The dominant narrative in legal AI coverage is that the gap between the haves and have-nots is widening. Big firms with seven-figure innovation budgets are buying enterprise tools, building internal AI teams, and pulling further ahead of everyone else. Small and mid-sized firms, on this telling, are destined to fall behind.
We think this gets the dynamics wrong, or at least incomplete. There is a real, time-limited window for smaller Australian firms to get ahead rather than fall behind, and the structural features of the market favour them more than they currently believe. The catch is that the window is closing, and most small firms are not exploiting the advantage they actually have.
There is a reason a 400-person firm takes 18 months to roll out a new research workflow, or commit to a new legal AI platform. The decision touches a steering committee, IT security review, vendor risk and procurement, change management, training across practice groups, partner buy-in across competing fiefdoms, and a long tail of compliance and conflicts checks. None of that is unreasonable. It is what running a 400-person professional services business looks like.
A 10-person firm deciding to change how it handles legal research has none of that. It needs one partner who is convinced and a tool that is worth using. The decision can be made on a Monday and operational by the end of the week. If it does not work, the decision can be reversed on Friday with a single conversation.
This is a structural speed advantage, and some small firms under-exploit it. The instinct in many smaller practices is to wait, watch what big firms do, and then follow. That instinct is upside-down for technology adoption in the current cycle. The advantage of being small is that you can move before the big firm has finished its second procurement meeting. The cost of waiting is that you get to the destination at the same time everyone else does, with no differentiation to show for it. There's also a training element here - the earlier your organisation is fostering a culture around genuine AI awareness and technical capacity, the better.
The second misconception worth addressing is the assumption that big firms have access to better tools, and that smaller firms are working with diminished versions of what the top end of the market is using. This is ironically the reverse of what is actually true.
The tools being procured by the largest firms are typically built for enterprise procurement criteria. They need to satisfy global IT security frameworks, integrate with existing document management systems, support thousands of users across multiple jurisdictions, and pass through 12-month vendor evaluation processes. These are real requirements for those firms. They are also requirements that shape the product. The tool a large firm ends up with is optimised for being deployable inside a large firm, not necessarily for being the best at legal research or drafting or contract review. Some larger firms take a different perspective, and opt for specialist tools, but many have adopted legal AI that is generalist and borad because of the belief it wll generally lead to better adoption across a range of practice areas.
Tools built for smaller Australian practices have different design constraints, and those constraints often produce better tools for the actual work. A platform that is built for Australian law specifically, with a corpus of Australian legal authority, with workflows designed around how Australian lawyers actually research and draft, with a small enough customer base that product feedback from a partner gets read and acted on within a week, is in many ways a more capable tool for the practitioner than a generic enterprise platform retrofitted for the Australian market.
This matters because the comparison most small firm partners run in their head is wrong. They imagine the choice is between "what big firms have" and "a worse version of that for us". The actual choice is between "what big firms have, which is general-purpose and optimised for procurement" and "what we can have, which is tailored, jurisdiction-specific, and optimised for actually doing the work". The second is often the better tool on the merits, before the speed-of-adoption advantage even enters the calculation. And indeed, more specialist tools often come at a much more reasonable price point.
Two things are happening at once in the Australian legal market.
The first is that client pricing pressure has intensified into a structural feature, not a cyclical one. Mid-market clients are increasingly asking for fixed fees, capped retainers, and itemised value justifications. They are not going back. A firm that turns around solid contract review or a research memo faster, at lower cost, with the same quality, wins the client comparison. Size does not enter the equation at the level of the individual matter.
The second is that AI in legal practice has moved past the speculative phase. The Federal Court's Generative AI Practice Note (GPN-AI), issued by Chief Justice Mortimer, sets out the framework for responsible use in court proceedings rather than banning AI outright. The Supreme Court of NSW Practice Note SC Gen 23 has been operational since February 2025. State courts in Queensland, Victoria, South Australia and the federal jurisdiction now have working protocols. The regulatory infrastructure is in place. The question is no longer whether AI is permissible in legal practice. It is how to use it well.
The productivity gap between a firm using AI properly for research, analysis and matter prep and a firm not doing so is now large enough to be visible to clients. Research that used to take a junior lawyer six hours can be foundationed in 30 minutes, leaving the senior lawyer with a properly briefed starting point rather than a partially formed one. Contract review that used to require two days of initial read-through and flagging can be reduced to half a day, with the remaining time spent on what actually matters, which is judgement on operative terms and deeper research.
There is a temptation to treat AI adoption as a procurement decision. Buy the tool, run the training, move on. We think this misreads the work.
Three things distinguish a firm that gets value from AI from one that ends up paying a subscription for software no-one is using.
First, the partner pushing the change has to use the tool themselves. Not in a demo. In live matters, on real researchquestions or everyday workflows, with real verification. If the partner cannot articulate from their own experience what the tool does well and where it fails, the rollout will not survive contact with junior lawyer scepticism.
Second, there has to be a clear written process for how AI output is reviewed before it leaves the firm. This is not a 30-page policy. It is a one-page document that says: here is what you can use the tool for, here is what you cannot, here is what verification looks like for each task, and here is who reviews and signs off. This is the document the Federal Court is implicitly asking you to have. Most firms do not have it.
Third, the tool or platformhas to be one that is built for the actual work. Generic LLMs do useful things and are appropriate for many legal tasks, but they are not a substitute for tools built on Australian legal corpora, with proper citation, with workflows that reflect how Australian lawyers actually work. The hallucination problem is dramatically worse on generic tools precisely because they are not grounded in jurisdiction-specific authority.
The point of all of this is that there is genuine competitive ground to gain right now, and that the window for gaining it is shorter than most small firms assume.
The firms that move in the next 12 months will have a year of compounding workflow improvement, a year of client-visible cost and turnaround advantages, and a year of internal AI-assisted practice that becomes harder to replicate the longer it goes on. Habits set under pressure become culture. The firms that move in the next 36 months will be at parity with everyone else. At that point AI-assisted workflows are baseline expectation, not differentiation.
The question for a small firm leadership team is which side of that line they want to be on, and how much of the available advantage they want to capture before it stops being available.
This is what we mean when we say there is genuine competitive ground to gain. Especially against the firm two suburbs over with a similar client base, similar fees, and similar quality, who is still waiting to see how AI develops before doing anything about it. The small firm has the speed advantage and can even use this advantage to scale rapidly, if so desired. The tools available now are often better suited to the actual work than what large firms are evaluating. Both of those facts run against the prevailing narrative, and both of them favour the firm that moves first.
