AI Agents Are Unbanked. So Are 1.3 Billion People
The framing that caught serious attention recently came from a16z: AI agents are effectively “unbanked.” Non-human agents already outnumber human employees by roughly 100 to 1 in financial services, yet they cannot open accounts, prove their identity in a way that transfers across systems, or settle payments through standard rails. They execute tasks and move money without the trust infrastructure that banks require before letting anyone in.
The proposed fix is blockchain. Portable identity, programmable wallets, verifiable credentials, stablecoin settlement. a16z laid out five specific gaps that public blockchain infrastructure can fill: identity, payments, attestations, coordination, and reputation.
What makes this argument more interesting than another “blockchain solves X” pitch is what it implies about the other 1.3 billion actors with the same problem.
The parallel that rarely gets named
The World Bank’s 2025 Global Findex report put the number of adults without access to the formal financial system at 1.3 billion. More than half are concentrated in eight countries: Bangladesh, China, Egypt, India, Indonesia, Mexico, Nigeria, and Pakistan. Around 900 million of them own a mobile phone. More than 530 million have smartphones.
The reason most of these people are outside the system has nothing to do with the absence of economic activity. It is the absence of portable, verifiable identity: the kind traditional banking requires before it will authorize access. No documentation, no institutional history, no credit record. The system cannot see them, so it excludes them.
That is the same bottleneck a16z identifies for AI agents. Not similar. The same.
What blockchain actually provides here
The a16z framework maps five things that blockchain infrastructure can give autonomous agents. Each one applies equally to underserved humans.
Identity: a “Know Your Agent” framework using cryptographically signed credentials that tie an agent to its principal, permissions, and history. The same mechanism that lets an agent prove what it is authorized to do can prove who a person is, without requiring a physical document or a branch visit.
Payments: stablecoin rails that require no bank account to access. Stripe and Tempo’s MPP marketplace processed over 34,000 transactions in its first week with fees as low as $0.003 per transaction. That is not a theoretical settlement layer; it is already handling volume at a cost structure that makes small-value transfers viable.
Attestations: verifiable claims about capabilities, constraints, and history that any counterparty can check without a centralized authority issuing them. The same mechanism that lets a DeFi protocol trust an unknown wallet can let a merchant trust an unknown customer.
Coordination: shared rules and state that do not require a single institution to own the ledger. No bank in the middle, no API dependency, no relationship to maintain.
Reputation: transaction history tied to an address, verifiable across platforms. An agent builds a record. So does a person who has never had a credit score.
Where the two problems converge
The programmable wallet being designed for an autonomous agent is also the wallet that works for someone in rural Indonesia with a smartphone but no bank account. The stablecoin settlement layer that an agent uses to pay for API calls at $0.003 per transaction is also what makes cross-border payroll viable without a correspondent banking relationship.
This convergence is not accidental. It is a consequence of designing financial infrastructure from the protocol layer up, rather than from existing institutional relationships down. When you remove the assumption that a user needs a bank account, a physical address, or a government-issued ID to participate, you end up with infrastructure that is genuinely neutral.
What is being built right now
The agent economy is creating commercial demand for infrastructure that did not exist two years ago. BNB Chain’s ERC-8004 standard, deployed in early 2026, creates verifiable on-chain identities for AI agents. Coinbase’s Payments MCP, launched in January 2026, lets large language models access blockchain wallets and execute transactions directly. The AI agents market is projected to grow from $7.84 billion in 2025 to $52.62 billion by 2030.
Every piece of this infrastructure that gets built for agents is infrastructure that also works for underserved humans. The financial access problem has been framed for decades as a social challenge requiring institutional solutions. The agent economy is building a technical solution that happens to address the same structural gap from a different direction.
For teams building crypto products, the implication is worth sitting with. The agent infrastructure layer is being built now. The teams building it well are building something that scales beyond the agent use case, whether they intend to or not.