CoCivil — Land Development Due Diligence

Won Google Studio AI at Hack Canada 2026 — turns a plain-English query about an Ontario property into a complete planning submission package.

The problem

Due diligence for land development means weeks of zoning bylaw reading, setback math, and policy archaeology before you know whether a lot is even buildable. We built CoCivil at Hack Canada to compress that into a query.

What it does

Type "What can I build at 123 Queen West?" and the platform:

• Parses the query into an address + intent with an LLM agent
• Pulls parcel data from Toronto Open Data and evaluates it against Zoning By-law 569-2013 — designations, overlays, setbacks, height limits, lot coverage — with adjacency analysis and frontage detection for full site context
• Reasons over planning law with an AI assistant grounded in the Planning Act, the Provincial Planning Statement (2024), and Ontario Regulation 462/24 — proposing variance strategies and approval pathways with citations
• Generates 3D massing from plain English (Three.js), plus an interactive 2D floor-plan editor with real-time Ontario Building Code compliance checking
• Assembles the submission package — compliance matrices, planning rationales, shadow studies — auditable and flagged for professional review

What I learned

RAG quality is retrieval quality. The demo moment that won the judges over wasn't the generation — it was that every claim in the package linked back to the bylaw section it came from. Grounding is the feature; the LLM is the plumbing.

Built in 36 hours with a team. My part: the RAG policy search and the agent orchestration.