AI for law firms, built around your duties.
Confidentiality, supervision, privilege: the constraints come first, the automation second. We build law firms their own Claude plugin, automate the routine around the practice, and train your team to use AI inside your professional obligations.
Most law firms are small businesses with heavier duties. You run on practice-management and document systems like Clio and NetDocuments, your client data is privileged, and your supervisory responsibilities are not optional. The AI wave has already arrived at your firm anyway — staff pasting matter details into chatbots nobody approved — while the tools actually built for small businesses speak QuickBooks and Shopify, not the systems a practice runs on.
That is the gap we work in. We wrote a full post about why Anthropic’s small-business plugin skipped law firms and what the law-firm version looks like. The short version: one constraint governs everything we build for a firm — nothing leaves the firm without a lawyer’s OK.
What we would do at law firms
A plugin teaches Claude how your firm works: your matters, your templates, your routines. It answers new enquiries at any hour with a draft in your tone, hands the managing partner a Monday brief of what moved, and drafts routine letters from your own precedents — read-only to start, every draft stopping at a lawyer’s desk.
Deadline watching with nudges to the right person — date arithmetic done by ordinary software, never the model, so it can be trusted. Time entries turned into billing narratives clients understand. Unpaid bills chased politely and persistently, on your schedule, so no one has to make that call again.
Hands-on training for lawyers and staff using your own templates and precedents as the curriculum: when AI helps, how to review a draft, and when not to trust it. Your team leaves with written playbooks that keep day-to-day AI use inside your confidentiality and supervision duties.
The client-confidentiality risk, found and fixed
Client: Professional services firm, 400 employees
Service: AI Risk & Security Assessment
Duration: 4 weeks
The engagement closest to law-firm work: a 400-person professional services firm — the same confidentiality duties over client data, the same audit pressure. Leadership believed AI use was limited to an approved pilot. Our assessment found sixty AI tools in use that neither the policy nor procurement knew about, three of them actively sending client data to unapproved offshore services.
All three were stopped within a week. The firm walked into its audit with a complete AI inventory, a risk register, and a policy teams could actually follow — and kept the productivity gains, inside the guardrails. Law firms carry stricter duties and the same blind spot; finding it before a privilege problem does is cheaper.
“We thought we were asking for a compliance document. What we got was a mirror — and we are much safer for having looked into it.”
Questions law firms ask us
Does this put client data at risk?
+
The system sees only the tools and folders you connect it to — one document library, if that is all you open — and starts read-only: it can look, never send. Anthropic states it never trains on your business data, on independently audited systems (SOC 2 Type II). The honest caveat: the larger risk today is usually the AI use already happening at your firm without any of this. The safety check finds that first.
Will AI send anything without a lawyer’s approval?
+
No. Nothing leaves the firm without a lawyer’s OK — that is how we build it, and it is what your professional duties around supervision and confidentiality already require. Sending, filing, anything touching client funds: a human signs for it, always.
We run Clio and NetDocuments. Can this actually connect?
+
Yes — that is precisely the gap. Anthropic’s own small-business plugin already reaches tools it does not officially support through read-only links that let Claude look but never change or send anything. Your plugin starts the same way: a read-only link to your practice systems, proving itself with summaries and drafts before it earns anything more.
Start with the safety check, not the software
Every law-firm engagement opens the same way: what AI may and may not see, so client confidentiality is airtight — most firms also discover staff are already using tools nobody approved. Thirty minutes, confidential, no jargon.