Google launched Gemini Enterprise for Legal this week, and Richard Tromans at Artificial Lawyer named the wider pattern: Legal tech and Labs fighting to become the center of legal work. Google threads a long list of legal tech partners,DMS providers, legal data groups, productivity platforms ,into GEL so lawyers pass through Google to reach their other tools. Microsoft ships a legal agent inside the software most of the legal market already runs on. Harvey and Legora keep announcing point-solution partnerships for the same reason. Kirkland builds internal applications to keep its know-how at its own center. The prize, in Tromans's framing, isn't any single feature. It's becoming the home of legal labour, with the long-term loyalty and locked-in revenue that follow.
artificiallawyer
The governance stakes are bigger.
A platform that becomes the center of legal work doesn't just capture attention and renewal revenue. It captures context. Deal histories. Negotiation positions. The playbooks and precedent that encode a legal team's accumulated judgment. And once agents are doing the work, the platform also holds the record of every action those agents took on a lawyer's behalf, what was read, what was drafted, what was sent, and under whose authority.
Whoever owns the workspace owns the evidentiary record of how legal work happened.
When I led counsel on Box's infrastructure migration to Google Cloud, the lesson that stuck was that infrastructure choices are governance choices. Where work lives determines who can see it, who can audit it, and what you can produce when someone with subpoena power asks. Nobody in that migration treated the platform question as a procurement decision. It went to the board, and it deserved to.
So the questions that matter in platform selection look less like a feature comparison and more like architecture diligence. Where does privilege sit when a query passes from your workspace through a partner tool and back? Who holds the audit trail of agent actions, in what format, and can you export it without the vendor's cooperation? What happens to your prompts, playbooks, and precedent , the know-how layer, if you leave? Kirkland's answer is to build internally and keep the center to itself. Most legal departments can't do that, which means the exit rights and record portability terms in the platform agreement are doing the work that Kirkland's engineering budget does.
There's a version of this story where centrality is just lock-in with better branding. I think the more useful read is that centrality is a trust proposition, and vendors should be forced to compete on it. A legal AI platform that wants to be the center of legal work is asking to hold the most sensitive record a company generates about its own risk. The vendors who win the long game will be the ones who treat that as a fiduciary-grade responsibility rather than a retention metric.
The practical move is to stop evaluating these platforms as tools and start evaluating them as jurisdictions. You're choosing where your legal work will be domiciled — which laws of data access, retention, and portability it will live under, and how hard it will be to emigrate. Centrality means the switching costs compound monthly as your context accumulates. Negotiate the exit before the entrance: data export in usable formats, agent action logs you control, deletion you can verify, and no partner data-sharing beyond what you've mapped.
The battle for centrality will produce a handful of winners, and Tromans is right that it won't be winner-takes-all. But the buyers get a vote. Legal teams that price trust into the selection — that treat the record as the asset and the workspace as its custodian — will pull the market toward platforms worth centralizing on. That's a better outcome than discovering, three years in, that your department's institutional memory lives in someone else's building.