The full workflow, every feature, and the trust model — for visitors who want depth before they reserve a slot. ← Back to home
Heavy-civil GCs receive 5–15 subcontractor bids per package. Estimators spend days normalizing apples-to-oranges scope, chasing missing line items, and second-guessing the low number. And then they pick a winner from gut feel — with no audit trail when the choice gets questioned six months later.
Every "bid management" tool on the market is a filing cabinet. They hold PDFs. They don't read them.
Five steps. No configuration. No template-fitting. Works with whatever your subs send you.
Forward sub bid emails to your tenant inbox — Bid Reasoner auto-matches them to the open invitation. Or invite subs by email; they get a magic link to a portal where they sign up, accept, and upload. PDF, Excel, Word, phone snaps — any format your subs send.
Every line item in every PDF gets extracted, mapped to your scope of work, and normalized so bidders are comparable. Scope-coverage gaps surface automatically — items in your SOW the bidder didn't quote, items the bidder bid that weren't in your SOW. For items a bidder didn't quote, Bid Reasoner plugs a projected price — the peer median across your own bidders by default, sharpened with a government baseline (New York, New Jersey) where the code matches. Works in any state.
Every bidder is scored on six dimensions: price, scope alignment, schedule, compliance, past performance, and risk. The Reasoning Workspace recommends a winner with a confidence score and a paragraph of reasoning — every finding backed by a quote from the actual bid PDF, with page number.
Switch decision modes (Lowest Responsible Bid, Best Value, Schedule Priority, etc.) to see the trade-offs. Run what-if scenarios — "what if the GC absorbs site welding?" — and watch the recommendation update. Click ✓ or ✗ on each finding to feed your judgment back into the system.
Pick a winner. If you agree with the AI, just capture rationale. If you disagree, an override category + reason are required — building a permanent audit record. Bid Reasoner generates the Reasoned Award Memo, Evidence Report, Bidder Scorecard, and four other Word docs automatically. Winner + decline emails go on click.
Every adjustment is labeled. Every projection shows its work. Your senior estimator can still argue with it — that's the point.
Per-package recommendation page with AI bidder evaluations across six dimensions, reasoning evidence backed by quotes from each PDF, click-to-explain confidence scores, and a what-if challenge box. Switch between seven decision modes to see how price-priority vs risk-priority changes the winner. The single most important page in the product.
When you award against the AI's recommendation, an override category and reason are required — not optional. Every disagreement creates a permanent audit record. When an owner asks "why this bidder?" six months later, the answer is one click away.
Seven Word document types — Reasoned Award Memo, Bid Reasoning Report, Evidence Report, Bidder Scorecard, Executive Summary, Clarification Recommendations, Project Handoff Rationale — populated from the underlying data. Real .docx, not screenshots. Pull a fresh copy any time.
For every bidder, surfaces the line items in your SOW that the bidder didn't quote — and the items the bidder bid that weren't in your SOW. Find scope gaps and scope creep before award, not after. Works on any project, any code system.
Not opaque AI: unbalanced unit prices (≤ $1.00), peer outliers (>2× or <0.5× peer median), total-bid outliers (>20% deviation), and front-loaded mobilization (>10% of total). Click any chip to filter the matrix to those flagged items.
Every bidder's line items are mapped to canonical names and normalized against the peer median of your own bids — works in any state, no government data required. Where a public catalog exists, it sharpens the estimate: select states such as New York and New Jersey DOT baselines are built in today, with more states added as partners come online. For items a bidder didn't quote, the projection plugs the peer median first, then a government baseline where the code matches. Specialty or non-catalog items the comparison view flags as coverage gaps explicitly — projected cost is a floor, not a ceiling.
Every Q&A with a sub is a tracked thread, not a buried email. Subs answer through their portal. The system flags when a response revises pricing — your leveling table updates accordingly. Closed clarifications stay attached to the award memo.
Invite subs by email — they get a magic link to a portal where they sign up, accept or decline, upload bids through a 4-step wizard, and acknowledge addenda. Or forward bid emails to your tenant inbox; sender is auto-matched to the open invitation.
When the project owner asks "why this bidder?", the answer needs to hold up to legal review.
Every customer's data is row-level isolated. We run 21 read-only invariant checks before every deploy that confirm zero rows leak between tenants. Your bids, scopes, and awards are never visible to another GC.
Feature changes, plan changes, budget changes, manual retries, email resends. Every super-admin action is logged with timestamp, reason, and before/after values.
When our team needs to look into your tenant for support, we enter Support Mode with a required reason + ticket reference. All writes are blocked at the middleware layer for the duration. 60-minute hard auto-expiry. Every blocked write is logged.
Every adjustment carries the source document, the page number, and the math. No black-box scoring. Override anything; the override is part of the record.
Bid Reasoner is live in private beta with a hand-picked group of heavy-civil general contractors. Founding partners get introductory pricing, founder-direct support, and direct influence on what the model learns from their team's decisions. Once the cohort is full, the founding rate goes away.