---
title: "Leveling subcontractor quotes"
role: estimator
slug: bid-leveling
status: works with supervision
updated: 2026-08-23
canonical: https://aiagents.construction/roles/estimator/bid-leveling/
---

## The job

Bid day compresses this job brutally: quotes arrive as PDFs, emails, and phone notes in the final hours, and someone has to line them up — same units, same scope buckets, exclusions surfaced — before award. Missed exclusions here become change orders later.

“Done”: one leveled sheet per trade with every quote mapped to the same scope lines, every stated exclusion listed, and every *silence* flagged for a phone call.

## Use

1. **Collect quotes into one folder per trade.** PDFs and emails as-is; don't retype anything.
2. **Run extraction.** A general-purpose model [[1]](#cite-1) or precon product [[2]](#cite-2) pulls price, scope lines, inclusions, exclusions, and alternates into a common table.
3. **Review the mapping.** Check that differently worded scopes were not incorrectly treated as the same.
4. **Chase the silences.** Flagged gaps become the call list before award. This is the point; don't skip it.

## Build

This is a strong DIY candidate: document extraction plus tabulation. A v1 is a prompt template asking for every scope line, inclusion, exclusion, and price, one row per scope item, with ambiguity marked. Run it per trade folder and output to a spreadsheet. Keep the transcript in the bid file.

## Deploy

Pilot on the last awarded trade package where a scope gap later caused trouble. If the agent's leveled sheet would have surfaced that gap, it earns a place on the next live bid—still with the estimator reviewing every mapping.

## Manage

- **The agent proposes; the estimator decides.** Every inclusion/exclusion judgment is human.
- **Extraction errors are silent.** Spot-check totals against the raw quote for at least one quote per trade, every time.
- **Sub pricing is confidential.** Confirm hosted-model data-retention terms before quotes go in.

## Tools cited

1. [Claude (Anthropic)](https://www.anthropic.com/claude) — General-purpose model with vision and document extraction; suits the DIY build described here. Usage-priced; no per-seat construction license. Last verified 2026-08-23.
2. [Ediphi](https://www.ediphi.com) — Cloud preconstruction and estimating platform; documented as an export target in takeoff workflows and offers structured bid-management features. Last verified 2026-08-23.

## Changelog

- 2026-08-23 — Page created.