AI adoption in law is moving fast — but so is the caution around it
Legal AI adoption has grown sharply, though the exact numbers vary widely depending on firm size and survey: one 2025 report found usage among mid-sized firms surging from 19% to 93% year over year, while a separate 2026 American Bar Association report put overall adoption closer to a third of firms as of the prior year. What's consistent across nearly every report is the theme: confidentiality, not capability, is the real barrier. As Law.com put it in a September 2026 piece, "clients increasingly view security and confidentiality as prerequisites for AI adoption, rather than add-ons that give firms a competitive edge."
Here's how firms are actually navigating that tension in practice.
1. Keeping privileged research entirely in-house
Rather than running case research through a general-purpose AI tool that sends data to an outside company's servers, some firms are building AI systems trained specifically on their own case history and precedent — run on their own infrastructure, so the query and the answer never leave the firm's control.
2. Using air-gapped systems for the most sensitive matters
For firms handling highly sensitive or privileged case types, a fully air-gapped AI system — one that runs entirely on local hardware with no internet connection — removes the internet as a possible path for that data to travel through at all. This is typically reserved for the specific matters that actually warrant it, not applied firmwide by default.
3. Automating intake and scheduling, not case strategy
Lower-risk administrative work — client intake forms, scheduling, routine status updates — is a common, safe starting point for AI automation, since it doesn't touch privileged case substance the way research or drafting does.
4. Using AI for first-draft document review, with a human final pass
AI-assisted drafting and document review can meaningfully speed up routine work — contracts, discovery materials — without removing the attorney's judgment from the final version. The efficiency gain comes from cutting the time on the first draft, not from removing oversight.
5. Choosing integration over replacement
Rather than adopting a disconnected AI tool, many firms are integrating AI directly with the practice management or case management software they already use — keeping data inside systems they've already vetted for security, instead of introducing a new outside tool into the mix.
The honest bottom line
Every one of these approaches still comes down to one decision: does this specific use case need a connected tool, or does it need to be fully air-gapped? That's not a question with a universal answer — it depends on the matter, the client, and the firm's specific risk tolerance.
Weighing whether your firm needs a connected AI tool or something fully air-gapped? Book a free call and we'll walk through it honestly.
