FIRMUpgrade
Book an Assessment
AI & LLM Adoption

Business AI for law firms — configured around confidentiality

Claude, ChatGPT, Gemini, Copilot, and agentic tools can change how your firm works — for internal work and client-facing work alike. Set up carelessly, they can also put client confidentiality at risk. We configure AI tools around your firm's confidentiality requirements, with documented access, training and retention settings and workflows for attorney review.

Why Configuration Is Everything

A personal account and a configured business plan are very different starting points

The differences sit in things most firms never look at: who can access inputs, retention, training use, which features send data elsewhere, and who controls the workspace. Whether a particular use is appropriate for particular client information is a judgment for your attorneys — it should be made on the actual agreement and settings, not the brand name.

We read the terms for the exact product and features you'll use, configure what they support, and give you a dated record of what we verified.

  • Training use — plans and settings where business content is excluded from model training, verified in the tenant
  • Retention — set to your records policy; zero-data-retention only where it is actually offered for your features
  • Feature paths — web search, connectors, file stores and agents reviewed, because they can change where data goes
  • Workspace governance — role-based access, audit logs, and offboarding that actually offboards
Closed green leather portfolio with a brass clasp on a walnut desk beside a fountain pen
The Toolkit

Platforms we deploy

Anthropic Claude

Claude Team & Enterprise

Strong research and drafting workflows, a growing catalog of legal connectors (including Midpage), and admin controls on Team and Enterprise plans. Business plans exclude your data from training by default; retention and connector settings are configured per firm, and API zero-retention covers only eligible features.

OpenAI

ChatGPT Business & Enterprise

The tool most of your staff already knows — which is exactly why it needs governance. Business and Enterprise plans add a DPA, no-training defaults, and retention controls; zero-data-retention and a HIPAA BAA are separate arrangements with their own eligibility. We move personal-account use into a firm-managed workspace.

Google Gemini for Workspace

For firms already living in Gmail, Drive, and Docs. With a qualifying Workspace edition, Google states your content isn't used to train models outside your domain without permission — but consumer Gemini, third-party apps used with Gemini, and Gemini Enterprise each have their own terms. We make sure you're buying the one you think you are.

Microsoft Copilot

Microsoft 365 Copilot & Agentic Tools

Copilot inside Word, Outlook, and Teams for firms living in Microsoft 365, under Enterprise Data Protection and Microsoft's DPA for prompts and responses. Generated web-search queries go to Bing, where Microsoft says its DPA doesn't apply, so we set the web-search policy deliberately. Agentic tools for back-office automation are adopted the same way.

A Third Option

Some work is better kept on your own hardware. We can set that up too.

A well-configured business plan fits a lot of day-to-day work. For particular matters — regulated clients, client terms that limit outside processing — we can deploy open-weight models on hardware your firm controls, and review the rest of that setup's data path just as carefully.

Explore Private & Local AI Vendor comparison
Solo attorney's office with a workstation and a compact AI server unit beside a brass banker's lamp
The Deployment

From shadow IT to governed advantage

Odds are your staff is already using AI — in Clio's 2025 Legal Trends Report, 79% of legal professionals surveyed said they do. The question is whether it's happening under your policy or outside it.

Audit

What's already in use (you may be surprised), what data is flowing where, and what your malpractice and ethics exposure looks like today.

Select & Configure

The platform mix that fits, chosen after a bounded pilot; the applicable agreement confirmed; the tenant configured — retention, training, access control, audit logging, feature controls — and every setting recorded with a date.

Policy

A written AI policy your attorneys approve — which tools are approved for which information, who reviews AI-assisted work, and how verification, correction and disclosure happen. For California firms, aligned with SB 574.

Train & Iterate

Hands-on sessions using your real workflows (sanitized), a prompt library for your practice areas, and quarterly refreshers as the tools change.

What AI Actually Does For Firms

Practical wins, not science fiction

Drafting & Review

First drafts of correspondence, memos, discovery responses, and contract summaries in minutes — with attorney review built into the workflow.

Document Intelligence

Summarize depositions, extract key dates and obligations from contracts, and query thousands of pages in plain English.

Admin Reduction

Intake triage, meeting notes, time-entry narratives, and client status updates — the unbillable hours that eat your day, automated.

We don't quote industry savings figures as a promise. We measure a baseline on one or two of your workflows, run a bounded pilot, and report what changed. The firms that get value aren't the ones with the most tools; they're the ones with clear rules and training. That's the part we build — starting with the fixed-fee assessment, and extending into AI-assisted legal research where your practice calls for it.

Common Questions

Is AI safe for a law firm? Your questions, answered.

Will using AI put our attorney-client privilege at risk?
It can, and no deployment model removes that question. Privilege and confidentiality depend on the facts, the agreements, who can access the information, and how the work was directed — matters for your attorneys' judgment. What we do is the technical side: identify exactly where inputs go for each tool and feature, configure and document the access, training, and retention settings the applicable agreement supports, and build attorney-review steps into the workflow. We won't tell you a setting guarantees privilege.
Our staff already uses ChatGPT on personal accounts. Is that a problem?
It's a common finding — in Clio's 2025 Legal Trends Report, 79% of legal professionals surveyed said they use AI in their work, and some of that happens on accounts the firm doesn't control. The fix usually isn't a ban; it's an approved alternative people can actually use. We move that work into a firm-managed workspace with reviewed settings, then train everyone on which information may go where.
Which AI platform should our firm choose — Claude, ChatGPT, Gemini, or Copilot?
It depends on how your firm works and on the exact plan and features you'd use. Claude has strong research and drafting workflows and legal connectors; ChatGPT benefits from staff familiarity; Gemini fits firms already on Google Workspace; Copilot fits firms living in Microsoft 365. Many firms end up with a primary assistant plus the one built into their office suite, and some add a locally hosted model for particular matters. We compare options against your workflows in a bounded pilot before you commit.
What is a DPA, and does it settle confidentiality?
A data processing addendum is a contract term describing how a vendor processes your data — often covering processing purposes, subprocessors, security, and deletion. It may be built into the main agreement rather than signed separately, and it rarely answers every question on its own. "No training on your data," zero data retention, a DPA, a HIPAA business associate agreement, and your duty of confidentiality are different things: a plan can exclude training but still retain data; zero-retention options often cover only some endpoints or features; web search, connectors, file storage, and agents can change where data goes. We record the agreement, the settings, and the features in use, with the date we checked them.
Do we really need a written AI policy?
Usually, yes. In Clio's 2025 Legal Trends Report, 53% of legal professionals said their firm has no AI policy or that they are unaware of one. A good policy says which tools are approved for which information, who reviews AI-assisted work, and how verification, correction, and disclosure happen. For California attorneys, SB 574 — effective January 1, 2027 — makes several of those steps statutory, so we align the policy with the enacted text. Your attorneys approve it.

Adopt AI deliberately, with the settings written down.

Book an AI readiness assessment: current-use audit, platform recommendation, policy, and training plan.

Book Your Assessment Explore Services