Two thirds of client-facing lawyers have seen AI change how clients behave. The answer isn’t to argue with the chatbot, says LawY CEO Shelley Burger, but to apply judgement.
Over the years, I have spent plenty of time listening to my private practice colleagues from a range of different practice areas. Through those conversations runs a common, thematic anecdote: that client who arrived in your office already armed with the results of their own legal investigations. In years past, that printout might have contained a Google search, a forum thread, or advice from a cousin who knew someone in property. Gently working through your client’s results in that first meeting was simply part of the job. My colleagues often referred to it as a key part of building trust with a client in the early stages of your first matter with them.
What I am hearing post-2023 is that the client arriving pre-armed today carries something far more persuasive. A web search used to give you ten links and then left the reader to reconcile them. A chatbot returns a single, fluent answer, and it tends to tell the person asking exactly what they want to hear, confidently. That tendency is well documented. Researchers from Stanford and Carnegie Mellon, writing in Science this year, found that 11 leading AI models endorsed users’ actions about 50% more often than humans did. People who interacted with them grew more certain they were right and less willing to negotiate and practically deal with the dispute. Consumer tech is designed to please you and agree with you. Law needs the opposite. It needs tools that push back and challenge your thinking. Because a tool that always agrees with you feels great, until you’re in court.
I write this as a lawyer who now leads a legal AI company, so I should be upfront about where I stand. I am optimistic about what this technology can do for our profession. I am also clear-eyed about what general-purpose chatbots are doing to the clients who consult them before they consult a legally-trained professional.
The Victorian Legal Services Board and Commissioner’s 2026 lawyer census found that two thirds of client-facing practitioners have seen client behaviour change because ofAI. Almost half encounter clients armed with inaccurate information. Four in ten receive AI generated documents that require correction. Thirty-seven per cent have had their advice challenged on the strength of a chatbot’s output. Only one in ten report that clients arrive better informed.
The burden does not fall evenly. The areas most affected are family law, small business, immigration, insolvency, and wills and estates. Clients in these areas often have limited legal literacy, little experience of the system and acute sensitivity to cost. They are the least equipped to identify an error, and the least able to fund its correction. That is what concerns me most.
Many firms are responding through their costs agreements, carving review of a client’s AI-generated material out of the fee estimate. I think that is prudent. A submission citing authority that does not exist cannot be glanced at. It must be carefully unpicked. However, a scope clause only prices the work. It does not tell the practitioner how to handle the client.
Our instinct as lawyers is to rebut the chatbot paragraph by paragraph. I understand that instinct well, but I believe it is the wrong response. It is expensive for the client, it elevates the machine to the status of opposing counsel, and it misreads what the client needs. They did not engage a lawyer to adjudicate between two sources. They engaged one for judgement.
What serves them far better is a short, candid conversation about the tool itself. This comes down to four key points:
It is designed to agree with them.
It is not confidential. Unless they are using an enterprise product with proper data terms, whatever they have entered sits on a third party’s servers.
It creates records. In the New York Times copyright proceedings, a US court ordered OpenAI to preserve ChatGPT output logs, including conversations users had deleted.
It can put privilege at risk. Remember the test in Mann v Carnell: privilege is waived where the client’s conduct is inconsistent with maintaining the confidentiality the privilege protects, whether or not the client intends that result. A client who pastes privileged advice into a chatbot for a second opinion may be doing exactly that. And the second robo-opinion will most likely agree with the privileged advice at first, but then eagerly change its tune if further prompting from the client indicates a preference the other way.
It is the human conversation that protects the matter. It also shows the client something they rarely see: what they are actually paying for.
None of this makes AI the villain, and the Australian case law suggests the problem lies elsewhere. UNSW’s Centre for the Future of the Legal Profession examined 87 court and tribunal decisions that recorded the use of generative AI. 68 decisions involved self represented litigants, 10 involved lawyers. The fabricated citations making headlines are overwhelmingly the result ofAI-generated output reaching a court without review by a trained legal professional.
The variable is not the technology. It is the judgement applied to it. That is the very premise on which our team, half of whom are lawyers, built LawY. A general chatbot is engineered to satisfy the person asking. A legal tool must answer to the law, with jurisdiction-specific grounding, and lawyers checking the work before anyone relies on it.
There is a harder lesson here for all of us in the profession. For generations, we have charged for two things: knowledge of the law and the craft of applying and expressing it. Law school trained both, and the billable hour measured both. Today, anyone with a phone can get a rough version of both.
What remains scarce is everything a chatbot cannot do. It cannot read a client, scope a matter, recognise when a deal is good enough to sign, dissuade a litigant from a fight they will lose, or deliver advice the client does not want to hear. When I moved in-house, first as a GC and then as a Chief Legal Officer, those were the qualities I valued most in the lawyers I built our teams with and the external lawyers I briefed. They were always the most valuable parts of the work. We priced the others because they were easier to measure.
The client who arrives with a self-service answer is here to stay, and it would be unfair to blame the client. Most are under pressure, out of their depth and conscious of every invoice. A tool that responds at 11pm, at no cost, and assures them they are right is an understandable place to turn.
The machine is built to please our clients. We, human lawyers, are bound to act in their best interests. I can think of no better reminder of why our profession matters.
Shelley Burger is the Global CEO of LawY and a commercial lawyer. To see how LawY can fit effortlessly into your practice and uplift the work you do, visit lawy.ai.