 10 working days from order and drawing approval.
Quotes valid 30 days; material prices valid 14 days.
Anything not on this card: estimator prices it.The last line is the most important rule on the card. It tells the automation when to stop and hand over, rather than improvising a price. For a workshop making one-off jobs, quoting custom jobs faster in a small workshop shows a rate card built around labour and materials instead of machine hours.
From enquiry to line items
Enquiries arrive as anything but a neat list. A typical RFQ email to the manufacturer (illustrative):
"Hi, can you quote for the drive shafts for our tray sealer, same as the ones you did for us last spring but we need 60 this time, maybe 100 if the price is right. Drawing attached (rev C). Also 20 of the white guide blocks, the drawing is on your system from before I think. Need them by end of next month ideally."
A prompt that turns this into structured lines and questions:
You prepare quote requests for our estimator.
From the ENQUIRY and any ATTACHED DRAWING TEXT, list every item as:
item, drawing_no, revision, material, quantity (list every
quantity mentioned), finish, delivery_needed_by (a date),
repeat_part (yes/no/unknown).
Then list QUESTIONS we must ask before quoting.
Rules: don't guess material or finish; if "same as before" is
mentioned, write "look up previous order" and leave the field
empty. Don't produce any prices.Illustrative output:
| Item | Drawing / rev | Material | Qty | Finish | Needed by | Repeat? |
|---|---|---|---|---|---|---|
| Drive shaft, tray sealer | DS-2210 / C | Stainless 304 (from drawing) | 60; 100 | Not stated | 30 Oct 2026 | Yes: look up previous order |
| Guide block, white | Not supplied | Look up previous order | 20 | n/a | 30 Oct 2026 | Yes: look up previous order |
Questions listed: Is revision C different from the shafts supplied last spring? Is passivation required as before? Which drawing number is the guide block?
That's a good result, with one thing to fix. The AI read "white guide blocks" and, in an earlier run, filled in "acetal (POM)" as the material, because white guide blocks usually are. It was probably right, but "usually" isn't good enough for a quote, which is why the prompt now says "don't guess material". Two other details show why structured extraction beats reading the email in a hurry: it caught that the customer gave two quantities (so the quote shows both, which often wins the larger order), and it flagged the revision letter, which matters if the drawing changed since the last batch.
Let the spreadsheet price it
With the lines and the rate card, the price is arithmetic. Keep it in a spreadsheet (or your quoting system) where every figure traces back to the card. The estimator enters only what needs judgement: the cycle times, taken from the previous job's records for a repeat part or estimated from the drawing for a new one.
The quick sum for the drive shaft at 60 pieces, from the card above (illustrative):
| Line | Calculation | Per part |
|---|---|---|
| Material | 0.8 kg × $6.50 | $5.20 |
| Turning (14 min) | 14 ÷ 60 × $65, less 5% (50+ break) | $14.41 |
| Milling (6 min) | 6 ÷ 60 × $78, less 5% | $7.41 |
| Setup (2 operations) | 2 × $120 ÷ 60 | $4.00 |
| Passivate | $1.60 (batch $96, above the $90 minimum) | $1.60 |
| Inspection report | $45 ÷ 60 | $0.75 |
| Cost | $33.37 | |
| Price | $33.37 ÷ 0.65 | $51.34 |
At 60 pieces that's $3,080.40; at 100 pieces the setup is spread further ($2.40 a part), so the price drops to about $48.40 each, roughly $4,840 in total. Showing both is the kind of detail that nudges a customer to the larger order. The spreadsheet does this in a second with lookups against the rate card; in Excel or Google Sheets, XLOOKUP (or VLOOKUP) pulls each rate by name, so changing the card updates every future quote. Asking an AI to do these sums in a chat is exactly where errors creep in; why AI is bad at maths, and what to check in quotes explains the reasons.
Catalogue items: the lane that can run on its own
Not every enquiry needs an estimator. About a third of the manufacturer's RFQs are for standard catalogue parts with fixed prices: belts, seals, knives and guide rails it keeps in stock. For those, the flow runs end to end with no one pricing anything:
- The enquiry arrives in the quotes inbox.
- The AI extracts part numbers and quantities, and checks each part number exists in the price list.
- The spreadsheet prices every line, applies quantity breaks and the minimum order, and checks stock.
- If every line matched, the total is under $1,000 and everything is in stock, the quote is generated and sent from the template. Otherwise it goes to the estimator's queue.
An illustrative example: a food plant's maintenance planner emails "please quote 12 x 4471-B drive belts and 6 x SL-220 seals, and can you confirm stock". Three minutes later the quote is in their inbox with both lines priced, "in stock, dispatch within 48 hours of order" on each, and a 30-day validity. The one rule that makes this safe is the match check in step 2: a part number with a typo ("4471-8" instead of "4471-B") doesn't match, so it goes to a person instead of being quoted as the nearest guess.
Built-in options in accounting and CRM tools
If your quotes are simpler than a machined part, your accounting software may already do much of this:
- QuickBooks Online. Intuit Assist can generate an estimate or invoice when you forward a customer's email or upload a photo of a document, for you to review and send; see Intuit's page on Intuit Assist. Check every drafted line against your products-and-services list before sending, and keep that list tidy so the drafts have the right items to pick from.
- Xero. Its AI assistant, JAX, has been in beta, and Xero describes forwarding an email to JAX to create a quote (see Xero's JAX update). Check it's available on your subscription; using Xero's JAX for invoices and cash-flow questions covers the rest of what it does.
- Your CRM. Pipedrive's Smart Docs, for quotes and proposals created from deal data, is included from its Premium plan and an add-on below that. Most CRMs can generate a quote document from a deal's products and prices.
These work best for price-list items. For anything calculated, keep the calculation in your spreadsheet and send the final lines to the accounting tool, so the quote, the order and the invoice share one set of numbers; connecting quotes, invoices and payments covers that hand-off.
A worked example: 40 RFQs a week
The manufacturer's estimator used to handle every RFQ from scratch. After a two-day set-up (rate card, extraction prompt, pricing spreadsheet, quote template and an approval rule), the illustrative numbers after a month:
| Measure | Before | After |
|---|---|---|
| Time per repeat-part quote | 35 minutes | 9 minutes |
| Time per new-part estimate | 90 minutes | 45 minutes |
| Quotes sent the same working day | About 30% | About 80% |
| Estimator hours on quoting per week | About 30 | About 12 |
| Pricing errors found after sending | 2 or 3 a month | 1 in the month (a stale material price) |
Eighteen hours a week back is the headline, but same-day quotes may matter more: in spares, a customer with a machine down often orders from whoever answers first. The one error came from a supplier raising stainless prices mid-month; the fix was adding "material prices valid 14 days" to the card and to every quote.
To see whether speed is winning orders, log two dates on every quote (enquiry received, quote sent) and whether it became an order. After two or three months, compare the order rate for quotes sent the same day with those sent later. If same-day quotes convert noticeably better, that's the case for putting more work into the automatic lane; if there's no difference, your customers care more about price or lead time, and the saved hours are the main benefit.
Photos and voice notes as enquiries
Customers often send a photo of a worn part with no drawing and no part number. AI can help here, within limits. Ask it to describe what it can see and list what you need to know, not to identify the part:
Illustrative AI draft reply: "Thanks for the photo. It looks like a sealing jaw from a flow-wrapper, worn along the leading edge. To quote a replacement we'll need the machine make and model (usually on a plate near the infeed), the overall length and width of the jaw, and the part number if one is stamped on the back. If you can send the old part, we can measure it and quote within two days of receiving it."
That reply is helpful precisely because it doesn't guess a part number. A realistic failure the manufacturer saw early on: an AI draft named a specific part number from the photo, the customer ordered on the strength of it, and the jaw was for a different width of the same model. Now photo enquiries always become estimates, never quotes, until the part is confirmed.
Voice notes from field engineers work the same way: "Customer needs four guide rails like the ones on line two, about 600 long, the white plastic ones, and a set of belts" becomes structured lines plus questions (which belt size, which rail profile). Ask engineers to say the part, the quantity and the key measurement in that order, and the extraction gets far more reliable.
The cover note and terms
The quote document carries the numbers; the email carrying it should be short and answer the customer's actual question. AI drafts this well from the extraction and the price sheet.
Before: "Please find attached our quotation as requested. Do not hesitate to contact us should you require any further information."
After: "Quote attached for 60 drive shafts to DS-2210 rev C at $51.34 each, or 100 at $48.40. That's 10 working days from order, so ordering by 16 October meets your end-of-month date. We've assumed passivation as on last spring's batch; tell me if that's changed. The guide blocks need a drawing number before we can price them. Could you check it for me?"
Keep standard terms fixed in the template rather than regenerated: validity period, the date material prices were set, lead time counted from order and drawing approval, what's excluded (for example, tooling for new parts), and payment terms. AI should never rewrite terms.
Errors that cost money, and the approval step
- Unit mix-ups. A drawing dimensioned in inches read as millimetres, or the reverse. Make "units confirmed" a field the estimator ticks.
- Quantity breaks applied wrongly. A discount meant for machining applied to the whole price. Keep break rules in the spreadsheet, never in the prompt.
- Old prices. A material price from before a supplier's increase. Date-stamp the card and set a review reminder.
- The wrong revision. Quoting rev B when the customer sent rev C. Show the revision on every quote line.
- Missed minimums. A $140 order quoted below the $250 minimum. Let the spreadsheet apply it automatically.
Then decide who must look at what before it goes out. The manufacturer's rule:
| Quote type | Approval |
|---|---|
| Repeat part, unchanged revision, under $1,000 | Estimator glances and sends |
| Repeat part over $1,000, or any changed revision | Estimator checks every line |
| New part, new material, or over $10,000 | Estimator plus owner |
| Anything from a photo or with no drawing | Estimate only, never a quote |
Setting up an approval step for AI-written quotes goes through building this into your tools so nothing skips it. And once a quote is out, the work isn't done: a polite, useful follow-up a few days later (confirming the drawing revision, or noting the order-by date for their deadline) wins orders that would otherwise go quiet, which is covered in automating follow-ups without being pushy.
Quoting with AI: follow-up questions
Can AI read a technical drawing and quote from it?
It can read the text on a clear PDF drawing, such as the title block, material, quantities and many dimensions, which saves typing. It can misread tolerances, surface finishes and small annotations, and it can't judge how hard a part is to make. Use it to pre-fill the job sheet and list questions, then have an estimator confirm every critical dimension and the machining time before any price goes out.
Should customers see an instant price on my website?
For standard catalogue items with fixed prices, yes: a price list, basket or simple calculator saves everyone time. For custom or variable work, show a starting-from figure or a range and collect the details for a proper quote, because an instant automated price for custom work is either padded to be safe or wrong. Keep automated website prices to items where your rate card fully decides the answer.
How do I stop quotes going out with old prices?
Put a 'prices last updated' date at the top of your rate card and make reviewing it someone's monthly job, with a quicker review whenever a supplier changes material prices. State a validity period on every quote, such as 30 days, and the date the material prices were set. The automation can also refuse to produce a quote if the rate card is older than a set number of days.
Further reads
- How to Write Business Proposals Faster With AI — For larger bids where a quote isn't enough.
- Can AI Price a Job Accurately? What Tradespeople Should Check — Where AI pricing works and where it misleads.
- How to Compare Supplier Quotes Side by Side With AI — The other side: checking the quotes you receive.
- Turning On-Site Voice Notes Into Job Sheets and Quotes With AI — Turning on-site voice notes into quotes.
- AI Document Processing: Turn PDFs and Scans Into Usable Data — Pulling data from PDFs and scans reliably.
- Can AI Tell You What to Charge? The Limits of AI Pricing Research — Why AI can't set your prices for you.
- How Caterers Use AI to Quote Events and Plan Menus — How a small caterer can use AI to go from messy enquiry to costed quote and proposal, with a worked 70-guest quote and prompts to copy.
- How Videographers Use AI to Edit, Caption and Quote Faster — Transcript editing in Premiere, Resolve and Descript, a caption proofreading routine, an enquiry-to-quote prompt with sample output, and rights checks.
- AI Estimating Software for Small Builders: Six Options Compared — Six AI estimating options for small building firms compared on what the AI does, list price, best fit and the catch, plus a way to test them on a real job.
- How Painters and Decorators Can Quote From Photos With AI — Send a photo brief, let AI read surfaces and condition, get sizes from a tape or scan, price from your rates and word the range so it holds up on the day.
- How Removal Firms Turn Customer Videos Into Quotes With AI — A seven-step workflow for removal firms: from the customer's phone video to an inventory, a checked volume and a priced quote with clear assumptions.
- Can a Plumber Use AI to Write Quotes? — Turn survey notes and your price list into a clear plumbing quote in minutes, with the exclusions that prevent arguments and the checks AI still needs.
- Can a Print Shop Use AI to Price Jobs and Check Artwork? — How a print shop splits the work between AI, its price matrix and preflight software, with prompts, a proofing example and a customer report.
- AI Proposal Software vs a General AI Assistant: Which to Pay For — What proposal tools add beyond the writing, what they cost per seat, and a question-by-question table for choosing between them and a chat assistant.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: QuickBooks Intuit Assist product page; Xero product update on JAX (quotes from forwarded emails); Pipedrive knowledge base on plan features (Smart Docs). Checked September 2026.