On this page
Construction Executive's August 5, 2026 Top Story, "A New Vision for AI: Construction Design-Review Platform Lends Extra Eye to Image Analysis," features Alexander Michalatos, CEO and co-founder of Buildcheck. The interview covers the platform's origins, the three-year model development process behind it, and where Alexander expects drawing-focused AI to go over the next decade.
What the article covers
Alexander traces the company's impetus to his own career — a construction family in Vancouver, where his father built single-family homes for decades, followed by roles across the owner, general contractor and design sides at Stantec, Honeywell, EllisDon and QuadReal. The pattern he describes recurring in every one of those roles: small inconsistencies scattered across hundreds of sheets that nobody caught until steel was going up or concrete was being poured.
His explanation for why AI stayed peripheral to construction risk for so long is the through-line of the piece: "Construction doesn't run on contracts or emails, it runs on drawings."
Don't conflate AI with LLMs
Asked for a closing thought, Alexander makes the argument the headline draws from. Language models are excellent at text — contracts, specs, RFIs — but construction's highest-value problems are visual: drawings, models, site conditions. Vendors addressing those problems are doing specialized computer vision work that, in his description, looks very different from a chatbot wrapper.
The test he offers buyers evaluating AI tools is a single question: what were the models actually trained on?
Why Buildcheck stays construction-only
The interview addresses focus directly. Drawings aren't standardized, symbol sets vary by firm and discipline, and layering conventions shift from project to project — so training a model that genuinely understands an MEP sheet versus a structural or civil sheet takes years of labeled data and domain expertise. Alexander calls that depth the moat, and says staying focused is the only way to build it.
He is equally direct about the build itself: three-plus years on proprietary models and on the reviewer experience around them, neither of which, in his words, "can be vibecoded." Adjacent drawing-heavy verticals — shipbuilding, aerospace manufacturing, industrial engineering — are acknowledged as possible later. The stated preference is to "be the best in the world at one thing than average at several."
Documented results
- $200 billion in estimated annual global overages attributable to design errors and coordination gaps
- 10x to 40x typical customer ROI under conservative assumptions
- Over $500,000 in cost avoidance and 27 days of schedule protected on a single 230-unit multifamily project
- Up to 64% of construction organizations now experimenting with AI, per recent industry surveys
Alexander identifies safety as the most underappreciated category of return. The highest-severity inconsistencies caught are life-safety issues — mismatched fire ratings between disciplines, undersized electrical feeds to fire pumps, missing sprinkler branches — and finding those in the documents, before installation, is meaningfully better than catching them at commissioning.
What the next decade looks like
Three expectations close the interview: automated review becomes table stakes on major projects, the way code check is today; tooling shifts from reactive flagging to real-time coordination feedback inside the design-authoring tools as drawings evolve; and genuine design optimization emerges, with AI proposing more efficient duct runs and structural layouts grounded in both code and constructability.
Underneath all three is a structural change — drawings ceasing to be static PDFs and becoming structured, machine-readable artifacts. Alexander argues that once that happens, everything downstream gets meaningfully faster: estimating, procurement, coordination and closeout.
Read the full interview at Construction Executive.
