Earlier this month, a story circulated in the legal press about a lawyer who billed a client for 34 hours of work in a single calendar day. The explanation was not fraud but a billing system that aggregated time across matters without flagging the arithmetic impossibility of the total. The story was absurd on its face. It was also a precise illustration of the structural problem with time-based billing in an era when AI can compress four hours of work into forty minutes: the system has no mechanism for separating time spent from value delivered.
The problems with the billable hour are not new, and Australian courts have not been shy about them. Kirby J observed in his 1998 speech 'Legal Professional Ethics in Times of Change' that hourly billing rewards inefficiency: 'the work of lawyer A, who spends 100 hours preparing a motion for summary judgment, costs the client 100 times the billing rate; the work of lawyer B whom it takes 200 hours to do the same work costs the client twice as much for the same service.' Martin CJ made the point in his Perth Press Club address about structural conflict: time billing 'creates an inherent and irreconcilable conflict between the interest of the client in the achievement of an expeditious resolution, and the interest of the lawyer in billing time.' These observations predated AI by a decade. The dynamic they describe has not improved.
The data on Australian firms suggests a profession that recognises this, at least in the aggregate. This rise is reflected in the surrounding media: in December 2025, Law.com ran the headline, 'It's Real Now: With Law Firm AI Use on the Rise, Expect Alternative Fee Arrangements to Pick Up Steam in 2026.' The firms that haven't worked out a position are running out of runway.
The standard defence of the billable hour in an AI context goes like this: clients pay for judgment, not hours, and AI improves quality without reducing the judgment component. A four-hour research task that now takes forty minutes produces a better memo; the value is the same or higher; the billing rate reflects that value.
That argument has some merit and some runway. It will not hold indefinitely, for a simple reason: clients are starting to ask. Large legal procurement functions, particularly in financial services, infrastructure, and government, are including AI disclosure requirements in their matter engagement terms. They want to know where AI contributed to work product. Some are explicitly raising whether AI-generated output should attract the same billing rate as solicitor time. The Legal Profession Uniform Law provides that charging more than a fair and reasonable amount for legal costs can constitute unsatisfactory professional conduct. As AI compresses the time component of a given task, what counts as fair and reasonable is a question that is going to get harder to answer with reference to hours alone.
The more substantive question is what happens to the underlying model if AI continues to compress time across a wider range of tasks. The billable hour works as a pricing mechanism when time is a reasonable proxy for value and effort. AI research tools can compress 30 to 60 per cent of repeatable legal tasks by up to 90 per cent, according to recent industry analysis. At that scale, the proxy breaks down.
The natural trajectory, if AI capability continues at its current rate, is fixed-fee and outcome-based pricing becoming the default for an increasing share of legal work. Complex disputes and novel transactions will hold on to hourly billing longer, where the judgment component is harder to replicate and the scope is harder to fix in advance. High-volume process work, and anything where AI can produce a reliable first draft, will not hold in the same way.
Many Australian firms now offer alternative fee arrangements, with growth concentrated in fixed-fee structures for defined-scope work. Whether firms are getting ahead of that shift or being pushed into it is increasingly the question.
There is a parallel tension inside firms that does not get enough attention. If an associate completes a task in half the time because they use AI effectively, that potentially reduces their billable output for the period. Under a system where practitioners have billing targets, this creates a direct disincentive to use AI, or at least to use it visibly. Partners and associates optimising for their own billing numbers have no structural reason to compress their own productivity measure.
This is not hypothetical. In Legal Services Commissioner v Scroope [2012] NSWADT 107, evidence described considerable pressure on employed solicitors to ensure all chargeable work was correctly captured, with partners concerned about 'leakage', meaning work that was chargeable but not entered into the billing system. That pressure ran in the direction of more billing, not less. AI creates a different kind of leakage: efficiency that practitioners may have an incentive to conceal rather than pass on. Firms that have not restructured their billing targets to account for AI-assisted productivity are likely running this problem right now without knowing it.
The firms handling this most coherently are running fixed-fee pilots on specific matter types where scope is predictable. Debt recovery is the most common starting point: the task is well-defined, AI compresses the drafting and review steps substantially, and the client already expects a fixed fee. The efficiency gain improves margin rather than creating a billing transparency problem, and it builds a track record before the conversation moves to more complex work.
80 per cent of Australian firms plan to implement AI solutions within the next year to drive efficiency and profitability, according to recent industry surveys. The firms that treat billing model reform as a separate, later question are likely to find that clients raise it for them, and that the conversation is substantially harder when it arrives as a dispute rather than a discussion.
The 34-hour billing day is an extreme case, but it illustrates something that runs through this entire debate: systems designed for a different era of legal practice produce increasingly strange results as that era ends. Billing systems that have no mechanism for what AI is changing will produce their own version of the arithmetic problem. Building a position on this now, while there is room to structure the conversation, is worth considerably more than waiting for a client procurement team to raise it first.
