TRACE AI / THE EVIDENCE BEHIND THE ANSWER

Construction AI that shows its work.

Ask a project question. Inspect the source. Prepare the next step with evidence your team can review.

TRACE is the platform. Models are supporting services.

TRACE is an independently operated construction-management platform, not an OpenAI, Google Gemini or Anthropic Claude product. It can use configured external AI services without presenting those providers’ models as TRACE’s own foundation model.

The distinction is in the software around the model: authorized project context, evidence retrieval and citations, workflow state, access controls, review routing and attributable history. These controls connect an AI-assisted response to the project record and the people responsible for deciding what happens next.

  • Context: project and source scope determine what records can support a response.
  • Evidence: supported findings link to their sources; missing evidence and uncertainty remain visible.
  • Authority: AI can prepare and explain. Authorized people retain approval responsibility.
  • Workflow: documents, RFIs, submittals, field records and commercial controls have their own project workflows.

Capabilities depend on the module, account permissions, source readiness and configured providers. This describes TRACE’s application and control architecture, not a claim that TRACE trains or owns the external foundation models.

Read the documented evidence boundaries ↗ · See actual TRACE screens ↗

Start with the right project record.

TRACE brings project-context assistance to authorized documents and supported structured records. Choose the current project, eligible general references, or a combined scope. Source permissions apply before evidence reaches the reasoning model.

Questions can lead to quick lookup, an answer with sources or detailed investigation. TRACE separates recorded facts, assumptions, missing information and supported implications so an answer can be checked.

Follow the citation back to its source.

Supported digital files and scanned PDFs enter through document extraction and OCR paths. Page context, source identity and revision links help reviewers inspect the original material rather than rely on a detached summary. Supported proof tools collect evidence and prepare PDF exports.

  1. Select the authorized project and source scope.
  2. Ask a focused question and review the answer’s qualifications.
  3. Open the cited passage or original document.
  4. Prepare a reviewable draft or supported proof export.

Prepare work across the project.

Supported assistance includes RFI follow-up drafts, daily narrative improvement, submittal questions, change-review preparation, risk interpretation and evidence-gap investigation. Drafts remain editable; they do not issue themselves.

Knowledge with source context.

The documented reference expansion contains 24 original workflow guides and eight retained federal publications totaling 524 PDF pages. Retained editions and applicability matter: a reference is not automatically a governing contract requirement.

Availability and boundaries. AI requires an enabled module, configured provider, usable evidence, current permissions and usage allowance. OCR does not establish design intent or measured geometry. Formal engineering, safety, legal and approval decisions remain with authorized people.

Try the evidence path.

The homepage’s synthetic proof demonstration shows a cited answer and a missing-evidence response without using customer project information.

Ask for the source you need.

Source-specific research respects explicit official-source requests and exclusions of project records, while retaining private-identifier protections. Supported proof downloads preserve text, tables, citations and eligible source-page images.

Keep the conversation moving.

The September release improves telephone turn-taking and caller interruption. Product guidance covers supported workflows, App Factory, training and billing boundaries. Get help now offers typed guidance, current workflow guides and a human-ticket handoff path.

Research results depend on available sources and service readiness. Telephone acceptance describes a reported real-call result; ticket handoff does not imply continuous live staffing.

Frequently asked questions

Can TRACE search another customer’s records?

Project retrieval is scoped to the authorized tenant and project. General references and project-private evidence remain distinct.

Does an AI answer approve the work?

No. An answer or prepared draft supports review. Approval, certification and contractual actions require the authorized person.

See how TRACE fits your team.

Explore the supported workflows, sources and account configuration in a focused demonstration. Book a working session or explore TRACE plans.