How to Use ChatGPT for Construction Estimating

Bid day usually comes with a stack of documents: a spec book, two or three addenda, subcontractor quotes in different formats, and last year's cost data in a spreadsheet nobody has opened since the job closed. Most of an estimator's week goes into reading and organizing that material before any pricing decision gets made.

ChatGPT can take over a large share of that reading and organizing. It can review specs, compare addenda, line up quotes, check math, and draft RFIs and proposal language. It shouldn't produce your quantities or set your prices, and it needs a few settings and habits in place before it's safe to use on real bids.

This guide covers how to set up ChatGPT for estimating, how to use it at each stage of an estimate, prompts you can adapt, and where its limits are.

ChatGPT works best as a document reader, analyst, and drafting assistant for estimators. It reads specs and quotes, runs calculations in its data analysis tool, and drafts written material, while measured quantities and pricing decisions stay with the estimator.

Create a Project for Your Estimating Standards

ChatGPT projects group chats and files together, with custom instructions that apply to every conversation inside the project. Projects accept PDFs, spreadsheets, documents, and images, and Pro, Business, and Enterprise plans support up to 40 files per project.

A practical setup is one project for your company standards, holding:

• Standard inclusions and exclusions
• Two or three strong past proposals
• Your cost code or CSI division list
• Scope templates for the trades you bid most
• Your RFI format

Then add project instructions like these:

You support the estimating team at [Company], a [trade] contractor.

For every scope item, cite the document, section, and page.

Run all calculations with data analysis and show the formulas.

Do not estimate quantities or prices unless asked.

If information is unclear or missing, list it as a question.

The calculation instruction is the one most estimators skip. When ChatGPT does arithmetic in its written response, mistakes can slip in. When it runs the numbers through data analysis, you can see the formulas and check them.

On Business plans, projects can be shared with teammates, so the whole estimating team works from the same exclusions, templates, and instructions.

Connect the Folders Where Your Bid Files Live

If your bid documents sit in SharePoint, Box, or Dropbox, you can connect those services to ChatGPT's file library. Once connected, you can add files to a conversation without downloading and re-uploading them, and ask ChatGPT to work across all the files in a folder, such as every quote received for one trade package.

Check How Your Plan Handles Client Documents

Drawings, specs, and quotes are often confidential. OpenAI states that on Free, Plus, and Pro plans, information in projects may be used to train its models if the "Improve the model for everyone" setting is turned on. Business and Enterprise plans handle data under OpenAI's business privacy terms.

Before uploading client documents, confirm your plan, check that setting, and follow your company's policy and any confidentiality terms in the bid documents.

Step 1: Review Specs and Pull Out Your Scope

Upload the spec sections for your trade and ask for a structured scope list.

Review the attached Division 26 specifications. List every item that affects our scope or cost: equipment, materials, testing, commissioning, warranties, submittals, training, and special requirements. Group items by spec section with page numbers. List separately anything that may belong to another trade.

Spot-check the costly items yourself, especially third-party testing, extended warranties, and owner training. The section and page references make that check quick.

Step 2: Compare Addenda Against the Original Documents

Addenda can change scope with one revised sentence. Upload the original and revised sections together.

Compare the original Section 26 05 19 with the version in Addendum 2. List every addition, deletion, and changed requirement. For each one, say whether it likely increases, decreases, or doesn't affect our cost, with a one-line reason and page references.

The cost direction on each change tells you what to reprice first.

Step 3: Level Subcontractor and Supplier Quotes

Quotes arrive as PDFs, emails, and spreadsheets in different layouts. With the quotes in a connected folder or uploaded to the chat, ChatGPT can line them up in one file.

Create an Excel comparison of these four electrical quotes. Use one row per scope item and one column per bidder. Include base price, alternates, inclusions, exclusions, qualifications, and quote expiration date. Highlight any item one bidder includes and another excludes.

The exclusions column is where most of the value sits, because a low number often reflects missing scope. Check the final comparison against the original quotes before carrying any number into your estimate.

Step 4: Check Quantities, Conversions, and Totals

ChatGPT's data analysis tool can check the math in your estimate spreadsheet: unit conversions, waste factors, quantity rollups, and totals.

Using data analysis, check the attached estimate. Verify unit conversions, recalculate every extended total, and list any line where quantity times unit price doesn't match the total. Also flag hardcoded totals, broken formulas, and lines with zero quantity or zero price. Show the formulas you used.

This works on quantities you've already taken off. It doesn't replace the takeoff itself.

Step 5: Compare Your Estimate With Historical Costs

If you keep past bids or job cost reports in spreadsheets, ChatGPT can compare a new estimate against them.

Compare the unit costs in this estimate with our job cost reports from the last three similar projects. Use data analysis to calculate the difference for each matching cost code, and list any line more than 15% above or below the historical average with the source project for each comparison.

The results are only as consistent as your old data. Spreadsheets with different cost codes or naming across jobs usually need cleanup before this comparison is reliable.

Step 6: Draft RFIs and Proposal Exclusions

Keep rough notes during your review and let ChatGPT turn them into RFIs and proposal language.

Turn these notes into RFIs using our RFI format from the project files. Each RFI should cite the drawing or spec section, describe the conflict in one or two sentences, and state the assumption we'll price if no answer arrives before the RFI deadline.

Notes:

[paste notes]

For the proposal, ask for a draft in the style of your past proposals, with pricing fields left as placeholders so no figure from an old job carries over by mistake.

• Quantity takeoff: ChatGPT can describe what's on a drawing, but measured quantities belong in takeoff software built for scale and symbol counting.
• Pricing judgment: Markup, contingency, and how competitive to be depend on your market, backlog, and client.
• Current material prices: Use supplier quotes and your pricing database, not general answers about material costs.
• Confident errors: ChatGPT can misread a document or fill a gap with a plausible answer, which is why section references and spot checks are part of every prompt above.
• Consistency across estimators: Without shared projects and instructions, each estimator ends up with different prompts and different results.