In 1971, Herbert Simon - Nobel laureate, AI pioneer, one of the few people who deserved the title "polymath" - delivered a lecture containing the most important insight about the economics of abundance anyone has articulated: "A wealth of information creates a poverty of whatever it is that information consumes. What information consumes is rather obvious: it consumes the attention of its recipients" [1].
That was before the internet, before email, before smartphones. Simon was reasoning from first principles about a world decades away. He was exactly right.
Now apply his logic to intelligence rather than information. When machine reasoning is abundant, what becomes scarce? Four things - and they're where every professional services firm should be investing.
What's the difference between intelligence and judgment?
Intelligence answers questions. Judgment decides which questions to ask.
A first-year associate with GPT-5 produces a competent legal memo in minutes. The senior partner's value was never in the memo. It was in knowing whether the memo was the right document to produce. Should the client settle? Is this judge sympathetic to this argument? Is opposing counsel bluffing?
These are judgment calls - pattern recognition built from decades of experience, contextual reasoning no model reliably replicates. Judgment sits above intelligence in the cognitive stack. It becomes more valuable when intelligence is cheap, because the number of competent analyses explodes and the ability to discriminate between them becomes the bottleneck.
What percentage of your professional value is intelligence - producing analysis - versus judgment - knowing which analysis matters? If the answer tips toward intelligence, your pricing power is eroding. If it tips toward judgment, it's increasing.
Who verifies the machine?
Intelligence is invisible. Unlike a physical product you can inspect, the output of an AI model is a probabilistic estimate dressed as confident prose. It may be correct. It may be subtly wrong. It may be hallucinating while maintaining the appearance of competence.
When a commodity is invisible and variable, trust infrastructure becomes enormously valuable. This isn't novel - it's the basis of credit ratings, financial auditing, and pharmaceutical regulation. In each case, buyers can't reliably evaluate quality, so intermediaries provide trust.
The intelligence economy has no equivalent. Trust is bilateral: you trust the lab. This won't scale. As providers multiply and capabilities converge, independent quality verification - the Moody's, the FDA of machine intelligence - will become essential. Whoever builds it occupies one of the most valuable positions in the economy.
Why does generic intelligence fail on specific problems?
Frontier models are general-purpose. The problems that matter are specific.
A tribunal judge in Leeds doesn't want a general analysis of unfair dismissal law. She wants to know how this employer's policies interact with this claimant's protected characteristics, in this region's case history. The gap between general reasoning and specific solutions is filled by context - proprietary data, institutional knowledge, regulatory nuance no general-purpose model possesses.
This is what Red Hat sold when IBM paid $34 billion. Not Linux. Knowledge of how to run Linux in your environment, with your compliance requirements, at your scale [2]. The context was the product. The commodity was the delivery mechanism.
What holds it all together?
Orchestration - the least glamorous of the four, and the most durable. Connecting intelligence to the systems, workflows, and decision structures where it creates value. APIs, compliance hooks, monitoring, change management. It's plumbing.
Nobody disrupts plumbing. Salesforce isn't a superior database. Its value is in the thousands of connectors that embed it into business operations. The technology is the entry point. The orchestration layer is the moat. Intelligence will follow the same pattern.
How should you invest?
Each of these scarcities is about making intelligence useful in context. Raw intelligence is becoming abundant. Useful intelligence - correctly applied, verified, contextualised, and integrated - remains scarce. Companies competing on raw intelligence are competing on a commodity. Companies competing on useful intelligence are building in the complement layer.
Map your firm against the four scarcities: judgment, trust, context, orchestration. Where are you strong? Where are you exposed? The answer determines whether AI is a threat to your pricing or a multiplier of your value.
Simon saw this in 1971. The scarce thing is never the abundant thing. It's whatever the abundant thing consumes.
Build there.
If you want to map your firm's position against these four scarcities, book a call with Lion Strategy.
Notes
[1] Simon, H.A., "Designing Organizations for an Information-Rich World," in Computers, Communications, and the Public Interest, ed. Martin Greenberger, Johns Hopkins University Press, 1971.
[2] IBM acquired Red Hat for $34 billion in July 2019.