Getting equity partners who have never used AI to say yes

By Michael Nadalin, Founder, Market Lead · 28 September 2026 · 5 min read

You have used an AI research tool for three months. It saves you time on first-pass research and on summarising long affidavits, and you have learned where it goes wrong. Now you want the firm to pay for it properly, with a policy, training and a licence for everyone. The problem is that the five people who vote on it have never typed a single prompt, and at least one of them has read about the lawyers who filed fake citations.

That partner is not being unreasonable. They carry the professional indemnity exposure, they sign off on the trust account, and they remember every earlier technology rollout that cost money and changed nothing. Enthusiasm from a senior associate does not answer any of that. A proposal that treats their scepticism as the starting point usually does.

This piece sets out what equity partners who have never used AI tend to worry about, how to design a small pilot they can say yes to, and how to run the partners' meeting so the decision rests on evidence from your own files rather than on vendor slides.

What partners believe, and what is actually true

Partners who have not used AI form their view from headlines and from what peers mention over lunch. The resulting beliefs are not all wrong, and a few of them should shape your proposal. The mistake is to argue against all of them at once. Pick out the ones that are simply out of date and deal with those, then take the valid concerns seriously and build your controls around them.

The fabricated citation concern is the most important one to get right. It is real, and courts have responded. The Supreme Court of New South Wales issued Practice Note SC Gen 23 on the use of generative AI, and other Australian courts and regulators have published their own guidance. A partner who raises this has done their reading. Your answer is not that the risk has gone away. Your answer is the verification process the firm will require before anything reaches a court or a client.

The objections that actually decide the vote

Not every objection carries the same weight. In most partnerships the decision turns on risk and money, and the conversation about working habits comes a distant third. If you spend your time showing clever outputs, you are answering the least important question.

Rank the concerns the way the decision makers will. Professional obligations come first: confidentiality under the Australian Solicitors' Conduct Rules, the duty of candour to the court, and competence. Then insurance and liability. Then the effect on revenue from time-based billing. Only after that will anyone care whether the tool is pleasant to use.

Design a pilot a cautious partner can approve

The easiest thing for a sceptical partner to approve is something small, reversible and measured. A request for firm-wide licences and a policy invites a debate about the whole future of the practice. A request for eight weeks, four users, one practice group and a report at the end invites a much shorter conversation.

Choose tasks where mistakes are easy to catch and where the output is checked before it goes anywhere: first-pass research memos, chronologies from discovery, summaries of long judgments, issue lists for a brief to counsel. Leave out anything filed or sent to a client without a senior lawyer's review. Record time spent and errors found on each task, including the ones the tool got wrong. A pilot report that lists no errors will not be believed.

Running the partners' meeting

Most proposals fail at the meeting because the presenter spends the time talking about AI in general. Partners who have never used the tool need to see it working on a problem they recognise, from a matter the firm has actually run, with the flaws shown rather than hidden.

A split like the one below keeps the demonstration short and leaves the most time for questions. The part that wins the room is often the section where you show the tool getting something wrong and then show how the verification step caught it. That is the moment a sceptical partner sees that the risk is managed rather than ignored.

After the yes: keeping it from stalling

Approval for a pilot is not adoption. The common failure is a pilot that quietly fades because users are busy and nobody is asking for the report. Put the review date in the partners' meeting diary on the day it is approved, and give one partner ownership of the result. If that partner has used the tool themselves by the end, the next vote is much easier.

Pick a tool that shows its sources so every proposition can be traced to a judgment or legislation that a lawyer can open and read. That single feature answers most of the verification concern. Legal Brain is built on that principle for Australian research, but whichever tool you trial, test the sources against your own knowledge of the area before relying on it, and let the partners see you do it. Sceptical partners are not persuaded by claims about the tool. They are persuaded by a colleague they trust showing the checking process.

Frequently asked questions

How do I convince senior partners to use AI in a law firm?

Start with their risk concerns, not the features. Propose a small, time-limited pilot with named users, clear verification rules and a written insurance position, then report the results honestly, including the errors. Partners who can see the controls working are far more likely to approve a wider rollout.

Is it a breach of confidentiality for Australian lawyers to use AI?

Not automatically, but it can be. It depends on what information is entered and on the tool's data handling, retention and training terms. Firms should approve specific tools after reviewing those terms and set rules about client identifying material, consistent with the confidentiality obligations in the Australian Solicitors' Conduct Rules.

Do Australian courts allow lawyers to use generative AI?

Several Australian courts have issued guidance rather than blanket bans. The Supreme Court of New South Wales, for example, has Practice Note SC Gen 23, which places limits on the use of generative AI in certain documents and requires verification. Check the current practice notes for each court you appear in, as they differ and are updated.

Should a law firm have an AI policy before a pilot?

Yes, at least a short one. It should name the approved tools, set out what information can and cannot be entered, require verification of every authority and identify who supervises use. A pilot is a good way to test and refine that policy before it applies firm wide.