Automating Bid Ingestion for a Commercial HVAC & Plumbing Subcontractor
A Detroit-based HVAC and plumbing subcontractor with commercial clients across the Midwest was drowning in manual bid processing. We built an intelligent pipeline that transformed their estimating workflow.
Industry
Construction & HVAC
Location
Detroit, MI
Timeline
10 weeks
Challenge
60+ hours/week on manual bid processing, missing 60% of opportunities
Strategic Approach
Knight
Knight Labs' approach
Rook
End-to-end pipeline execution
Bishop
Precision document classification
Queen
Full-scope system integration
Client
Midwest Mechanical Services
Timeline
10 weeks
Focus Areas
Tech Stack
The Challenge
The client's estimating team was spending 60+ hours per week manually reviewing incoming RFPs, construction drawings, and bid invitations from general contractors across Michigan, Ohio, Indiana, and Illinois. Each bid package contained hundreds of pages of architectural plans, mechanical specifications, and scope documents that needed to be parsed, categorized, and entered into their estimating software. The manual process meant they were only able to respond to roughly 40% of incoming bid opportunities, leaving significant revenue on the table. During peak season, estimators were burning out and making costly errors in scope interpretation that led to underbid projects.
The Solution
We designed and deployed an end-to-end bid ingestion pipeline that automatically receives bid invitations via email and FTP, extracts and classifies documents using computer vision and large language models, and structures the data for their estimating platform. The system parses architectural drawings to identify HVAC and plumbing scope, extracts key project details (square footage, building type, mechanical specifications), flags addenda and change orders, and routes categorized bid packages to the appropriate estimator based on project type and capacity. We integrated directly with their existing estimating software via API, eliminating manual data entry entirely.
The Result
Bid processing time dropped by 90%, from an average of 4 hours per bid package to under 25 minutes. The team went from responding to 40% of opportunities to over 85%, capturing an estimated $2.3M in previously missed commercial contracts in the first year. Estimator overtime was eliminated, and scope interpretation errors decreased by 75%.
Bid Processing Time Reduced
Bid Response Rate
Revenue Captured (Year 1)
Scope Errors Reduced