AI pre-construction for commercial GCs

Turn a 2,000-page RFP into trade scope packages in hours, not weeks.

TeraContext.AI reads the specs, drawings and addenda, classifies every section to your WBS, and builds the scope packages, sub bid analysis and GC proposal. Your estimators approve every classification.

Bring a current RFP See the 6-step workflow
  • Estimators approve every classification
  • Answers cited to section, sheet and page
  • Can run in your environment
WBS tree for a municipal library bid set: 5,998 classified items, with spec sections listed by code under Procurement and General Requirements
Classifications list: each item shows its WBS code, source text, classification method and a confidence score, with Confirm and Flag buttons

Review & Classify, on a public municipal library bid set

How it works

From the RFP landing to the proposal shipping, in six steps

Step 01

Upload the whole bid set

Drop in spec volumes, drawing sets, addenda, contracts and geotech reports. Processing starts on its own.

You give it
PDF, DOCX, XLSX, PPTX, images
You get
Every spec section split out and browsable by code
Time
10–20 minutes for a 1,000+ page spec book
Specs screen: spec sections listed by WBS code, from General Requirements through Concrete, Masonry and Metals
Step 02

Every section classified to your WBS

A 7-phase pipeline built for construction documents maps each spec section and drawing note to a WBS code and scores its own confidence.

Formats
11 industry-standard WBS formats, or your own
You get
A code, a source and a confidence score on every line
Drawings
A vision model reads every sheet
Review and Classify screen: a WBS tree totalling 5,998 items beside a list of classifications, each with its code, source and confidence score
Step 03

Your estimators make the call

Confirm the high-confidence classifications in bulk, typically 70–80% of sections, and spend your time on the edge cases. Nothing is packaged until your team approves the WBS.

Bulk
Confirm everything above 90% in one click
Edge cases
Correct, flag or reclassify line by line
Control
WBS approval gates scope packages
Classification review: Confirm All above 90%, Confirm All, Reclassify All and Verify buttons above a list of items with confidence bars and per-line confirm and flag buttons
Step 04

Scope packages and SOWs by trade

Approved sections roll up into trade packages. Rename, split and merge until the breakdown matches how you bid, then generate a statement of work per trade in your own template.

Packages
Export as CSV, PDF, Word or Markdown
SOWs
Scope items, alternates and exclusions per trade, as .xlsx
No overlap
Each task item belongs to exactly one trade
Scope Packages screen: 122 trade packages such as Access Doors, Air Barrier and Aluminum, each with CSV, PDF, Word and Markdown export
Step 05

Sub bids, checked against the scope

Invite subs from your directory to bid on specific packages. Upload each response and the AI reads it against the package it was meant to cover.

Directory
29 standard construction trades
Compare
Bids side by side, per package
Each bid response is analyzed for
  • Scope coverageWhat the sub actually priced
  • ExclusionsIncluding the ones buried in the fine print
  • AlternatesOffered substitutions and options
  • QualificationsConditions attached to the number
  • Risk flagsItems your team should look at first
Step 06

The GC proposal, assembled

Accepted sub bids come together into your proposal response, checked against the original RFP before it goes out.

Check
Compliance check against the RFP
Export
PDF or Word, for final editing
The assembled response includes
  • Coverage matrixWhich bid covers which scope
  • Gap analysisScope nobody has priced yet
  • Per-trade narrativesWritten from the accepted bids
  • Executive summaryThe response at a glance

See every step in detail →

Before and after

Weeks of reading become hours of review

500–2,000+pages per spec book
50–500drawing sheets, 9+ disciplines
29trades to scope and bid
By handWith TeraContext.AI
Classify the spec bookRead every section and sort it by trade. Days of estimator time.Every section classified with a confidence score. Bulk-confirm the 70–80% that are high-confidence.
Build scope packagesAssembled section by section, for each trade.Generated by division, then renamed, split and merged to match how you bid.
Review sub bidsRead each response hunting for exclusions and gaps.Each response analyzed for coverage, exclusions, alternates, qualifications and risk flags.
Assemble the proposalPulled together under deadline pressure.Coverage matrix, gap analysis and narratives generated, then checked against the RFP.
Find a requirementSearch PDFs volume by volume, sheet by sheet.Ask in plain English. Get an answer cited to spec section, sheet and page.

Missed sections become change orders. The point of the pipeline is that every section is accounted for before the packages go out.

On a real bid set

A public municipal library and parking garage, decomposed

Not a mock-up. These are screens and counts from the platform after processing one public bid set.

10documents uploaded
5,998spec and drawing items classified
246drawing sheets indexed by discipline
122trade scope packages and SOWs
Spec query answering 'What are the concrete strength requirements?' with values from structural sheets PS-001, S-101 and S-403 and spec sections 03 30 00, 31 23 00 and 05 50 00, citing 25 sources
Ask the spec book, get a cited answer. "What are the concrete strength requirements?" returns values from the structural notes and four spec sections, with 25 sources.
Drawing index: 246 sheets counted by discipline, 54 architectural, 41 civil, 26 electrical, 55 structural and more, above a sheet index with number, title, discipline and scale
Every drawing sheet read and indexed. 246 sheets sorted into eight disciplines, with 2,803 text sections extracted from notes, keynotes and schedules.
Generated SOWs: trades such as Site Utilities, Roofing and Testing, each with its count of scope items, alternates and exclusions and an .xlsx download
A statement of work per trade. Scope items and exclusions counted for each trade, downloadable as .xlsx in your template.
Scope Packages screen listing trade packages with section counts and export buttons
122 trade packages, ready to send. Each one exports to CSV, PDF, Word or Markdown.
Why estimators can trust it

Built so you can check the work

Your team has the final say

Every classification carries a confidence score and waits for an estimator. Scope packages are built only from an approved WBS.

Every answer shows its source

Query answers cite spec section codes, drawing numbers and page ranges, so checking one takes seconds.

Specialized, not a generic chatbot

Seven phases purpose-built for construction documents, with RAG, GraphRAG and a vision model for drawings.

Your documents stay yours

Encrypted in transit and at rest, role-based access, audit trails, and never used to train AI models. It can also run in your environment.

  1. PDF extraction
  2. Table extraction
  3. Section splitting
  4. Embedding
  5. WBS classification
  6. Graph building
  7. Cross-ref validation
Jim Smith, founder of TeraContext.AI
Who builds it

Jim Smith, Ph.D., P.E. is a registered professional engineer whose training spans electrical and structural work. He builds TeraContext.AI in McLean, VA as part of Joshua8.AI. Design partners work with him directly.

About the founder and the company →
Built for commercial construction

11 industry-standard WBS formats, or your own

WBS formats, including

  • Commercial spec divisions
  • UFGS
  • PennDOT-408
  • FERC Generation
  • EPRI GN-COA
  • DOE Nuclear
  • DOE Buildings
  • NAHB Residential
  • + Custom taxonomies

Clone any standard format and add, rename or reorganize codes to match your firm's own system.

Project types

  • Multi-Family
  • Office
  • Municipal
  • Parking
  • Retail
  • Hotel
  • Strip Mall
  • Mixed Use

Also in the platform: automated document review for spec and drawing conflicts, and six documented workflows.

Early Access · design-partner program

Bring a current RFP. See it decomposed on your documents.

TeraContext.AI is in Early Access. We work with a small group of general contractors, on their own bid sets, and they shape what gets built next.

  1. Intro call. We learn how your team bids today.
  2. Pick a project. A current or recent RFP you know well, so you can judge the output.
  3. Run it together. Your estimators check the classifications and packages against what they would have produced by hand.
Request Early Access Read the datasheet

A personal reply within one business day.

A good fit if

  • Your spec books run 500+ pages
  • Decomposition and scope packaging take days or weeks
  • You bid to subs across many trades
  • Missed scope has cost you in change orders
  • You carry several RFPs at once

Probably not a fit if

  • You build small residential work without complex spec books
  • Your projects have no multi-trade sub bidding
  • Specs run under 100 pages per project
Perspectives

One set of documents, every seat at the table

General Contractor Owner Architect Subcontractor Lender Insurer

Watch all the role videos, plus the affordable housing view →

From the blog

Latest on AI and construction documents

All posts →

Early Access program

Your next spec book, decomposed with your team in the room.

Send us a line about your projects and we will set up a working session on your own documents.

Request Early Access See How It Works