Summary
Alliance co-founder Imran Khan argues that the internet’s primary participants may increasingly shift from humans to AI agents, creating practical demand for crypto-native payments, digital ownership and programmable economic infrastructure.
AI may supply the missing demand
Crypto’s early vision for the internet was broader than digital tokens. It included user-owned assets, permissionless markets, portable identity and reputation, open social graphs, programmable finance and payment systems embedded directly into internet protocols. In a recent essay, Alliance co-founder Imran Khan argues that these ideas did not develop along the path many early advocates expected. The rise of AI agents, however, could create a new and more concrete source of demand for some of that infrastructure.
Khan’s central argument is that crypto may become more useful when the internet’s principal participants are no longer only humans. If software agents increasingly act on behalf of people and organizations, they may need to discover services, access data, negotiate tasks, make payments and coordinate with other agents. Those activities could expose limitations in today’s account-based and platform-dependent internet economy.
The argument is not simply that AI and blockchains are adjacent technology trends. Rather, it asks whether autonomous software will require new systems for identity, authorization, payment and ownership—and whether crypto networks are capable of providing them at useful scale.
From human users to machine participants
Most internet services were designed around human users. A person creates an account, accepts terms, navigates an interface, initiates an action and completes a payment. Platforms typically manage identity, permissions, transaction history and customer relationships within closed systems. Online commerce ultimately came to rely on cards and payment companies such as PayPal and Stripe, rather than on a native monetary layer built into the web itself.
Khan points to the history of HTTP status code 402, which indicates that payment is required. Coinbase’s x402 initiative takes its name from that earlier concept. He also recalls efforts during the Netscape era to integrate payment into the browser, including a reported collaboration with Visa. Those attempts did not establish an internet-native payment standard. The web became a highly effective commercial platform, but its payment infrastructure remained connected to traditional financial institutions and centralized intermediaries.
AI agents could change the design assumptions behind that model. Instead of waiting for a person to select every service and approve every step, an agent might search for information, call software tools, acquire computing resources, coordinate a workflow or commission a service. In a network with large numbers of such participants, agents may need to transact with services and with one another without establishing a bespoke commercial relationship with every counterparty.
Khan cites Cloudflare data indicating that automated traffic exceeded human activity in 2026 and represented approximately 57% of all web requests. That figure is presented in his essay and does not, by itself, establish what share of automated traffic is generated by AI agents. Still, the broader direction is significant: automated systems are already a major part of internet activity, and agentic software could make that role more active and economically meaningful.
Why open infrastructure could matter
For human users, the complexity of identity, payments and permissions can be hidden behind a familiar application interface. A platform can maintain an internal account ledger and make the process appear simple. For agents operating across thousands of services, however, the cost of creating separate accounts, credentials and payment relationships may become a material constraint.
Portable identity, machine-readable permissions, open payment conditions and shared transaction state could reduce that friction. An agent working for a company might be authorized to obtain data from several providers, compare the results, pay for specific API calls and distribute compensation to other participants. A human may experience this as a single outcome, while the underlying process could involve many small interactions between independent software systems.
This is where Khan sees a possible connection between AI and crypto. Crypto networks can represent ownership and authorization in software, settle transactions through programmable rules and allow assets or records to move across applications. Tokenization, decentralized identity, portable reputation and programmable finance are examples of capabilities that might become more relevant if agents need to operate across platform boundaries.
The economic case for openness may also be stronger at machine scale than at human scale. People often tolerate centralized platforms because they value convenience. A single company can provide a smooth interface, store user data and handle payments with little visible complexity. Agents, by contrast, may need to coordinate continuously across systems that do not share a common operator. Interoperability and portability could therefore have direct operational value, even for users who have little interest in decentralization as a political or philosophical principle.
That does not mean decentralized systems will automatically outperform centralized ones. Khan explicitly acknowledges that most people prioritize ease of use over platform ownership or architectural ideology. Any crypto-based agent infrastructure would need to be competitive on reliability, speed, cost, user experience and regulatory clarity. Open access alone is not enough to make a system useful.
Agent payments create a control problem
Once software can move assets or authorize payments, the key question is not only whether a transaction is technically possible. It is who granted the authority, what the authority covers, and how the action can be reviewed or reversed. An agent acting for an individual or institution might need access to a wallet, a payment account, a smart contract or an external service. Poorly defined permissions could expose users to model errors, prompt injection, malicious instructions, compromised credentials or failures at a third-party service.
This is where institutional wallet and custody infrastructure becomes relevant to the discussion. The important issue is not merely where assets are stored, but how agent actions are constrained and supervised. Enterprise systems may require transaction limits, allowlists, purpose-based permissions, multi-level approvals, time-bound credentials, policy engines and detailed audit trails. A firm would also need to establish which person or entity is accountable for an agent’s actions and how those actions fit within existing internal controls.
Compliance questions would be equally important. Agent-mediated payments could involve customer identification, sanctions screening, anti-money-laundering controls, data protection and recordkeeping obligations. The legal status of an agent’s instruction, the identity of the beneficial owner and the responsibility for an erroneous or unauthorized transaction would need to be addressed. These questions cannot be solved by cryptographic signatures alone.
Blockchain-based settlement instruments and other blockchain-based settlement instruments may feature in this debate because they can express payment conditions through software and support automated settlement. Their suitability, however, would depend on liquidity, network fees, confirmation times, counterparty exposure, legal classification and availability in the relevant jurisdiction. A system that is technically programmable is not necessarily suitable for every commercial or regulated use case.
Moving from narrative to evidence
The AI-agent thesis remains an emerging hypothesis rather than a broadly validated business model. It is not yet clear how much financial autonomy users will be willing to grant software, how often agents will transact independently, or whether blockchain-based identity and payment systems will offer a meaningful advantage over centralized APIs and account platforms.
The next phase will likely be judged less by broad claims about an autonomous economy and more by measurable operational outcomes. Can an agent safely obtain authorization? Can it make a bounded payment across multiple providers? Can institutions audit every action? Can users revoke permissions without creating new security risks? Can a system support these functions while meeting applicable legal and operational requirements?
For crypto, AI could provide the demand that earlier visions of open infrastructure lacked. But it could also force the industry to revisit its longstanding trade-offs between sovereignty and convenience, openness and control, programmability and safety. For the AI sector, crypto is not an automatic answer to the challenges of agent coordination; its infrastructure would have to demonstrate that it reduces real friction without introducing unacceptable risks.
If those conditions are met, AI agents could become a new class of users for payment, identity and digital-ownership systems. Crypto’s relevance would then depend less on persuading people to embrace decentralization in the abstract and more on whether open, programmable and auditable infrastructure can solve concrete problems in an increasingly machine-operated internet.
Source: link