Last updated: 2026-10-26
Structured AI, a Y Combinator Fall 2025 company, announced a $4.2M seed round on 2026-06-11. It reviews architectural and engineering drawing sets with AI agents, checking PDFs and Revit models against codes and a firm's standards. Each finding cites a sheet and location, and a person accepts or dismisses it.
About Structured AI
Structured AI is a quality assurance platform for architecture, engineering and construction firms. It reads a drawing set the way a senior reviewer does, checks it against building codes, specifications, schedules and the firm's own standards, and returns a list of findings. The company is a Y Combinator Fall 2025 startup, co-founded by CEO Raymond Zhao with former practitioners and computer vision researchers, and announced a $4.2M seed round on 2026-06-11 led by FCVC, bringing its total funding to $5M.
The core product is QA/QC and code compliance review. Each finding names the exact sheet, the spot on the page, the problem and the suggested fix, and the vendor says the output is a deterministic list rather than a confidence score. Beside it sit drawing set chat, which answers questions with a link to the source sheet, and an overlay view that stacks architectural, structural and MEP sheets covering the same area so cross-discipline clashes show up. A Revit add-in lets the agent query the live model and report confirmed failures and items needing a human look, tagged with element IDs and levels. Bluebeam markups carry findings back as redlines.
Teams that have already adopted general assistants such as ChatGPT, Claude or Microsoft Copilot can paste a few sheets into a chat, but those do not run a maintained library of checks across a 1,000-page package. Structured AI sits closer to vertical review products such as Legora and Harvey, which do the same job for contracts, and to document parsing infrastructure like Reducto. The difference is that its checks are written by people who have stamped drawings. Firms already on Google Workspace may try Gemini for ad hoc questions, and teams that research building codes online sometimes use Perplexity for the citations. For firms that mostly need to find answers across a company wiki rather than review drawings, Glean is the nearer fit.
Pricing is not published. Evaluation starts with a vendor-run pilot on one of your own projects, and results are promised within a week. Integrations listed on its site include Autodesk Revit, Autodesk Construction Cloud, Procore, Bluebeam, SharePoint, Google Drive, Outlook, Gmail, Word and Excel. Security claims are limited to two points: customer data never trains the models, and every organisation, pilots included, runs in an isolated environment. No SOC 2 or ISO report is listed on the security page. Software teams will find the equivalent job for code in our roundup of the best AI code review tools, and the best agentic AI list shows where review agents sit among the rest. Not sure whether this or a general assistant suits your workload? Smart Match can narrow it down.
Pricing
No public price list. Access is through a demo and a pilot on a live project; the vendor quotes after that. For ad hoc questions only, ChatGPT or Claude are quicker to try, since they need no sales call.
Key Features
- QA/QC and code compliance: Each check returns the exact sheet, the spot on that sheet, what is wrong and how to fix it, drawn from a library the vendor puts at 120+ checks.
- Drawing set chat: Questions about a set come back with a link to the exact sheet and location, after the system has read the symbols, tables and code references on every sheet.
- Overlay and clash detection: Sheets covering the same area are found and aligned automatically across architectural, structural and MEP, even when drawn at different scales.
- Revit add-in: The agent queries the live Revit model and reports definite fails and items to verify with element IDs and levels, with no export step.
- Custom checks in plain English: A firm describes its own standard in words and the platform runs it across every sheet of a set.
- Bluebeam and ACC workflows: Findings arrive as Bluebeam redlines, and sets can be read in place from Autodesk Construction Cloud, Procore, SharePoint or Google Drive.
Pros
- Every finding is cited to a sheet and location, so an engineer can confirm or dismiss it in seconds, which the vendor frames as its main trust feature.
- Works inside Revit and Bluebeam, the two tools many design firms already use, instead of asking reviewers to move to a separate report.
- The vendor says 9 in 10 findings were confirmed by client engineers in its own tracking, and a published case study cites 400+ issues found in one 1,000-page set.
- Customer drawings are kept out of model training and each organisation is isolated, pilots included, per the vendor security page.
Cons
- No public pricing or self-serve signup, so every evaluation begins with a sales conversation.
- The security page lists no third-party attestation such as SOC 2 or ISO 27001, which procurement teams at large firms usually ask for.
- Early-stage company with a $5M total raise and a 10-person team per its YC profile, so roadmap and support depth are less proven than older AEC software.
- Results quoted for accuracy and speed come from the vendor and its design partners, and no independent benchmark of the checks has been published.
Data Handling
- Training-data policy
- Customer project data is never part of model training; core models are trained on curated public datasets and construction standards (vendor security page).
Frequently Asked Questions
What does Structured AI cost?
The vendor does not publish a price list. Access begins with a vendor-run trial on your own drawings, and the quote follows that. Ask for the scope of the pilot in writing, since the vendor says results arrive within a week.
Can I try Structured AI before buying?
There is no self-serve free plan on the site. The route in is a pilot where the vendor runs its review on your own drawing set, inside an isolated environment. Bring a project that has already been through manual review so you can compare what each finds.
What can I use instead of Structured AI?
A general assistant such as ChatGPT or Claude handles a handful of sheets for one-off questions. Legora and Harvey do comparable review work for legal documents, and Reducto suits teams who want to build their own pipeline over parsed files. None of them ships a drawing check library.
Does Structured AI work with Revit and Bluebeam?
Yes, the site lists an Autodesk Revit add-in and Bluebeam markup support, plus Autodesk Construction Cloud, Procore, SharePoint and Google Drive as sources. Results land as redlines in Bluebeam or as element-level findings in Revit. AutoCAD and Navisworks are not on the integration list.
Is my drawing data used to train the models?
The vendor security page says project data never enters training and that each organisation runs in its own silo, demos and pilots included. It lists no SOC 2 or ISO 27001 report, so ask for a security questionnaire before sharing client drawings.
Top Alternatives
- Legora: Pick Structured AI for drawing sets and Revit models; pick Legora when the documents under review are contracts and legal text.
- Harvey: Pick Structured AI for construction QA/QC; pick Harvey for legal research and drafting work.
- Reducto: Pick Reducto if you want to build your own pipeline on parsed documents; pick Structured AI for a finished review product.
- ChatGPT: Pick ChatGPT for quick ad hoc questions on a few sheets; pick Structured AI for repeatable checks across a full set.