Compare Versions

Compare Versions is an AI-powered drawing revision review tool built for architecture, engineering, and construction teams who need to know exactly what changed between two versions of a project — quickly, accurately, and without manual page-by-page cross-checking. Whether you're moving from a 30% to a 65% submittal, reviewing a permit revision against the original construction documents, or auditing a prototype drawing set against a new location, Compare Versions produces a complete, structured record of every change across the set in minutes rather than days.

The tool ingests two PDF versions of the same drawing set and begins by intelligently pairing matching sheets across both documents. Sheet pairing is resilient to renumbering, reordering, and discipline reorganization, so an A-101 that became A1.01 in the new set still gets recognized as the same drawing. Sheets that exist in only one version are flagged as added or removed, while paired sheets are scanned for substantive content changes.

For each matched sheet pair, Compare Versions uses multi-pass ensemble verification to detect added, removed, and modified content across all major disciplines — mechanical, electrical, plumbing, fire protection, architectural, structural, and civil. Findings include changes to equipment tags, fixture counts, schedules, spec sections, code references, dimensions, notes, room programs, and drawing index entries. Every finding is tagged with its discipline, change category, severity (high, medium, low), and a confidence score, and is anchored to its exact location on the page with a precise bounding box so reviewers can click straight from the report to the change itself.

Results can be viewed by sheet pair, grouped by room and building level, filtered by discipline or severity, or summarized in a per-discipline breakdown that shows how many sheets were modified, unchanged, added, or removed in each trade. An executive summary is generated automatically, and the full revision log can be exported for transmittals, coordination meetings, owner reports, or QA/QC documentation.

Compare Versions is especially valuable for MEP coordination, multi-discipline submittal review, fast-moving renovation projects, and chain or prototype rollouts where the same drawing set iterates through many milestones — giving design teams, project managers, and reviewers a fast, auditable answer to the question "what actually changed?"

Structured is building the AI workforce for construction design engineering. Our platform automates QA/QC on engineering drawing sets, using cross-discipline vision-language agents. Every finding is returned deterministically with the exact page, location, rule, and fix. Our agents review drawing sets against code, spec, and firm standards in 30 minutes, solving the QA/QC bottleneck behind $177B in annual rework. Structured AI deploys vision-language agents that review every set against code, spec, and firm standards in 30 minutes, turning a weeks of monotonous work into a frictionless QA/QC gate.

The platform combines vision-language models with OCR to parse drawings and identify symbolic and image-based elements (MEP symbols, structural callouts, annotations), a multi-agent review system that runs cross-discipline checks in parallel, and internal fine-tuned LLMs for natural-language reasoning over codes, specs, and firm standards. A built-in chatbot copilot handles software navigation and onboarding for a frictionless user experience.

70% of construction rework starts with errors in design drawings. Structured AI deploys vision-language agents that review every set against code, spec, and firm standards in minutes through a frictionless QA/QC gate, solving a 177bn rework problem.

Our vision is an AECO industry where design errors no longer reach the field, where the tacit expertise engineers is translated to every project on earth. The documents that govern the built environment become a verified, defensible source of truth. The billion-dollar tax that rework imposes on construction goes to zero.

Our competitive advantage is years of domain expertise, compounded with elite commercial talent. Our CTO has spent nearly a decade building AI for the AEC industry, scouted at 18 to build innovation tools for Syska Hennessy Group. We've compounded this depth as we've scaled. Our engineering hires include software engineers with direct mechanical and electrical engineering experience, and architects. People who have stood on jobsites and reviewed drawings themselves. Our team understands the customer because, in many cases, our team has been the customer.

Structured AI is built on agentic AI throughout the stack: vision-language models with integrated OCR parse engineering drawings and recognise symbolic, image-based, and tabular elements; a multi-agent design-review system runs code, coordination, spec, and standards checks concurrently; proprietary fine-tuned LLMs reason over project documentation with engineering-grade accuracy; and a natural-language chatbot provides in-product navigation and contextual onboarding — so the entire experience, from review to interrogation, runs on emerging AI technology. We’d love to work with architecture and engineering firms across MEP, structural, civil, and architectural disciplines: the teams that author the drawing sets every building, road, and piece of infrastructure is built from.

We’d love to hear from design engineers, QA/QC managers, engineering directors, and principals: the people responsible for ensuring drawing sets are accurate, code-compliant, and aligned with firm standards before issuance. We serve the full size spectrum, from boutique multi-disciplinary studios to enterprise AE firms producing thousands of sheets per month. We also love to work with general contractors and owners who inherit these drawings downstream who feel the same pain (RFIs, change orders, rework) and benefit directly when the upstream QA layer is automated.