Give an assistant one supplier quote at a time and ask for one row per priced line: description, unit, unit price, exact source text, validity and exclusions. Convert units with factors printed on the quote, paste the rows into a rate sheet and check them against the quote's total. A rate counts only with its source text beside it.
Takeoff counts what is on the drawing. The rate sheet answers the other half of an estimate: what each unit costs from the suppliers you really use. It is also where small builders leak money, through a price per 1,000 read as a price each, a quote that expired last week or a delivery charge nobody noticed. AI is quick at the typing; the checks below make the output safe to price from. Quantities from drawings are a separate job, covered in the small firm's takeoff test.
Where the rate sheet sits in an estimate
The chain is short. Drawings give quantities through a takeoff. Supplier quotes give rates through the rate sheet. Each estimate line multiplies one by the other, and the quote to the client is built from the estimate lines. Without a rate sheet, rates live in old emails, in last month's job and in memory, and the estimate quietly uses whichever one came to mind first.
A rate sheet has three jobs: hold the current rate for every item you buy regularly, show where each rate came from, and say when it stops being true. That last part, the date, is the one most builders skip.
The cost of the tooling is small. As of October 2026, the individual plans that handle a handful of quotes a week cost about the same: ChatGPT Plus and Claude Pro at $20 a month, Google AI Pro at $19.99. A free plan works for occasional use, with tighter limits. Supplier quotes are commercially sensitive, so on a personal plan switch model training off in the privacy settings before the first upload; team and business plans keep your content out of training by default.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowGet quotes an assistant can read
The extraction is only as good as the document. A few habits make a large difference:
- Ask for itemised, written quotes in a digital PDF, one per supplier, with the job reference on each. The RFQ email template tutorial shows wording that gets comparable quotes back.
- Name files so they sort and trace: supplier, date, quote reference and revision, for example brickmerchant_2026-10-14_Q-A-1043_rev2.pdf. Never rename a file after you have extracted from it.
- Avoid photos and scans of printed pages when a digital copy exists. If a PDF's text can be selected and searched, it is a digital document. If it cannot, it is an image, and every number needs a human check.
Try Excel before AI for clean digital quotes. In Excel on Windows, Data, Get Data, From File, From PDF opens a navigator that lists the tables Excel detects in the file, and you can load them straight into a sheet. Microsoft's support page describes the steps. It is deterministic, so it cannot misread a digit the way a language model sometimes does, and it is worth a test on your own suppliers' layouts. Where it fails, because a quote is not laid out as a table, the assistant takes over, with the checks below.
The extraction prompt, a messy quote and the row it got wrong
Here is a quote of the kind a merchant sends, with the abbreviations and footer small print that make extraction hard. The figures are illustrative.
QUOTATION Q-A-1043 rev 2 14 Oct 2026
Valid 14 days from date above. Prices in USD.
1 Buff multi facing brick, pallet 500 ........ 780.00 per 1000
2 Concrete block 440x215x100 7.3N (pk 96) .... 1.92 ea
3 Cavity insulation bd 100mm 1200x450 ........ 26.40 /sht
Delivery incl. on orders over 1,500.00. Kerbside offload only.
Prices exclude cutting and wastage. Subject to stock at time of order.Extract every priced line from the supplier quote below into a table with exactly
these columns: Line | Description | Supplier unit | Unit price | Pack size or coverage
(only if printed) | Printed line total (only if printed) | Source text (copy the exact
words and numbers from the quote).
Rules: copy numbers exactly as printed. If a price or unit is unclear, write UNCLEAR.
Do not convert units and do not calculate anything.
After the table, list on separate lines: quote reference and revision, quote date,
validity wording, delivery wording, and anything marked excluding, subject to, POA
or to be confirmed.
[paste the quote text or attach the PDF]Sample output, with one deliberate fault:
| Line | Description | Supplier unit | Unit price | Pack or coverage | Source text |
|---|---|---|---|---|---|
| 1 | Buff multi facing brick | each | 780.00 | Pallet 500 | "780.00 per 1000" |
| 2 | Concrete block 440x215x100, 7.3N | each | 1.92 | Pack 96 | "1.92 ea" |
| 3 | Cavity insulation board 100 mm, 1200x450 | sheet | 26.40 | Not printed | "26.40 /sht" |
Row 1 is wrong. The unit says "each", but the assistant's own source text says "per 1000". Paste that into an estimate and 6,200 bricks cost $4,836,000 instead of $4,836. Nobody would sign that, but the same slip on a smaller scale, a price per 100 read as each, produces a number that looks plausible and quietly loses money. The source-text column is what exposes it: the evidence and the interpretation disagree, and you can see it in two seconds. That is why the prompt demands the exact words, and why a row without them is not allowed into the sheet.
The notes after the table matter just as much. The reply should list Q-A-1043 rev 2, 14 Oct 2026, valid 14 days, delivery included on orders over 1,500.00, and three warnings: kerbside offload only, excludes cutting and wastage, subject to stock at time of order. Those lines go into the rate sheet's notes and the scope-gap review later, because a rate without its conditions is half a rate.
Unit traps: per 1,000, per sheet, per length and per tonne
Suppliers sell in units that estimates cannot use directly. A rate sheet needs one estimating unit for every item, and a factor that converts between the two. The factor comes from the quote or the product datasheet, never from the assistant's memory.
| Supplier sells | Estimate needs | Factor comes from | Rate |
|---|---|---|---|
| Bricks per 1,000 at $780.00 | Each | 1,000 per unit, by definition | $780.00 ÷ 1,000 = $0.78 |
| Block at $1.92 each | m² of wall | Supplier's stated coverage, 9.9 per m² | $1.92 × 9.9 = $19.01 |
| Insulation board at $26.40 per sheet | m² | 1.2 m × 0.45 m = 0.54 m² per sheet | $26.40 ÷ 0.54 = $48.89 |
| Timber at $31.60 per 4.8 m length | Linear metre | Length printed on the quote | $31.60 ÷ 4.8 = $6.58 |
| OSB board at $29.50 per sheet | m² | 2.44 m × 1.22 m = 2.9768 m² | $29.50 ÷ 2.9768 = $9.91 |
| Aggregate per tonne | m³ placed | Bulk density from the supplier for that product | Price per tonne × tonnes per m³ (at $30 a tonne and 1.6 t/m³ from the supplier's sheet, $48) |
| Delivery $65 per drop | Job overhead line | Number of drops planned | $65 × drops, unless free over a threshold |
The aggregate row is where assistants are most dangerous. Ask for tonnes per cubic metre and you will get a confident, plausible figure, and it will be a generic one, not the density of the product your supplier actually delivers. Put the conversion in your rate sheet only when the supplier has given it to you in writing, and say "UNCLEAR, ask the supplier" in the notes until then. The same rule covers coverage rates for blocks, pack sizes for fixings and the area of a sheet when the quote prints only a nominal size.
The rate sheet layout, and how the estimate pulls from it
Use one Excel table called Rates with these columns: Code, Item, Supplier, QuoteRef, QuoteDate, ValidTo, Price, SupplierUnit, UnitsPer, Rate, EstUnit, Source and Notes. Two points about the design. First, store the supplier's price and the conversion factor separately from the rate, so you can always see how the rate was made. Second, make the rate a formula, not a typed number:
UnitsPer (blocks) =1/9.9 (m2 per block, from the quote)
Rate =ROUND([@Price]/[@UnitsPer],2)
Estimate rate =INDEX(Rates[Rate],MATCH([@Code],Rates[Code],0))
Estimate total =ROUND([@Qty]*[@Rate],2)If your Excel has XLOOKUP, =XLOOKUP([@Code],Rates[Code],Rates[Rate],"NO RATE") does the same job. Either way, do not wrap the lookup in IFERROR returning zero: a missing rate should shout, not become a free item. With the three quotes above, plus a second plasterboard quote, the live part of the sheet looks like this:
| Code | Item | Quote ref | Valid to | Rate | Unit |
|---|---|---|---|---|---|
| BR-01 | Buff multi facing brick | Q-A-1043 rev 2 | 2026-10-28 | $0.78 | each |
| BL-01 | Concrete block 440x215x100, 7.3N | Q-A-1043 rev 2 | 2026-10-28 | $19.01 | m² |
| IN-01 | Cavity insulation board, 100 mm | Q-A-1043 rev 2 | 2026-10-28 | $48.89 | m² |
| TB-01 | C24 timber 47x150, 4.8 m | T-7731 | 2026-11-11 | $6.58 | lm |
| OS-01 | OSB3 18 mm sheet | T-7731 | 2026-11-11 | $9.91 | m² |
| PB-01 | Plasterboard 12.5 mm, supplier C | PB-5520 | 2026-11-12 | $3.40 | m² |
| PB-02 | Plasterboard 12.5 mm, supplier D | D-0218 | 2026-11-05 | $3.55 | m² |
Keep each supplier's rate on its own row with its own code and let the estimate choose on purpose. The comparison is then one line: after converting C's per-sheet price, supplier C is $3.40 per m² and supplier D is $3.55, so D is 4.4% dearer on the same board. Before conversion the two quotes could not be compared at all.
Now the estimate side. Suppose the takeoff, done separately, gives these quantities for the extension. The estimate lines carry only a code and a quantity, and the rate arrives by lookup.
| Code | Qty | Unit | Rate | Total |
|---|---|---|---|---|
| BR-01 | 6,200 | each | $0.78 | $4,836.00 |
| BL-01 | 64 | m² | $19.01 | $1,216.64 |
| IN-01 | 64 | m² | $48.89 | $3,128.96 |
| TB-01 | 120 | lm | $6.58 | $789.60 |
| OS-01 | 38 | m² | $9.91 | $376.58 |
| PB-01 | 140 | m² | $3.40 | $476.00 |
| Materials | $10,823.78 |
Wastage, cutting and delivery are not in those rates, because the quotes excluded them. They are separate estimate lines, and that is the point of recording the exclusions. A single sheet that holds both the rate and its conditions stops the same 10% waste being added twice, or not at all.
Five tests before a rate goes live
An extracted rate is a claim. These five tests turn it into something you can price from, and a Checked column holds the date and initials of whoever ran them.
- Line test. Where the quote prints line totals, check that quantity times unit price equals the printed total:
=ROUND([@Qty]*[@UnitPrice],2)=[@PrintedTotal]. A misread digit fails here. - Total test. The extracted lines, less any discount and plus delivery, should add up to the quote's printed total. If the lines add to $4,150.00 and the quote prints $4,510.00, one row is wrong by $360, and the source-text column tells you which.
- Sanity test. Compare each new rate with the one it replaces:
=[@NewRate]/[@OldRate]-1, flagged when the absolute change is above a threshold you choose (15% is a sensible start). An ordinary price rise passes, as the bricks' move from $760 to $780 per 1,000, which is 2.6%, would. A unit slip fails loudly, as $780 each would. - Second pass. Run the same quote through a different assistant and compare the two tables. Differences are not errors; they are the rows to look at first. Two assistants are less likely to make the same slip on the same row.
- Spot check. Open the PDF and search for the source text of at least one row in ten, and every row above a value you set (say, anything that will make up more than 5% of the estimate). It takes minutes, and it is the only test that catches a quote the assistant read correctly but mis-filed.
The same ideas appear for a different job in checking a quote with ChatGPT, and for plumbers in the tutorial on updating a price book from supplier invoices.
Quotes that arrive as photos, scans or odd number formats
Not every supplier sends a clean PDF. A photo of a printed quote, a scan with a coffee ring across the prices or a handwritten discount scribbled in the margin all go through the same extraction, and all of them raise the odds of a wrong digit. Assistants can misread image digits in the familiar ways of any optical reading: 5 for 6, 1 for 7, a dropped decimal point, a crossed-out figure treated as live. Four habits reduce the damage:
- Photograph flat and crop to the table. One page per image, good light, no shadow across the figures. A skewed phone shot taken across a van bonnet costs you accuracy for nothing.
- Make the assistant admit doubt. Add to the prompt: mark every number Clear or Unsure, write each price digit by digit, and treat any crossed-out or handwritten figure as UNCLEAR. Every Unsure row goes to the top of the spot-check list.
- Watch the number format. Some suppliers print a thousand as 1.000,00 with a decimal comma. Tell the assistant to report prices with a decimal point and to show the original beside it, then check one large figure by eye.
- Ask for a digital copy on big orders. A scanned quote is fine for $200 of fixings. For the brick or timber package that drives the estimate, a two-minute email to the supplier is cheaper than a mistake.
The total test from the previous section is the safety net here. If the extracted lines do not add up to the printed total, the picture was read wrongly somewhere, whatever the assistant says about its confidence.
Addenda and revised quotes: never overwrite a rate
Rates change in two ways. Suppliers revise a quote, and the tender documents change under you through addenda. Both need a trail, because the question you will one day be asked is "what was the price when you tendered?" Keep a small addenda log beside the rate sheet:
| ID | Date | Source | What changed | Rates affected | Action | Status |
|---|---|---|---|---|---|---|
| ADD-01 | 2026-10-14 | Supplier A, quote rev 2 | Brick price up, validity cut, delivery threshold raised | BR-01 | Rev 1 moved to the archive; rev 2 live | Closed |
| ADD-02 | 2026-10-20 | Tender addendum 1 | Cavity insulation changes from 100 mm to 125 mm | IN-01 | Re-quote requested from supplier A | Open |
| ADD-03 | 2026-10-24 | Drawing revision C | Window schedule changes from three windows to four | None (quantity) | Passed to the takeoff | Open |
Three rules keep it honest. Never type over a rate: add the new row, move the old one to an Archive sheet and record why. Give the Rates table a Status column (Live, Needs re-quote or Superseded) so an addendum that changes a specification can flag every rate it touches. And let the estimate shout when it uses a stale one, for example =IF(INDEX(Rates[Status],MATCH([@Code],Rates[Code],0))="Needs re-quote","RE-QUOTE","") in a spare column next to each line.
Comparing two versions of a quote is a job an assistant does well, within limits. Attach rev 1 and rev 2 and ask: list every difference between these two quotes, in prices, units, quantities, validity, delivery and exclusions, including the footer small print; show each as old wording, new wording. Sample output lists the two changes in the table and header: line 1 from 760.00 to 780.00 per 1000, and validity from 30 days to 14 days. It misses the footer, where free delivery moved from orders over 1,000.00 to over 1,500.00 and "kerbside offload only" appeared for the first time. You find those by reading the footer yourself, and the lesson is to ask for the diff twice, once in each direction, and to name the footer in the prompt. On 6,200 bricks the price rise alone is $124, but the delivery threshold could move a whole order's cost by more.
Scope-gap review: what each quote leaves out
People read what a quote says and miss what it does not. An assistant is useful here because it can be told to mark absence. Write your own scope list for the materials package first, then ask for a matrix with a third category, Silent:
Below is my scope list for materials supply on a single-storey extension, then three
supplier quotes (A, B, C). For each scope item and each supplier, mark Included,
Excluded or Silent, and quote the words that justify it. Assume nothing a quote does
not say. Finish with a list of questions to send to each supplier.
SCOPE LIST: delivery to site; offloading onto site; cutting to size; fixings and
accessories; wastage allowance; lead time; price validity; minimum order or delivery
charge; packaging or pallet returns.
[paste the three quotes]| Scope item | Supplier A | Supplier B | Supplier C |
|---|---|---|---|
| Delivery | Included over $1,500.00 | $65.00 per drop, free over $1,000.00 | Silent |
| Offloading | Kerbside only | Silent | Silent |
| Cutting to size | Excluded | Excluded | Silent |
| Fixings and accessories | Silent | Excluded | Silent |
| Lead time | Silent | Silent | Silent |
| Price validity | 14 days | 30 days | To 2026-11-12 |
| Pallet returns | Silent | Silent | Silent |
The grid makes three gaps obvious. Nobody has stated a lead time, so the programme is guesswork. Supplier A offloads kerbside only, so someone has to carry bricks to the rear of the site, which is labour the rates do not include. And supplier C has said nothing about delivery at all, so its low plasterboard rate may not be comparable to D's. The assistant's list of questions becomes a short message to each supplier, with [first name] and [your name] left for you to fill. For the wider job of comparing bids, AI bid levelling and comparing supplier quotes side by side go further.
Keep the sheet alive: validity dates and re-quote triggers
A rate is a statement about a date. Two formulas in the Rates table do most of the policing, and a conditional format can turn the row red:
Expired =[@ValidTo]<TODAY()
Expiring =AND([@ValidTo]>=TODAY(),[@ValidTo]<=TODAY()+14)
Age in days =TODAY()-[@QuoteDate]Re-quote when any of these happens: the validity date passes, an addendum changes the specification, a supplier announces a price change, a new supplier enters the picture, or the item will make up more than a few per cent of a tender. Treat the shortest validity among your major items as the limit on how long your own quote to a client can hold, or add a clause that lets you adjust for supplier price changes. A client quote valid for 30 days built on a brick price valid for 14 is a promise you cannot keep.
Plan on an hour to set up the sheet and extract the first three quotes while the layout settles. After that, each new quote is a few minutes of extraction and a few more of checking, and the checking is the part to protect. The assistant removes the retyping; the person who knows what a pallet of bricks should cost keeps the final say.
Questions about quote-to-rate-sheet work
Can I upload all the supplier quotes in one go?
One quote per request is safer. Mixing several files makes it easy for an assistant to attach a price to the wrong supplier or merge two units, and a mistake is harder to trace. Extract each quote separately, keep the quote reference on every row, and combine the tables yourself in the rate sheet once each has passed the checks.
Which AI assistant reads supplier quotes best?
Any of the main assistants handles a clean, text-based PDF. The differences show up on scans, photos and odd layouts. Test two on three of your own quotes, compare each result with the printed totals, and choose on the error rate you actually see. Rankings change monthly, so your own test is the only one worth trusting.
Do I need takeoff software for this?
No. A rate sheet is about prices, and takeoff is about quantities from drawings. They meet at the estimate line, where quantity times rate gives the cost. You can build the rate sheet in a plain spreadsheet and fill quantities by hand or from whichever takeoff method you already use.
Further reads
- Masonry Bidding: Review Plans and Specs for Scope Gaps With AI — Scope gaps in plans and specs, from the masonry side.
- Concrete Estimate Template: Volume, Pump and Finishing Lines With AI — A filled-in estimate sheet that could draw rates from this one.
- AI Estimating Software for Small Builders: Six Options Compared — If you decide you want software for the quantities side.
- AI Consulting for a Construction Company: What a First Project Costs — Three gated phases priced in consultant days for a fourteen-person building firm, plus five candidate first projects and the payback test for each.
- AI Tools for Construction Estimating: Bid Checks After the Takeoff — Takeoff tools make quantities; checking the bid is a different job. A comparison with list prices and a five-gate bid check on a worked example.
- Asphalt Price Increase 2026: Update Quotes and Tell Customers With AI — How a paving firm sorts open quotes, re-prices only the asphalt line and writes customer notices, with clause wording and AI prompts that do the drafting.
- Bathroom Remodel Estimate Template: Catch Missing Lines With AI — The hidden lines that slip out of bathroom estimates, a checklist prompt that audits your last three, and a reusable template with allowances for unknowns.
- Best AI for Building Companies: Quotes, Programmes and Site Admin — Tool types for a small building company, sorted by quotes, programmes and site paperwork, with one rear extension priced, programmed and documented.
- Bill of Quantities Template: A Worked Example Checked With AI — A worked bill of quantities with Excel check formulas and AI prompts for codes, descriptions, units and totals, with measurement left to the QS.
- Construction Job Costing With AI: Find Which Jobs Actually Made Money — Export costs from Xero or QuickBooks, add timesheet labour, build a margin table by job and let AI flag the coding errors that hide losses.
- Data Cabling Quote: Count Drops, Cable Runs and Test Lines With AI — A contractor-side data cabling quote for a 12-drop office: drop schedule, cable metres, test and label lines, exclusions and a count check done with AI.
- Demolition Estimate Template: Scope, Waste and Disposal Lines With AI — A demolition estimate template for a rear extension: scope, asbestos and services exclusions, tonnes by waste stream, and skip and disposal lines.
- Finding Construction Bids Worth Pricing: An AI Bid/No-Bid Screen — Where bids come from, a weighted score for fit, margin, competition and client, a prompt that summarises tender packs, and the break-even hit rate.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Microsoft Support page on importing data from a PDF into Excel (Data, Get Data, From File, From PDF); vendor pricing pages for ChatGPT, Claude and Google AI plans; vendor data-control documentation on training defaults for business and consumer plans.