The #1 AI myth that lawyers need to know
These four assumptions are fuelling the myth that “AI needs to be 100% perfect to be useful”. Pulling them apart reveals that the real question is not whether to use AI, but how.
One of the most common myths about AI is that it “needs to be 100% perfect to be useful”. The myth rests on four assumptions worth pulling apart.
Assumption 1: AI can replace lawyers
This assumes AI is designed to replace legal judgment. It cannot. AI cannot discharge the responsibilities at the heart of legal practice: managing a competent practice, upholding fiduciary duties to clients, and meeting your paramount duty to the court. Those sit with you, the lawyer in charge.
A line from a 1979 IBM training manual sums it up: "A computer can never be held accountable. Therefore a computer must never make a management decision." That applies just as well to AI today.
When I had a graduate lawyer working under me, I never let their work go out the door without reviewing it first. Not because I did not trust them, but because it was going out under my name. AI is no different. Think of it as a tireless contributor sitting between a well-read junior and a fast search engine. Use it accordingly and never skip the review.
Assumption 2: AI sounds human, so it thinks like one
When an AI tool gives you coherent responses and uses legal terminology correctly, it is easy to assume legal reasoning is happening behind the scenes.
There is not. AI does not think. It predicts.
A large language model translates words into "tokens" and analyses how each relates to others across vast amounts of text. The crucial limitation: LLMs prioritise fluency over accuracy. They predict what the next word should be based on patterns, but do not actually know whether that word is correct.
As AI researcher Jey Han Lau from the University of Melbourne puts it, LLMs "were never made to distinguish between facts and non-facts, or reality and generated fabrication."
For lawyers, this means AI can sound confident about something wrong, miss nuances that require judgment, and hallucinate details that fit the pattern but are not true. Understanding this makes you a better user of AI, because you know where to apply your judgment.
Assumption 3: Humans are 100% accurate
Humans make mistakes too. It is why we create workflows, checklists, and review processes. And yet we apply a curious double standard to AI, holding it to a standard we would never apply to a junior lawyer.
When I was in private practice, my insurer reminded me regularly that even the best lawyers make mistakes under fatigue and time pressure. The same logic applies to AI.
The pairing works because the weaknesses are complementary. AI offers consistency and speed but struggles with nuance. Lawyers bring expertise and context but experience fatigue. Together, you get work that is faster, more thorough, and more consistent, without sacrificing professional judgment.
Assumption 4: If I avoid AI, I avoid risk
This last assumption offers a false sense of security. Being a lawyer in an AI-driven world is not just about whether you use AI. It is about recognising when others have, whether it is a colleague, a client, or an expert witness.
A striking example came from the ACT Supreme Court in 2024. In DPP v Khan, an AI-generated character reference was submitted during a sentencing appeal from the brother of the accused Justice Mossop was sceptical, noting the vague praise and oddly formal tone. His Honour was particularly alerted by phrases suggesting the reference provider had known the accused “personally and professionally for an extended period of time”. His Honour gave the reference no weight and criticised both counsel and solicitor for failing to recognise the reference had been written using an LLM, and for failing to enquire about it.
AI proficiency is fast becoming an essential professional skill, both for using AI well and for recognising it when others have.
So, where does this leave us?
Most of the anxiety about AI in law comes from treating it as either a magic lawyer-in-a-box or a career-ending asteroid. In reality, today's AI is closer to a fast junior who never sleeps, sometimes misunderstands instructions, and occasionally invents case citations with heroic confidence.
Used well, AI removes friction from first drafts, summaries of complex documents or cases, checklists, and administrative work that must be done even when it is not billable. The question was never whether AI is perfect. It is whether, with the right processes around it, AI can deliver efficiency gains and help you produce better work. The answer is yes. However, the lawyer harnessing these tools remains the decision maker who assesses the quality of the output and the right use case for the AI tool.
Prathna Tiwari is Head of Legal Content at LawY, and a commercial and property lawyer. To explore how LawY can help your firm reclaim hours every week, visit lawy.ai.
