Structured AI

Structured AI is a construction technology company that provides an AI-powered quality assurance and quality control (QA/QC) platform for architectural and engineering drawing packages. The platform works across both PDF drawing sets and native Revit models, reading every sheet of a set, applying the relevant building codes and a firm's own standards, and returning a list of findings that identifies what is wrong, where it is located, and how to fix it. Capabilities described by the company include compliance reviews across thousands of pages, natural-language querying of drawing documents with sourced answers, custom checks written in plain English, version comparison and pixel-aligned discipline overlays, and comment close-out, all with an audit trail attached to each finding.

The company positions its product as a replacement for manual "red pen" review at the end of a design deadline, arguing that errors missed in manual review resurface downstream as RFIs, change orders, and field rework. Its check library covers architectural, structural, MEP, and civil disciplines and, per the company's about page, contains more than 120 checks. Findings are cited to the exact sheet, detail, and code citation, and the system is framed as recommending rather than deciding: proposed changes must be accepted by a human engineer, a workflow the company and its design partner compare to spell-check inside the drawing software.

Structured AI describes itself as building "the AI workforce for construction design engineering," with a stated longer-term ambition to move from quality-controlling plans to generating and optimizing them.

Founding story

According to the company's Y Combinator profile, CTO Brandon Abreu Smith began coding at 12 and, after a video he made at 17 showing a fighter jet trained with deep reinforcement learning to dodge missiles circulated widely, was recruited by a US construction contractor to apply similar methods to MEP pipe-routing optimization; he spent roughly six years contracting and building AI tools for construction and engineering before leading innovation and scaling in-house AI tools at a top US engineering firm, then spinning out with his team to form Structured AI. CEO Raymond Zhao initially pursued architecture and engineering but switched to Maths & Statistics at Oxford after an internship exposed him to the manual work involved in junior architecture roles. The founders describe starting the company after having performed drawing reviews themselves at architecture and engineering firms.

Business model

Structured AI sells a software platform to architecture and engineering firms. The company's stated sales motion involves an intro call and live demo followed by a pilot on the customer's own drawing sets with checks configured to the firm's standards. Deployment options include uploading a drawing set, connecting the system where the set already lives, or running checks directly in the model, with integrations for Bluebeam Studio (two-way markup sync) and Revit (checks against model elements and writing approved fixes back to the source file).

Traction

The company reports that it reviews a full drawing set in about 20 minutes, that 9 in 10 findings triaged by client engineers are confirmed and fixed, and that 6x fewer issues reach the field. In one case study cited by the company, its design review agent caught more than 400 issues in a single 1,000-page set headed to the jobsite. Syska Hennessy Group, an MEP engineering firm, has been testing the platform and helping develop it; Syska also invested in the company as a seed-stage investment, per its executive. Total disclosed funding stands at $5 million following the June 2026 seed round.

Latest developments

On 10-11 June 2026 Structured AI announced a $4.2 million seed round led by FCVC, with participation from Y Combinator, 20VC, Cherry Ventures, Zero Prime Ventures, Transpose Platform and Sequoia Capital's Scout Fund, plus angels including Charlie Songhurst and founders of Supabase, Enpal, Pebble and Privy, taking total funding to $5 million. Proceeds are earmarked for further development of its optical recognition and computer vision models. The company is expanding integration into design-authoring software so recommendations can be accepted directly in native drawings, and is hiring across founding engineering, design engineering, go-to-market and operations roles in New York.