Roughly one in five building permit applications arriving at Toronto’s counter is missing basic paperwork. On August 31 the city put an AI screen in front of the queue to catch that before an application is ever submitted.
The tool is called Building Permit Application Pre-Check. It lives at toronto.ca/PreCheck and runs on CivCheck, the Guided AI Plan Review platform from Vancouver-based Clariti. Residential projects only for now, and using it is voluntary.
What the AI plan review actually checks
Pre-Check screens for missing documents, incomplete information, and in-scope zoning and Ontario Building Code issues. Every finding comes back with a citation to the underlying requirement, which matters more than it sounds. An unsourced flag is a support ticket; a cited flag is an instruction. The governance language in Clariti’s release is blunt: “CivCheck does not approve, refuse, or delay any permit.” City staff make every decision and can validate or override a finding during normal review.
That framing is the vendor’s and the city’s description of the workflow, not an audited constraint. But it’s the right design question, and it’s the one that sank earlier municipal AI pilots that blurred the line between screening and deciding.
The Honolulu numbers, and who counted them
Clariti’s pitch rests on Honolulu. Its Department of Planning and Permitting deployed CivCheck, and in the first quarter of 2026 applicants using it saved more than 40 days per permit on average, while residential permits reached decisions 55% faster than non-CivCheck applications. Read those figures with the source in mind. They’re Clariti’s analysis of its own customer, not an independent audit or a DPP publication.
Denver recently selected the same product. Three cities on one engine starts to look like a procurement question for every mid-size building department heading into 2027, rather than a pilot curiosity. Clariti also launched AI Studio, which gives government teams direct access to its permitting implementation staff, and was named to the Center for Public Sector AI’s “AI 50.”
Permit throughput is the constraint money can’t move
A contractor can pay for more labor, more equipment, a longer crane rental. It cannot pay for a faster plan reviewer. Permit review sits on the critical path of most housing work and is one of the few schedule risks that cash doesn’t touch, which is why cities have been attacking it from the staffing side for a decade with modest results.
Toronto flips the target. If the Honolulu result holds at Toronto’s volume, the binding constraint was never reviewer throughput. It was application quality, and quality is fixable upstream at close to zero marginal cost. That’s the reading worth testing against deep-affordability projects where entitlement and permitting timelines eat the pro forma outright, like the 320-unit Arverne East Building D now under construction in Far Rockaway.
The pilot aligns with Toronto’s Digital Infrastructure Strategic Framework and sits against an Ontario target of 285,000 new homes by 2031 plus Toronto’s own HousingTO goal of 65,000 rent-controlled homes by 2030. Those are policy numbers, not forecasts. Whether an AI pre-check moves them at all is the open question, and nobody has a year of data yet. Clariti’s announcement is here.