The Dependency Layer
"When intelligence is free, value lives in what intelligence cannot generate."
The Argument
AI models are commoditising intelligence. GPT, Claude, Gemini — they all get smarter, cheaper, faster. Intelligence is becoming a utility. Soon, it will be nearly free.
But AI agents — systems that act in the world, not just think — need things that intelligence alone cannot provide:
1. Trust
Who built this agent? Is it reliable? Who is liable when it fails? Trust cannot be generated — it must be earned and verified.
2. Identity
Is this agent who it claims to be? Is the human behind it verified? Identity cannot be self-asserted — it must be externally attested.
3. Physical Attestation
Did this actually happen in the real world? Can you cryptographically prove it? Physical reality cannot be simulated — it must be measured.
4. Governance
Who decides what this agent can do? Who audits it? Who shuts it down when it goes wrong? Governance cannot be automated — it must be designed.
These are the dependency layers. Intelligence depends on them, but cannot generate them. They must be built. The companies that build this infrastructure will be the most valuable companies of the agent economy — not the ones building more intelligence.
Why This Matters
$52.62B
AI agent market by 2030 (41% CAGR)
~3 in 4
enterprise agent projects never reach production*
$15T+
projected B2B digital commerce by 2028**
2026
first billion-dollar single-employee company (Amodei prediction)
The gap isn't intelligence — it's the dependency layer. The majority of enterprises can't get agents to production not because agents aren't smart enough, but because the trust, identity, and governance infrastructure doesn't exist yet.
*Based on patterns observed across 150+ enterprise engagements and corroborated by
Gartner Hype Cycle for AI and McKinsey State of AI reports on enterprise AI deployment gaps.
**Projected global B2B digital commerce volume; an increasing share of which will be
agent-initiated as autonomous procurement and supply chain agents reach production.
India's Unique Position
India built Aadhaar — digital identity for 1.3 billion people. India built UPI — real-time payments processing billions of transactions daily. India built the India Stack — the most ambitious digital public infrastructure project in human history.
India knows how to build trust infrastructure at civilization scale. The agent economy's trust layer — identity, attestation, governance for billions of AI agents — is India's to build. Not because India is cheap. Because India has done this before.
Further reading — the Dependency Layer cluster
Supporting essays that extend the dependency layer thesis into specific domains:
When Intelligence Is Free
The core economic consequence of commoditising intelligence — value migrates to what intelligence cannot generate.
India Will Build the Agent Trust Layer
Why India's Aadhaar/UPI track record makes it the natural home of the agent economy's trust infrastructure.
Why AI Agents Fail in Production
The practitioner truth behind the dependency layer — three in four enterprise agent projects never reach production.
15 Falsifiable Predictions
Specific, dated, falsifiable claims about how the agent economy unfolds through 2031.
Glossary: Dependency Layer
Canonical definition of the term and its sub-components.
The AI Agent Economy — Book 1
The full thesis, developed across ten chapters and fifteen predictions.
From the Book
"The AI Agent Economy" — Book 1 of an ongoing series. Specific, time-stamped predictions about the agent economy. The dependency layer thesis, fully developed. Subscribe to get chapters as they're written.