AI policy for law firms, starting with what is really happening
A policy written from the partners' meeting room misses the paralegal's browser tab. This survey finds out what is actually in use.
Many law firms that have an AI policy wrote it in one room. Most AI use in law firms happens somewhere else: a paralegal summarising a long affidavit, a solicitor tightening a letter of demand, a secretary turning a dictation into a file note. A good AI policy for law firms begins with an honest inventory of those moments. That is the gap this page is about, and it is a much smaller job to close once you can see it.
Calibre Forum is a short survey every person in the firm answers on their phone in about eight minutes, from partners to reception. You receive one branded report with an AI readiness assessment for your firm: a scored view across six areas, a register of the tools people really use, and a plain roadmap. It does not review any matter, and it never shows you individual answers.
Where AI shows up in law firms
Summaries of disclosure, briefs and long documents
Worth asking who feeds discovery bundles, expert reports or lengthy contracts into an AI tool to get a first summary, and where the document goes when they do. A summary produced from client material raises different questions from one produced from a public judgment. The point is not to catch anyone out, but to learn which tools are in play so that rules can be set around real habits.
Drafting letters, advices and submissions
Find out whether lawyers use AI to tighten correspondence, outline an advice or produce a first draft of submissions, and what, if anything, they check afterwards. Citations and case references are the specific area to ask about, because they must be verified by someone qualified. Verification habits vary widely between lawyers, so the report is useful for showing where supervision needs to be written down.
Precedents, clause banks and conveyancing packs
Practice management and document assembly systems increasingly include AI features. Ask who has them switched on, what the vendor does with the content, and whether your precedent library has been uploaded anywhere that was never reviewed. Content a firm spent years refining can leave the building through a convenient feature that nobody reviewed. Both are worth checking.
Transcription, file notes and time recording
Client calls, conferences and dictation are often captured by apps that sit outside the practice management system. Find out which recording and transcription tools are used, who consents, and where transcripts are stored. Time entry tools may also read emails and calendars. Recording a conversation raises questions about consent and storage that are easy to overlook when the tool is simply a phone app.
What puts law firms at risk
Confidentiality and legal professional privilege
Solicitors owe clients a duty of confidentiality, and privilege can be affected by careless handling of material. Pasting client matter content into a consumer tool, on a personal account, is the scenario most partners want to rule out. The survey shows whether it is something to rule out or already happening. Partners usually want this answered first, because it shapes every other rule in the policy.
Supervision, verification and court expectations
Courts and regulators in several states have issued guidance or practice notes on generative AI in proceedings, and professional conduct rules still make solicitors responsible for their work. Principals need to know where AI-assisted drafting enters the file, and how supervision of junior staff handles it. A policy that says every output must be verified is only useful if people know who does it and when.
Insurer and client due diligence
Professional indemnity proposal forms and cyber renewals can ask about AI use and controls. Corporate and government clients also add AI and data handling terms to panel arrangements. A controls checklist, and a register of tools, lets the firm answer from evidence. Being able to answer from a register and checklist saves time, and avoids promising controls that are not yet in place.
What each role's answers show
- Partners and principalsTheir answers on policy, insurance, risk appetite and client expectations show how aligned principals are with each other and with the people doing the work. It also shows how consistent the partner group is on what is acceptable.
- Senior associates and solicitorsDescribe drafting, research and review habits, and how matters are supervised. It shows where verification steps are strong and where they rely on habit.
- Paralegals and law clerksExplain the document-heavy tasks that fill their week, from bundles to chronologies. They tend to know which tools save time, and which ones nobody approved.
- Legal secretaries and practice supportDescribe dictation, formatting, diary management and filing routines. This reveals transcription and writing tools that never appear in a software list.
- Practice manager or operations leadCovers practice management software, subscriptions, device rules and trust accounting systems. It fills out the data and technology side of the report.
Everyone completes the same core survey. The job titles you choose in setup decide how people are grouped and how the results read.
Questions your report answers
- Which AI tools are lawyers, paralegals and support staff using on matters, and which of them sit outside systems the firm chose and pays for?
- At what points in a typical matter does AI-assisted work enter drafts, and is there a named person who verifies it?
- Which controls would a PI insurer or a panel client expect to see, and which of them can the firm evidence today?
- How many hours a week go on repetitive tasks such as chronologies, summaries and transcription that better-governed tools might shorten?
Law firms operate under the conduct rules set by their state or territory regulator or in New Zealand under the Lawyers and Conveyancers Act 2006 and the Law Society's rules. These include duties of confidentiality and competence. Firms also handle personal information under the Privacy Act 1988 (or the NZ Privacy Act 2020), and several courts have issued practice notes on generative AI. Professional indemnity insurers set their own conditions. Check current requirements with your regulator and insurer. The report is not legal or insurance advice.
Questions from law firms
What should an AI policy for a law firm cover?
At a minimum: approved tools, what client information may and may not be entered, verification of anything AI-assisted, supervision of junior staff, client disclosure, and who owns decisions. The survey report gives you the facts about current use so the policy matches real practice.
How do I find out if lawyers are putting client matters into ChatGPT?
Ask everyone, safely. The survey asks each person which tools they use for which tasks, including personal accounts, and you see results only as groups of three or more. That tends to produce more honest answers than a memo reminding people of the rules.
Does the survey look at our matter files?
No. It collects what people say about their own role, tasks and tools. It does not connect to your practice management system, read documents or review any client matter. Not every firm needs the same rules, but every firm needs to know which tools are actually in use. The report is built for that.
Is the report a compliance certification?
No. It is an assessment and a roadmap based on staff answers and your own policy and systems answers. It does not certify compliance with any rule, and it is not legal advice. Your regulator, insurer and legal advisers remain the authority. It is a snapshot of what people say, not a technical audit of devices or networks.
How much does it cost for a firm of 40 people?
Pricing is a flat A$1,990 + GST once per organisation, not per head. You complete a short setup wizard and add staff emails. The report arrives automatically when the survey closes, with no software for you to run. Optional membership re-invites the team about every 90 days and emails you a fresh report.
See how your business really works.
Ten minutes of setup. Your team answers on their phones. A report arrives when the survey closes.