In 2001, China joined the WTO. Over the next decade, c5 million American manufacturing jobs vanished - Autor, Dorn, and Hanson showed that Chinese import penetration accounted for c59% of US manufacturing losses between 2001 and 2019 [1]. The aggregate was positive. The distribution was savage. Nobody built the institutions to manage the difference.
AI is producing the same asymmetry. But the commodity being redistributed isn't manufacturing labour. It's intelligence itself.
What does this mean if you sell expertise?
If you run a law firm, a consultancy, or an accounting practice, your pricing depends on intelligence being expensive. Clients pay £500 an hour because structured thinking is scarce and hard to produce. A frontier model now produces competent legal research, financial modelling, and market analysis in minutes - work that used to fill the first five years of a professional career.
The internet equalised information - a farmer in Zhejiang got the same market prices as a New York retailer. Alibaba, founded in Jack Ma's apartment in 1999, became the mechanism through which millions of micro-entrepreneurs accessed global demand. But the internet didn't equalise intelligence. The MBA could still do things with that information - analyse, model, strategise - that required years of expensive training. That gate held.
AI is opening it.
Who benefits most when intelligence gets cheap?
Not the partner at Clifford Chance - for her, AI is a productivity tool. Useful but incremental. The transformative impact is elsewhere. A first-generation entrepreneur in Lagos can now access structured legal analysis, financial modelling, and market research that previously required a $500-per-hour consultant. England and Wales have c170,000 practising solicitors serving 60 million people [2]. Nigeria has a fraction of that capacity for 230 million. The unmet demand is not a gap. It is a chasm.
The leapfrogging precedent is real. Kenya's M-Pesa brought mobile financial services to millions who never had a bank account. Bangladesh's bKash has reached over 82 million users [3]. Both skipped infrastructure that Western nations spent a century building.
And there is a geopolitical accelerant. China's DeepSeek - spun off from a Hangzhou hedge fund [4] - has released capable open-source models at prices that undercut Western competitors by an order of magnitude. If open-source models drive the price of base intelligence toward zero, the primary beneficiaries aren't Chinese firms. They are the billions of people currently priced out of professional cognitive services entirely. The geopolitical intent may be commercial competition. The geopolitical effect may be the largest accidental development intervention in history.
The marginal value of cheap intelligence is highest where prior access was lowest. That is the asymmetry.
Where does the value actually sit for incumbents?
Most commentary focuses on cognitive disruption. But the biggest moat for incumbent firms isn't intelligence - it's structural. Client relationships built over twenty years. Professional networks. Institutional trust. Regulatory access. The ability to call the right person at the FCA. The reputation that gets you in the room.
AI can draft the memo. It cannot have lunch with the general counsel.
Your firm's value has two components: cognitive (research, drafting, analysis) and structural (relationships, trust, access, reputation). The first is being commoditised globally. The second is becoming more valuable. The professionals who thrive will be those who understand which component they're actually selling.
What happened when we ignored this pattern last time?
The China Shock's deepest failure wasn't trade. It was the policy vacuum: inadequate retraining, non-existent transition support, no effort to redirect gains toward affected communities [1]. The Rust Belt hollowed out. Communities that had been middle-class for generations fell into despair.
The professional services version: a five-year-qualified associate at a 200-person City firm who built her career on due diligence reviews and research memos. That work trained her judgment - but it's also the work AI now handles. She's not redundant, but her role needs fundamental redesigning. Multiply that across the profession, and the scale of the transition becomes clear.
What should you do?
You have the advantage of seeing the pattern before it repeats.
Audit your team's work this quarter. Separate the cognitive output that AI can replicate from the judgment and relationships it cannot. Restructure toward the second.
Build new training pathways. If you automate the apprenticeship, you lose the pipeline to mastery. The professions need their equivalent of simulation and war games - before the gap opens.
Track the exposure. After every major model release, ask: what percentage of our billable output could a frontier model produce at 80% quality? That number is your vulnerability.
The firms that restructure first will absorb the talent and clients from those that don't. Intelligence is not geography. But the assumption that it will always command a premium is exactly the kind of belief that gets destroyed by a commodity shock.
If you're navigating this transition, book a call with Lion Strategy.
Notes
[1] Autor, D., Dorn, D., and Hanson, G., "On the Persistence of the China Shock," Brookings Papers on Economic Activity, Fall 2021, pp. 381-447.
[2] Solicitors Regulation Authority, Annual Financial Statements, October 2024. Approximately 170,000 practising solicitors regulated across c9,200 firms.
[3] bKash company data, January 2026. See also: GSMA, "Mobile Money Usage in Bangladesh," October 2025.
[4] Fortune, "Meet the Hedge Fund Manager Who Founded DeepSeek," January 27, 2025.