Struction
Also known as: Struction.co, Struction AI
AI preconstruction workspace that reads a project's drawings, specs and BIM model, drafts trade scopes of work, finds risks, RFIs and scope gaps, and cites the source sheet for every item, with approval before any AI change.
Struction is an AI preconstruction workspace for general contractors. Teams upload a project's drawings, specifications, bid documents, schedules and BIM model, and Struction reads and labels each document, organizes the set by discipline, and writes a first draft scope of work for every trade it finds, in about an hour for a typical project. It also surfaces risks, RFIs and scope gaps, and supports manual, assisted and automated quantity takeoff.
The product is built around traceability. Every scope item, quantity and risk links back to the sheet, spec section or model element it came from, and project chat, which has every document in context, answers questions with clickable citations and can create, combine or split scopes, resolve risks and pull quantities from schedules.
Any change chat proposes appears as an approval card to accept, edit or deny first, estimators mark items as verified before sending to subcontractors, and every scope keeps a full version history of AI and human changes with restore. Organizations can upload example scopes and write preferences so drafts match their own format, and finished scopes export to Excel for bid invitations.
Struction offers one free project, a $50 monthly plan, a $500 per user per month plan for general contractors, and a custom Enterprise tier that adds unlimited users, unmetered chat and a choice of chat model across Google, Anthropic and OpenAI. The company says it does not train on customer project data.
Vendor details
Canonical URL
https://struction.com
Category
Enterprise operations agent
Funding status
Private company (Struction Incorporated).
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Struction ingests uploaded drawing sets, specifications, bid documents, schedules and BIM models, reading notes, labels and schedules across the set, and supports manual, assisted and automated quantity takeoff. Its AI chat can create and modify scopes of work, resolve risks and RFIs, and pull quantities from schedules, and finished scopes export to Excel. Struction names no outside system integration, for example with ERP, document management or project management tools.
In practice
A general contractor has five days to bid a school renovation with 45 drawings and 47 spec sections. Struction reads the set, drafts a scope for every trade with each inclusion linked to its sheet, and flags the scope gaps nobody has picked up.
An estimator notices the drawings include access control that never got its own scope. She asks chat to create one without overlapping Electrical, reviews the approval card, accepts it, and refines two items.
A preconstruction manager wants every sub's bid form in the company's usual wording. The team uploads past scopes as exemplars, and new projects generate and export scopes in that format.
Sources & related URLs
Agentic Index coverage score
6.0 / 14 capabilities · 43%
| Integrations & Tool Calling | Partial |
|---|---|
|
The agent reads and writes only inside Struction's own workspace. Chat can create, rename, combine or split scopes of work. It can create, resolve or rerank risks and RFIs, pull counts from a schedule into a quantity and find duplicates of a symbol. Documents come in by upload (drawings, specs, bid documents, schedules and BIM), and scopes go out by export to Excel for subcontractors and ITBs. A person carries the files in and out. Struction names no construction platform, document management, ERP or estimating integration, and no connector types, credential scoping or custom tools. Sourcestruction.com/docs/editing-with-chat, /docs/upload, /docs/scopes-of-work; readread 2026-09-15 |
|
| Workflow Orchestration | Full |
|
A multi stage generation pipeline runs on upload, with deterministic steps, approval gates and version control. Struction reads the documents, labels each one, organizes them by discipline and writes a first draft scope of work for each trade. It then surfaces risks, RFIs and scope gaps. Items that fit no trade go to scope gaps instead of being forced into the wrong scope. Deterministic and agent steps mix. Quantities can be manual, assisted or automated, counts land on a scope item with the sheet measured, and approval cards sit between chat actions and data changes. Scopes carry major.minor version history with restore. Background jobs run on the item while the user keeps working, and one job per item prevents collisions. Struction names no workflow authoring surface, explicit retries, fallback paths or reuse across teams. Regeneration reruns on request. Sourcestruction.com/docs, /docs/scope-creation, /docs/scope-editing, /docs/editing-with-chat; readread 2026-09-15 |
|
| Knowledge Grounding & RAG | Full |
|
Struction builds a persistent, linked map of a project's own documents and answers from it with citations. Users upload drawings, specs, bid documents and schedules, and Struction reads them, labels each one, organizes them by discipline and handles replacement sets. Struction lists five ideas behind the product, and one is that every source is connected. Scopes, drawings, specs and issues are linked by their relationships and navigated as trade and document maps that include the federated BIM model, with callouts, details and spec references linked into one chain. For project chat, Struction has read every drawing, spec, addendum and model element on the project. Cited answers carry a clickable chip to the page, scope, issue or spec section, and every output has a source. Context attached to chat is shown before sending, and chat never works with context. New project knowledge enters without retraining, and Struction says it will not train on project data. Struction names no source permissioning and no handling of conflicting documents beyond manual replacement. Sourcestruction.com/docs, /docs/cited-answers, /docs/context-aware, /landing; readread 2026-09-15 |
|
| Human Oversight & Guardrails | Full |
|
An approval card appears before any chat action changes project data. The user accepts, edits or denies it before it happens, so chat never silently changes the project, and bulk actions are grouped into a single card. Scope Editing adds a verify step, where users mark items as verified as they confirm them, turning an AI first draft into a scope they stand behind. Regeneration warns before overwriting edits a user has made. Context attached to chat is shown before sending. The gate is always on, and Struction names no setting by action or approver and no way to route approvals to other channels. Sourcestruction.com/docs/editing-with-chat, /docs/scope-editing, /docs/context-aware; readread 2026-09-15 |
|
| Security, Identity & Governance | Not documented |
|
When a project is created, the creator and their team are the only ones who can edit it, and changes are stamped with who made them. Struction says it won't train on customer project data. The Enterprise GC tier is for larger builders with scale, security and unmetered chat, without naming a control. Struction names no roles, SSO, SAML, SCIM or MFA, and no attestation, certificate or trust center. It shows no compliance badge and publishes no security or privacy page. Sourcestruction.com/docs/sharing, /pricing, /landing; readread 2026-09-15 |
|
| Observability & Auditability | Full |
|
Every AI change and every user edit is recorded and reversible, with version history a customer can inspect and restore. Every scope carries its full history. An AI change, such as a regeneration, bumps the major version number, and a user edit bumps the minor. Version history lists previous versions with a summary of what changed, a count of items added and removed, timestamps and who made the change. Risks and RFIs are stamped with who changed them and when. Background jobs show progress on the item being worked on. The record of what the agent changed and when is kept separately from the chat. The record covers changes to outputs, not each tool call. Struction names no step level inspection of prompts and model calls, no export to monitoring tools and no retention by plan. Sourcestruction.com/docs/scope-versioning, /docs/version-control, /docs/editing-with-chat; readread 2026-09-15 |
|
| Memory & State Persistence | Not documented |
|
An organization can upload exemplar scopes so Struction matches its format. Organizations and individuals can also write preferences, which are system prompts that steer generation and chat. These persist because a person wrote them down. The project map of documents, scopes and issues persists as project data. Struction names no conversation or long term memory that the agent keeps across sessions. Sourcestruction.com/docs/scope-customization, /docs; readread 2026-09-15 |
|
| Deployment & Data Residency | Not documented |
|
Customers reach a hosted web workspace by signing in at struction.com, and projects are uploaded and processed in the hosted product. Struction names no dedicated, VPC, on premises or self hosted option, including on the Enterprise GC tier, which mentions security without naming a deployment mode. No hosting region or residency choice is published. Sourcestruction.com/docs, /pricing, /faq; readread 2026-09-15 |
|
| Prebuilt Agents / Templates / Packs | Not documented |
|
Scopes are created automatically when a project is created, one for each trade Struction identifies in the documents. Each first draft reflects that specific project, not a generic checklist. The only templates are the customer's own exemplar scopes, uploaded in Settings, and Struction names no prebuilt agents. A Specialty Contractors plan is offered as early access by request. Sourcestruction.com/docs/scope-creation, /docs/scope-customization, /pricing; readread 2026-09-15 |
|
| Triggers & Channel Coverage | Partial |
|
Project chat inside the product carries the agent's work, and work starts only when a person acts. Chat can do almost anything a user can do by clicking and is context aware to the current view. Processing starts when a user uploads documents. Struction names no event triggers, and a new drawing set arriving does not start processing on its own. Document version replacement is manual, and Struction says it will soon handle replacement itself. The Dispatch email is a newsletter, not a channel into the product. Sourcestruction.com/docs/editing-with-chat, /docs/upload, /docs/version-control; readread 2026-09-15 |
|
| Model Flexibility & Routing | Full |
|
Customers on the Enterprise GC tier choose a chat model across Google, Anthropic and OpenAI, and the choice is theirs to control. It applies to chat only, not to scope generation, takeoff or issue discovery. Struction names no model policy by workspace or action and no fallback visibility. Sourcestruction.com/pricing Enterprise GC tier; readread 2026-09-15 |
|
| APIs / SDKs / MCP Extensibility | Not documented |
|
The product is UI only. Struction names no API, SDK, webhook or MCP surface, and the only outbound path is export to Excel. Sourcestruction.com/docs, /docs/scope-exporting (title in index), probes; readread 2026-09-15 |
|
| Testing, Debugging & Optimization | Not documented |
|
The Verify step marks individual scope items as confirmed. Struction calls its accuracy high fidelity and says it only ships automated takeoff customers can rely on, which is its own internal release standard. Struction names no way to evaluate a change to the agent before or after release, no testing with fixtures or datasets and no scoring of output quality over time. Sourcestruction.com/docs/scope-editing, /faq, /landing; readread 2026-09-15 |
|
| Browser / Computer-use | Not documented |
|
Struction reads uploaded drawings, specs and BIM models with computer vision, detects symbols and acts inside its own workspace. It names no browser or computer control. A person can drive the BIM viewer with a game controller. Sourcestruction.com/docs, /landing; readread 2026-09-15 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
$50 per month (Struction plan, billed monthly); $500 per user per month for GCs
Not published; presumed direct enterprise pricing
Cost watchouts
Pricing model unknown; construction estimating tools sometimes bill per project, which can add up across many bids.
Variable cost rationale
Struction's pricing model is not published, so variable exposure is uncertain; construction estimating tools are commonly priced per seat or per project, either of which could scale with usage.
Sales call required
Mixed (some tiers require a call)
Free / trial
One project free with a work email, no credit card required
Lowest paid plan
Not published
Key ambiguities
The $50 plan omits scope-of-work generation and RFI, risk and scope-gap discovery, which start on the $500 per user GC plan, so the entry price does not buy the full product. Enterprise is usage-based with no published rate.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Adopt AI — AI CPA firm whose agents run reconciliations, close, AP and AR, and…
- Aera Technology — Decision intelligence platform where an always on agent executes…
- Akro AI — On premise operational intelligence that automates document heavy…
- alfred_ — AI executive assistant that triages the inbox overnight and scores…
- Alloy.ai — Commerce intelligence system for consumer brands that unifies four…
Alternatives to Struction
The closest documented capability profiles to Struction among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- alfred_5.5 / 14Fuller documented coverage on Triggers & Channel Coverage
- Archimetis5.5 / 14Adds documented Security, Identity & Governance
- Balerion AI4.5 / 14Adds documented Security, Identity & Governance
- Planned6.5 / 14Adds documented Security, Identity & Governance and Testing, Debugging & Optimization
- Shortwave5.5 / 14Adds documented APIs, SDKs & MCP Extensibility
- Smart Bricks3.5 / 14A lighter documented profile than Struction
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded