Jevons paradox has been turning up a lot in legal conversations lately. It surfaces on panels about AI and the future of the department, in vendor decks, in the occasional LinkedIn post arguing that efficiency will not shrink legal work the way everyone assumes.
The concept is older than the legal-tech market by more than a century, which is part of why it is worth trusting rather than dismissing as hype. When William Stanley Jevons studied coal consumption in the 1860s, he observed that increasingly efficient steam engines had not reduced Britain’s coal use.
Efficiency lowered the cost of using steam, cheaper steam opened uses that had never penciled out before, and total coal demand rose. The pattern has held across a great many technologies since: when you reduce the cost of doing something valuable, you usually get more of it rather than less.
The question for a legal department is not whether the paradox is real, but where the recovered work will come from and how to resource it once it arrives.
Why the Jevons Paradox Matters for Legal Teams
The term has earned its moment, because it names something most general counsel are already feeling: the cost of routine legal work is falling as AI absorbs more of it, and the matter count has still climbed for most of the year. Those two trends sit side by side on the same desk, and the instinct is to treat them as a contradiction that reality will eventually resolve in one direction. The paradox is worth taking seriously precisely because it suggests they will not resolve, and that the department planning for one number to correct the other is planning for the wrong future.
It is worth checking whether the idea actually holds up in legal, rather than borrowing it as a slogan. Both halves of it are grounded in something real. Anthropic’s labor-market research suggests AI could accelerate a substantial share of the tasks that make up legal work, and the direction of travel on cost is not in serious dispute.
On the demand side, most legal leaders can confirm the trend from their own docket before they see a survey. FTI Consulting’s 2026 General Counsel Report found that 97% of the general counsel interviewed reported rising work volume in at least one category, with new regulation the most frequently named driver. The result is a department getting cheaper to run per matter and busier in absolute terms at the same time, which is exactly the shape the paradox predicts.
How AI Efficiency Increases Legal Work Volume
Most departments are probably carrying more work than the volume figures suggest, in the form of matters that were priced out rather than genuinely resolved. Some work gets handled at a lighter touch than it would warrant in a world of unlimited resources, some gets deferred, and some gets written off entirely because pursuing it costs more than it appears to return.
None of that necessarily reflects an absence of demand. In many cases it reflects a price that was too high to clear, and that work tends to sit in a kind of shadow docket that never shows up in the volume figures because it was never formally taken on.
If that is right, then lowering the price should start to bring some of it into view. Work that was not worth doing at the old cost becomes worth doing at the new one, whether that means a fuller review, a closer look at something that used to get a sample, or a matter that finally justifies the effort to pursue.
Alongside that recovered demand, the technology tends to generate work of its own, from AI governance frameworks and model output verification to regulatory interpretation and the disputes that follow when an automated system makes a consequential mistake. The document-heavy phase of any single matter may well compress even as the number of matters a department is expected to stand behind keeps climbing.
Why a Growing Legal Queue Is Not One Resourcing Problem
This is where the resourcing conversation usually goes wrong, and it goes wrong in a specific and avoidable way. Faced with a growing queue, most departments reach for whichever lever is closest to hand. The internal team absorbs the overflow until someone is visibly underwater, the panel firm receives the excess at panel-firm rates, or a headcount request goes up the chain.
Each of those is a defensible answer to a particular question. The failure is answering every question with the same lever, because a growing queue is almost never one problem. It is a stack of different problems that happen to have arrived through the same intake.
Pull a single contracts backlog apart and the distinctions become obvious. A run of standard NDAs and order forms is a candidate for self-service tooling or automated review, and putting a qualified lawyer on it is a waste of a scarce resource. A large portfolio of similar commercial agreements is a managed-service or flexible-lawyer problem, where the value is in throughput and consistency rather than bespoke judgment.
A genuinely novel regulatory question is a boutique problem, where narrow expertise is worth paying for. A business-critical negotiation is an internal problem, because it turns on knowing what the company will actually sign and who has to be consulted before the signature page goes out. Four kinds of work, four different right answers, one queue.
How to Route Legal Work: A Resourcing Framework
What separates the departments that route well from the ones that route by reflex is that the former ask a consistent set of questions before they assign anything.
- How repeatable is this work?
- What is the cost of getting it wrong?
- How much company-specific context does it require?
- Where does final accountability have to sit?
- Is the need temporary, recurring or permanent?
- What blend of technology and human expertise actually produces the outcome.
Those questions are not sophisticated on their own. Asking them deliberately, matter after matter, rather than defaulting to the familiar lever, is what turns resourcing from a reflex into a discipline, and it is what makes the resulting allocation something a general counsel can defend to a board, measure against results and adjust when the facts move.
Why Your Legal Resourcing Mix Has to Keep Changing
They will move. A decade ago the routing decision had two settings, keep it inside or send it to the panel, and a competent department could set that allocation once a year and largely leave it alone. A single matter today might run through an internal lawyer working from an AI-generated first draft, an alternative provider handling the volume review, a boutique advising on one narrow point of law and a flexible specialist coordinating the whole thing.
The boundary between what software can execute and what still needs a human sits in a different place every few quarters, and regulation, business priorities and the department’s own appetite for risk keep shifting alongside it. A resourcing model built for a world that holds still is the one that quietly breaks, usually in the form of an experienced lawyer who should be advising the business instead spending a Friday sorting routine agreements because the queue was full and nobody stopped to route it.
The Legal Judgment AI Cannot Replace
For all the movement, one thing stays stubbornly scarce, and naming it correctly is the point of the whole exercise. Most consequential negotiations reach a moment when everything depends on one person being reachable, someone who understands what the company will and will not sign, can translate legal risk into terms the business can act on and knows exactly who needs to weigh in before the deal closes. Faster review does nothing to make that person more abundant.
Odini Gogo, a technology transactions lawyer who practices flexibly through Priori, frames the distinction as the ability to give legal advice the business can digest, and the strongest in-house lawyers do more than flag the risk. They show the business how to get where it wants to go in spite of it. That blend of judgment, context and accountability is the resource worth protecting, wherever it happens to sit, and the corollary matters just as much: a department that spends senior judgment on work a tool or a repeatable service could have handled is misallocating the one input that efficiency will never make cheap.
It is worth saying plainly that flexible talent has not retreated to the low end of this market as the technology has improved. Departments still use it for standard agreements and high-volume projects, and that has not changed. What has changed is the ceiling. An experienced specialist can now supply senior-level judgment on a specific matter without the department having to buy the full traditional provider model around it or commit to a permanent hire it may not need in a year.
Building a Legal Department Resourcing Model for AI
The through-line is uncomfortable for anyone hoping AI would simplify the resourcing question. Cheaper execution does not lower the stakes of deciding who does the work. It raises them, because the mistakes are now more visible and the range of available answers is wider than it has ever been.
The departments that come out ahead will not be the ones that bet everything on automation or everything on headcount. They will be the ones that know what is actually entering the queue, hold a clear and consistent set of routing criteria and run a capacity mix built to change as fast as the work does.
