Written by the Miriad team with AI assistance. Figures link to their sources; the quote set below is illustrative, not a real case. Updated 2026-10-21.
TL;DR: To compare subcontractor quotes that arrive in different formats, fix the columns first, then let AI copy every quote into that one table without calculating anything. The useful output is not the table but the list of gaps: lines one supplier priced and another did not, units that differ, exclusions hidden in footnotes. The estimator checks every figure against the original.
Every estimator in building services knows the moment. The bid is due on Wednesday, and by Monday afternoon you have three answers for the plumbing and heating materials: a tidy PDF, a spreadsheet with its own idea of what a line is, and a photo of a printed quote taken on a van dashboard. On top of that come subcontractor quotes for insulation and controls, each with a different scope split. Before you can price anything, you need them side by side. This article belongs to our guide on how an MEP contractor can use AI to win more bids, and covers the comparison step in detail.
Why are supplier and subcontractor quotes so hard to compare?
Because every supplier answers a different question. One prices exactly your list, one bundles items into lump lines, one quotes per box instead of per metre, and a specialist subcontractor quietly excludes the part of the scope you assumed was theirs. The prices are rarely the hard part. The scope boundaries and units are.
The format is the visible problem. The invisible one is what each quote leaves out. A controls subcontractor who prices "BMS integration" may or may not include the commissioning visit; an insulation sub may exclude valve jackets; a wholesaler may quote fittings "as available". If those gaps reach the bid unnoticed, they come back later as a loss on the job or as a change order argument with the general contractor.
This work lands on the people who are hardest to replace. In AGC's 2025 workforce survey of firms with $50.1 million to $500 million in revenue, 75% of firms with estimating openings reported difficulty filling them (AGC, 2025). Every hour an estimator spends retyping quotes is an hour not spent on the takeoff or the labour allowance.
What columns does the comparison table need?
Fix the columns before you open the first quote. An estimator needs item, size or specification, unit, quantity, unit price, line total, delivery or lead time, and a note column for exclusions and conditions. For subcontractor quotes, add a scope line that says in one sentence what the sub is and is not doing.
| Column | Why it is there |
|---|---|
| Item and specification | So "ball valve" from A and "BV" from B land on the same row |
| Unit | Each, metre, box, lot; mismatches are where comparisons break |
| Quantity | Copied as quoted, so you see if a supplier changed your quantity |
| Unit price and line total | Both, so a lump line is visible as a lump line |
| Delivery or lead time | Long-lead items affect your programme, not just your price |
| Included / excluded | Scope boundaries for subcontractors, conditions for suppliers |
| Validity | When the quote expires against your bid date |
| Source | Quote reference and page, so every figure can be checked |
The last column is the one people skip and the one that matters most. Without it, nobody can check the table quickly.
How do you get AI to build the table without inventing prices?
Tell it to copy, not to estimate, and to mark every gap. A general AI assistant reads PDFs, spreadsheets and photos well enough to transcribe them, but if you ask it to "compare the quotes" it may smooth over missing lines or calculate figures you never asked for. A strict instruction prevents most of that.
The instruction we use as a starting point:
You are helping an estimator compare quotes. Build one table from the quotes
below with these columns: item, specification, unit, quantity, unit price,
line total, delivery, included/excluded, validity, source (quote ref + page).
Rules: copy values exactly as written. Do not calculate, convert or guess.
If a value is missing or unreadable, write MISSING or UNREADABLE.
After the table, list: (1) items present in one quote but not another,
(2) rows where units differ, (3) every exclusion or condition you found.
Two habits make it safer. Paste one quote at a time if they are long, and ask for the gap lists separately. And keep the original files next to the table, because the AI output is a draft, not a record. Dodge Construction Network and CMiC found that only 26% of contractors say their data is in good shape (Roads & Bridges, 2025); a table without traceable sources makes that worse, not better.
Write the instruction down and share it with everyone who prices work. In a 2025 Houzz survey, 92% of contractors who used AI had no formal AI training (Houzz, 2025). In practice that means each estimator invents their own prompt, and the tables stop being comparable from one bid to the next.
Example: three quotes for a plumbing and heating package
Example (illustrative, not a real case): A 15-person building services firm is pricing the plumbing and heating package for a two-storey office fit-out. The bid package asks for copper distribution, a small plant room with two boilers, and radiators on the upper floor. Wholesaler A sends a PDF that follows the estimator's material list line by line. Wholesaler B sends a spreadsheet with "valves and fittings" as one lump line and a separate delivery charge. Wholesaler C sends a phone photo of a printed quote, with pipe priced per box and a handwritten note: "rads 4–6 weeks".
The estimator pastes the three quotes into the assistant with the instruction above. The table comes back with the A lines complete, the B lump line on its own row with MISSING in the unit price column for every individual valve, and the C pipe lines flagged as a unit mismatch. The gap list shows something nobody had spotted: neither B nor C included the pressure-reducing valve set shown on the plant room drawing.
The estimator does three things by hand. Rings B for a breakdown of the lump line. Converts C's box prices to metres from the box length printed on the quote. Adds the pressure-reducing valves as a separate line to request from all three. The radiator lead time goes into the proposal as a clarification. None of these steps needed AI; all of them were easier to see because of it.
What should the estimator check before the table goes into the bid?
Check every figure against its source, every unit against your takeoff, and every exclusion against the bid package. The table is only trustworthy when someone has confirmed that each price exists in a real quote, that quantities match your takeoff, and that the scope gaps are either priced or written into your exclusions.
A short checklist:
- Source check. Every price in the table has a quote reference and page, and you have opened it.
- Unit check. Units match your takeoff; conversions were done by a person and noted.
- Quantity check. Where a supplier changed a quantity, find out why before you accept it.
- Gap check. Each missing item is either requested, priced by you, or listed as an exclusion in the proposal.
- Validity check. No quote expires before the expected award date without a note.
- Lead time check. Long-lead items are flagged for the programme and, where relevant, for the submittal schedule.
If quotes keep arriving late, the problem sits earlier in the process. Why MEP bids are lost before price is compared maps where those days go.
Where should the comparison live after the bid?
Keep it with the bid, in the same structure every time. A comparison table saved in a fixed format becomes pricing history: which supplier is reliable on lead times, which sub always excludes commissioning, which unit prices moved. That history is the part of estimating knowledge that usually lives only in a senior estimator's memory.
Save the table, the original quotes and a two-line note on the outcome in the bid folder. After a dozen bids, you can ask an AI assistant to summarise patterns across them, as long as it reads your files rather than its general knowledge. How to get the rest of that knowledge out of one person's head is covered in capturing estimating knowledge before it leaves.
Our Business Brain Starter Kit, a workbook and a chain of prompts that turn company knowledge into plain files with a source next to each line, is being built for this kind of record. Get notified when it launches. The full bid cycle is in the MEP contractor guide to AI and bids.
FAQ
Can AI read a photo of a printed supplier quote?
Usually yes, if the photo is sharp and the page is flat. General AI assistants transcribe printed quotes reasonably well, but handwriting, stamps and folded corners cause errors. Ask the AI to mark anything it cannot read as UNREADABLE rather than guessing, and check those lines against the photo yourself. For long quotes, ask the supplier for a PDF or spreadsheet instead.
Should the AI convert units, for example boxes to metres?
No. Let the AI flag the unit mismatch and do the conversion yourself, using the box length or pack size printed on the quote. Conversions are where silent errors enter a bid, and an AI that converts without being asked may use a pack size from general knowledge instead of the supplier's own figure. A flagged mismatch costs a minute; a wrong conversion can cost the job.
How do I compare subcontractor quotes with different scope splits?
Add a scope column and write, for each subcontractor, one sentence on what they include and exclude, using their own wording. Then list the scope items from your bid package and mark who covers each one. Anything covered by nobody becomes either a request, a line you price yourself, or an exclusion in your proposal to the general contractor.
Is it safe to upload supplier quotes to an AI tool?
Quotes contain prices that suppliers give you in confidence, so treat them like client documents. Use only an AI tool your firm has approved for business data, check its settings on whether inputs are used for training, and do not upload drawings or prices to a personal account. When in doubt, remove supplier names before pasting.
Sources
- Associated General Contractors of America, 2025 Workforce Survey Results – $50.1 Million to $500 Million Results, 2025. https://www.agc.org/sites/default/files/users/user21902/2025_Workforce_Survey_50to500M_V2.pdf
- Roads & Bridges, reporting Dodge Construction Network and CMiC, Report: Most Contractors Expect AI to Change How Projects Get Built, 2025-12-08. https://roadsbridges.com/technology/news/55337001/report-most-contractors-expect-ai-to-change-how-projects-get-built
- Houzz, Houzz Releases Inaugural 2025 U.S. State of AI in Construction and Design Report, 2025-07-16. https://www.houzz.com/press/993/Houzz-Releases-Inaugural-2025-U-S-State-of-AI-in-Construction-and-Design-Report