AI Is Making Lawyers Busier: Jevons Paradox, Live on Odd Lots

In the northern deep there is a fish; its name is Kun.
It transforms into a bird; its name is Peng.
— Zhuangzi, “Free and Easy Wandering” (c. 4th century BC)

First, the pitch: what is Odd Lots

One minute on the show itself, because it earns it.

Odd Lots is Bloomberg’s flagship finance podcast, hosted by Joe Weisenthal and Tracy Alloway — two Bloomberg veterans with bottomless curiosity about how the world actually works. While other finance shows discuss what went up or down, these two interview people about paperclip supply chains, grid transformer lead times, uranium enrichment capacity — the actual plumbing of markets. Several episodes a week, about an hour each; guests range from central bankers to hedge fund managers to port union bosses.

I follow the show regularly, and this episode demanded a write-up.

The original episode (strongly recommended): Why AI Might Actually Create More Work for Lawyers (55 min)

What the episode is about

The guest is Gary Wingens, chair of Lowenstein Sandler — the top manager of a major US law firm. Law is the archetypal sell-hours-by-the-clock business, in theory the exact business model AI should destroy. Instead, he came on the show to explain why AI is making his firm busier and more profitable.

The whole episode plays like a live broadcast of the Jevons Paradox.

My highlights

• Costs down 70%, and a dead deal comes alive. A matter requiring review of thousands of trust agreements was quoted around $10M three years ago; the client passed — bad risk/reward. With AI compressing the cost to roughly $3M, the client signed immediately. The work didn’t disappear; demand that had been priced out of existence got unlocked. That’s Jevons: efficiency raises total consumption.
• Patent work quadrupled. A client’s own engineers use AI too, so invention output exploded and internal patent filings quadrupled. AI cuts cost on the supply side while creating demand on the demand side — burning the candle from both ends.
• Billing rates up, total bills down. Big-law average hourly rates rose 10.1% in 2025 (CPI: ~3%). But the other half matters: the hours needed per matter collapsed. Clients pay less in total while each lawyer-hour is worth more. Time is depreciating; judgment is appreciating — I expect that scissors gap in every professional service.
• A 180-degree turn in two years. Two years ago, malpractice insurers asked “are you letting lawyers use AI?” (as a risk). Now they ask “are you making sure lawyers use AI?” Clients flipped identically: from “don’t use AI on my matters” to “you must use AI to cut my bill.” The one red line left: citing AI-hallucinated cases in court. Deeply embarrassing.
• Building your own model doesn’t pencil. A rival firm announced $500M over five years for AI infrastructure; Wingens does the math out loud — training a frontier model from scratch starts around $1.5B, and you’d never catch up. His firm instead layers proprietary assets on top of existing models: their own clause libraries, historical briefs, deposition archives. The moat isn’t the model; it’s the corpus you feed it.
• Pricing foreshock. Legal AI tools today are mostly all-you-can-eat subscriptions, about to shift toward usage-based billing — and he admits the industry “doesn’t yet know the true cost” of AI services. What I hear: AI’s price discovery has barely begun.

Where my thinking goes

1. I am a walking sample of the Jevons Paradox. I run a one-person company. AI didn’t shrink my workload — it moved everything from “impossible for one person” onto my calendar: a self-built market database, a transcript pipeline, a multi-model engineering team. Efficiency up → ambition up → total work up. True for lawyers, true for individuals; I suspect it will be true for corporate IT budgets.

2. A clean test for investors (educational, not advice). When a company talks about AI, is it describing costs saved or previously-uneconomic demand unlocked? The former is defense — competition arbitrages it away (most of that 70% saving ends up as the client’s negotiating leverage). The latter is a growth story. Wherever Jevons holds, compute and token consumption only compound — which is why I keep watching the AI infrastructure chain.

3. “Time depreciates, judgment appreciates” is now everyone’s personal problem. It’s true of billable hours and equally true of investment research: once AI drives the cost of finding, organizing, and comparing toward zero, only two things retain value — asking the right question, and owning the conclusion. Over the past six months I’ve moved my own center of gravity from doing to specifying and verifying, and the mechanics Wingens describes felt uncannily familiar.

References

• The episode: Odd Lots — Why AI Might Actually Create More Work for Lawyers (or search “Odd Lots” on any podcast platform)
• New to Odd Lots? Start with an episode in a domain you know — the “so that’s how the plumbing works” feeling is the hook
• Jevons Paradox source: William Stanley Jevons, The Coal Question (1865) — steam engines got efficient, coal consumption exploded

---

Listening notes and educational discussion, not investment advice. Companies and products appear only as episode context; I hold no interest in the private companies mentioned — for listed ones, do your own research. Quoted content belongs to the original show; go listen to the source.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.