In October I built a due-diligence report on a single property, entirely from public data. One location in, nine sections out: zoning, heritage, the regional plan, hazards, neighbours, census, businesses, transit, safety. The script makes about forty queries to the city’s, the region’s, the police service’s and the conservation authority’s map services, and finishes in thirty seconds.
For a planning consultant, that’s roughly three to six hours of reading ten-plus map viewers and PDFs per property, by my estimate. The speed isn’t the story, though. What the report admitted it didn’t know is.
The test site was a real file
I picked the test property blind: the first downtown zoning area carrying a “hold” symbol. It turned out to be a site with a live application for more than 1,300 homes in towers, and a 2024 amendment to the city’s plan allowing 1,622 units per hectare.
So the report had something to be checked against. It held up. The zoning, the hold, the plan amendment and the surroundings all matched the public record of the application.
The best page was the one that said “we don’t know”
Eight things the open data could not answer, and each became an item on the checklist the buyer takes to the city.
- What the hold actually requires. The zoning says there is one. The conditions are in a by-law published only as a PDF. The report can flag the hold and say that an application to remove it costs $4,775. It can’t list the conditions.
- The city’s own planning designations. They’re published as map images, not data.
- Which development-charge area the site sits in. Not published as data.
- Complaints about the block. Parking, property standards and noise complaints carry no location.
- Whether neighbouring rentals are licensed. The list has no download.
- Traffic on city roads. The counts cover regional roads only.
- Building outlines downtown. The layer returned nothing for a lot that clearly has a building on it.
- The application itself: towers, units, parking, what happens to the heritage building. That came from the city’s public notice and the trade press, not open data.
None of these is a failure. They’re the questions a professional would have to ask anyway. Writing them down turns a gap into the next step.
The layout that earned trust
Six colour-coded flags first: hold, watch or OK. Then the tables, with a source column on every row. Then a list of actions. Then a section on sources and limits that names what open data can’t say.
A reader decides in the first ten seconds whether to keep going, so the flags carry the verdict. The source column means anyone can check any number without calling me. And the limits section is the reason to believe the rest.
The line that only appeared because I looked wider
The single most valuable line in the actions list was about the neighbours, not the property. Three sites on the same block carried their own holds. Three holds within one block usually means a question about water and sewer capacity, and it only showed up because the report pulled zoning 150 metres out, not just at the address.
The same habit caught a quieter problem. The zoning area was bigger than the property (8,687 square metres against 6,034), because it also covered the lot next door. The report now gives both numbers and says which is which.
Check the comparison before you sell it
My first safety section compared incidents near the site to a city-wide average. Assaults within the ring came to 137 against 7.9 expected, and other categories ran 5 to 17 times the average.
Then I noticed the city-wide area includes farmland in the north. That makes the urban average look about a third lower than it is, which makes every downtown site look worse. The fix is to compare against the urban area only, before anyone pays for the ratio.
What this means for any report an AI tool helps you produce
Your board pack, your client proposal, the market summary someone generated on Tuesday. The tool will fill every section confidently. It won’t, unless you ask, tell you which sections rest on a guess, an old number or a source it couldn’t reach.
So ask for three things: a source on every figure, the verdict first, and a closing section on what the report couldn’t answer and who to ask. A report that lists its own gaps is the one people believe.