Powered by
MOMENTUM
MEDIA
Clients now arrive with chatbot analysis of their matter. Firms that read it, correct it, and protect it will win more trust than those that brush it aside, writes Andrew Cooke.
Picture a first meeting. The client runs a 30-person engineering firm in Bunbury, Western Australia. She slides 11 pages of chatbot analysis across the desk. Its verdict: a former employee’s unfair dismissal claim is baseless, and she should fight it on principle. She is calm, certain and wrong on two of the three points that matter.
The reflex is to push the pages aside and start again. Resist it. That printout is the most candid intake form you will ever receive.
Your clients now consult AI before they consult you
The question is no longer whether clients use AI first, but what you do with what they bring. The Legal Services Research Centre’s pilot V-LULU survey found that four in five people facing a legal issue look for help online. Preliminary Fair Work Commission research suggests 40 to 50 per cent of applicants in unfair dismissal and general protections matters use AI to prepare their claims, and the commission’s president attributes a workload rise of more than 70 per cent over three years mainly to litigants’ use of AI. Respondents are doing it too: the employer across your desk has often run the same exercise from the other side. In the UK, research by the charity JUSTICE found some people use chatbots to check the service their own lawyers are providing.
Chief Justice Stephen Gageler has warned that judges are acting as “human filters” for AI-generated arguments. Practitioners are now the first filter, and a good filter can do more than block.
The printout records what the client wants to hear
General-purpose chatbots tend to mirror the tone of the person asking. As Professor Jeannie Paterson observed in The Australian Financial Review, a litigant who feels deeply wronged is likely to have that grievance reflected straight back. So, the document on your desk is less a legal analysis than a map of the client’s fears, hopes, and the story they are telling themselves. That is exactly what a first meeting is meant to uncover, and the client has written it down for you.
The insight: That printout is the most candid intake form you will ever receive.
3 moves turn the printout into trust
First, read it as evidence, not error. Ask what they typed in. The prompt often reveals more than the answer: the facts they led with, the ones they left out, and the outcome they were hoping for.
Second, correct it out loud. Go through it with them. Agree where it is right, because conceding the sound points costs nothing and earns credibility for the disagreement. Then show where it is wrong and why. This is where your judgement becomes visible, and visible judgement is what clients now pay for. The explanatory layer of legal work, what a claim is and how the process runs, is now free to anyone with a phone. Firms that keep billing for it will feel the pressure on fees first. The answer is to price the review, not hide it: quote a fixed fee at intake for reviewing the client’s AI output, so the correction becomes a named service rather than a write-off.
Third, close the privilege gap. In February 2026, Judge Jed Rakoff of the US District Court in New York held in United States v Heppner that a defendant’s exchanges with a consumer chatbot were protected by neither privilege nor the work product doctrine. Whether an Australian court would take the same view remains to be seen, and no client should be the test case. A client who pastes your advice into a public chatbot to understand it better may be putting that advice at risk. One plain paragraph in the engagement letter, repeated at the first meeting, protects them and shows you are on their side.
Make it a firm habit, not a personal skill
One partner handling this well is a talent; a whole firm handling it well is a service. Add a standard intake question: have you used AI on this, and will you bring us what it said? Give juniors a short script for the conversation and a supervisor to call. Apply the same verification discipline to client material that the courts now demand of yours. In NSW, Practice Note SC Gen 23 makes clear that verification cannot be done by a generative AI program.
AI is 20 per cent technology, 80 per cent psychology. The hard part here is not the tool. It is the client’s emotional investment in the answer it gave them, and that is human work.
The client who leaves with more than she brought
Back to the engineering business owner. Forty minutes later, she has kept the one point that held, understood why the other two fail, and stopped pasting her lawyer’s emails into a chatbot. She agrees to a strategy she believes in, because the firm took her work seriously before improving it.
That is the standard for every client who walks in with AI-generated homework: not dismissed, not rubber-stamped, but turned into advice they can act on. Clear, safe, and usable.
Andrew Cooke is the founder and managing director of GPS-AI (Growth & Profit Solutions AI), a boutique AI transformation and governance consultancy based in Bunbury/Perth, Western Australia.
