Relativity Bought Gavel. Drafting Was the Easy Half

The legal AI market spent 2026 consolidating, and the most telling deal of the month was not the biggest. On June 12, Relativity, the eDiscovery and legal data company, acquired Gavel, a document-automation and drafting company whose tools are used by firms across 28 countries. The strategic logic is clean, and it points at where the value in legal software is being contested, which is not where most real estate teams actually feel pain.

What Relativity actually bought

Gavel automates the production of legal documents. It combines generative AI with rules-based templates so a firm can turn structured inputs into drafted, edited, and redlined work product, and it does this inside Microsoft Word, where lawyers already live. Relativity's plan is to connect that drafting surface back to the matter, so that a document opened, edited, and finalized in Word stays synced to the underlying data in RelativityOne.

In plain terms, Relativity owned the data behind a matter and wanted the place where lawyers write. Gavel was the place where lawyers write. The deal extends a large platform from the discovery and analysis end of a file all the way into the drafting end. For litigation and large-firm transactional work, that is a real expansion.

The drafting layer is getting rolled up

Step back and the pattern is unmistakable. The drafting and document-automation layer is being absorbed into bigger platforms, as Legal IT Insider noted when the deal landed. Owning the editor is strategically attractive because it is the surface lawyers touch most, and whoever owns the surface owns the workflow and, increasingly, the data exhaust.

This is the same instinct that has Microsoft pushing Copilot into Word and the legal-specific players racing to plant a flag in the same document. The bet is that document creation is the center of gravity, and that controlling it lets a platform pull in everything around it. For a wide band of legal work, the bet is correct.

Why drafting was the tractable half

Drafting got automated first because it is the part of the job that starts from structure. A template, a clause library, a set of merge fields, and a model that can phrase variations: those are well-defined inputs producing a well-defined output. The hard cases exist, but the median document is generated from things the system already knows.

Real estate transactional work has a drafting component that fits this mold. Closing letters, requisition letters, and standard undertakings are template-shaped, and automating them is genuinely useful. The problem is that drafting is the second half of the job. The first half is reading.

The intake problem nobody acquired

A title review file does not arrive as structured data. It arrives as a title package, a stack of registered instruments, a survey, a tax certificate, and a chain of emails, often as scanned PDFs of varying quality. Before anyone drafts a single letter, someone has to read that material, identify every encumbrance, figure out which ones survive closing and which get discharged, and reconcile what the documents say against what the file needs. That extraction step is messy, jurisdiction-specific, and unforgiving, and it is the step a generic drafting tool does not touch.

This is the gap. The acquisition extends a platform deeper into generation, the part that was already the most automatable. None of it makes the inbound title package any easier to read. A drafting engine that produces a beautiful requisition letter is still waiting on a human to have read the register and decided what the requisitions are.

Where a vertical tool still earns its keep

This is the case for building narrow and deep instead of wide and shallow. Cleardeal is title-review software for real estate legal teams, and it lives on the side of the problem the platform deals do not solve. It pulls requests directly from a Microsoft 365 inbox, runs vision-based extraction on the title PDFs to pull encumbrances and registered interests off the page, and only then generates the letters and sends them back through the Microsoft Graph. The drafting at the end is the easy part, and it works precisely because the reading at the front was done by a system built for title documents specifically, not a general-purpose drafter pointed at a hard input.

A horizontal platform can credibly say it drafts in Word. It is much harder to say it reads an Alberta title and a Quebec index of immovables with equal reliability, because that knowledge is not in a clause library. It is in the document types, the registry conventions, and the failure modes of a specific kind of file. That depth is the moat a vertical tool keeps even as the platforms swallow the drafting surface.

What to watch as the platforms move in

The consolidation is not going to stop, and the platforms will keep buying their way toward the editor. The question for any focused tool is whether the hard, unglamorous part of its workflow is something a horizontal player can replicate by acquisition, or whether it requires understanding a document type deeply enough that no general drafter will bother. For title and real estate diligence, the inbound file is that part. Watch whether the next wave of legal-AI deals reaches for intake and extraction the way this one reached for drafting. When a platform pays up for the company that reads the messy file rather than the one that writes the clean letter, that is the signal the real bottleneck has finally been priced.