The Economics of AGI

Andrew koh, alex imas, & shane legg.

GoogleDeepMind 2026 https://institute.deepmind.com/essays/the-economics-of-intelligence/

“Smarter AI will increase productivity but real economic transformation hinges on agents learning to coordinate.”

Is AGI really going to transform the economy as fast as the benchmarks suggest? 🧠📈

A landmark essay from Google DeepMind (“The Economics of Intelligence”) offers a compelling reality check on why exponential AI benchmarks haven’t immediately translated into macroeconomic productivity gains.

Here are three key takeaways from the research:

1️⃣ The Individual Model is the Wrong Benchmark: Standard tests evaluate AI agents in isolation. But in the real world, value is generated by organizations through “mosaic intelligence”—pairing specialists whose complementary strengths cover each other’s blind spots. An AI can beat a solo human, but still lose to a well-coordinated human team.

2️⃣ Competence vs. Coordination: Breakthroughs in individual skill (math, coding, writing) smoothly adjust wages and prices. But real structural disruption—market concentration, firm sizes, and labor shifts—only happens when AI achieves supercoordination across agents.

3️⃣ Scaling Without Frictions: Human teams hit hard growth limits due to communication bottlenecks and misaligned incentives. Multi-agent systems, however, can pass entire context windows and codebases instantaneously—rewriting the classical economics of firm size once agent-to-agent coordination is solved.

#ArtificialIntelligence #economicsofai #artificialgeneralintelligence #GoogleDeepMind #AIAgents

Discover why solo AI benchmarks miss the big picture and how agent coordination thresholds reshape markets, wages, and organizations.