AI for Real Estate Due Diligence Beyond Title

Title Review Was the Easy Slice

Title review is where every AI tool in real estate started, and there is a clean reason for it. The document set is bounded: a register extract, a list of inscriptions, a set of discharges, the occasional notarial deed referenced by number. Each inscription has a registration number, a date, a right, and named parties. The output is a defensible letter that opines on a small list of encumbrances. Every vendor in the space picked this slice first for the same engineering reason. The corpus is structured, the rules for reading it are codified, and the lawyer's product is a list of facts plus a recommendation.

That tractability is also the slice's main limitation. A title package is the smallest piece of pre-closing diligence on a commercial deal, and often on a residential one too.

The Rest of the Pre-Closing File

On a typical commercial closing, the documents that arrive alongside title look something like this. A zoning compliance certificate from the municipality describing permitted uses and any non-conformities. A Phase I Environmental Site Assessment and, if the Phase I flags a Recognized Environmental Condition, a Phase II with sampling. A property condition report for buildings with mechanical or structural risk. A current survey, which in Quebec means a certificate of location, in Alberta a Real Property Report with municipal stamp, in BC a building location certificate. For condominium units, a disclosure package including the declaration, by-laws, recent minutes, the reserve fund study, and a status certificate. For shopping-centre or industrial deals, an REA or reciprocal easement agreement. Tax certificates, lien searches outside the land registry (PPSA, Bank Act, judgment registries), and operating leases.

Every one of these is a long PDF, structured but inconsistent across jurisdictions and authors, read by a lawyer or a clerk who pulls a small number of facts out, flags the unusual ones, and either opines on them or asks for a representation. From a model's perspective, this is the same shape of problem as title.

The Product Category Is Merging

What was missing in 2022 was the product layer. By the first half of 2026 it is real. For environmental, V7 Go and CaseMark generate or assist Phase I ESA drafts under ASTM E1527-21, and FlipType ships an AI-assisted screening report. For condominium documents the Canadian field is more developed than most lawyers realize: Eli Report and StrataReports pull risks out of board minutes, budgets, and reserve fund studies on BC, Alberta, and Ontario condos in under fifteen minutes, with Eli carrying errors-and-omissions insurance to handle the bad-summary risk. For zoning, Plotzy AI runs a use-by-use compliance check against municipal codes and produces a lender-ready report. For lease abstraction, V7 Go reports 99% accuracy with citation links back to the source page. For property condition, CAPE Analytics ships automated reports off photo and drone footage. For the closing folder itself, Closing Folders (an iManage product) already routes signature pages and assembles closing books, and is the obvious place for an extraction layer to slot in.

None of these tools removes the professional. The Phase I still gets signed by an environmental professional. The certificate of location still comes from a Quebec land surveyor. What they do is collapse the prep work that a clerk used to do over two days into a model output a reviewer can audit in twenty minutes.

In 2024, vendors organized themselves by document type: a title company, a zoning company, a Phase I company, a condo company. In 2026, those slices are starting to feel like internal modules of one product. The reviewer does not want four tabs. The reviewer wants one file view that says: here is the title package and the encumbrances, here is the zoning certificate and the non-conformities, here is the Phase I and the RECs, here is the condo declaration and the unusual restrictions, here is the survey and the encroachment.

How We Think About the Next Readers

We started Cleardeal because title review was the cleanest first slice. The encumbrances are itemized, the rules for opining on them are well known, and the output is a defensible letter. The harder question was always going to be where it goes next. A real estate file rarely stops at title. Zoning certificates, property condition reports, condo declarations, sometimes survey readings: every one of these arrives as a separate PDF and gets a separate read. The pattern that fits is the same. Extract the structured facts. Surface the unusual ones. Let the lawyer keep judgment. The product expands by adding readers, not by changing what the lawyer does.

Each new reader is a prompt, a JSON schema, a confidence calibration on real client files, and a UI for the side-by-side review. The hard parts are not the model. They are the source crop that grounds the extraction, the audit trail the bar wants when a notary signs, and the cost routing that keeps the file profitable.

What the 2027 Due Diligence Stack Looks Like

A lawyer in mid-2027 closing a mid-market commercial deal will work in one surface that holds the title package, the zoning certificate, the Phase I, the survey, the condo file, the lien searches, and the lease abstracts, with extractions on each and a single review queue that surfaces only the items needing a human call. Each output will be cited back to a pixel-level source crop. The signing professional will defend a list of facts the model produced, not generate it from a blank page.

The shops that win started with one well-bounded document type, got the audit trail and confidence layer right, then added readers. Title review software and due diligence software have been separate categories for a decade. The next eighteen months are when they stop being separate.