Summary
Alliance co-founder Imran Khan argues that as AI Agents become more active participants on the internet, crypto may finally gain a large and concrete source of demand. The opportunity would center on open systems for identity, payments, data and asset coordination, but major usability, security and regulatory questions remain unresolved.
The question crypto has been trying to answer
Crypto has long presented itself as a way to add an economic and ownership layer to the internet. Information can move across the web with relatively little friction, but identity, payments, data ownership and the transfer of value have generally remained dependent on centralized platforms and traditional financial institutions. Open protocols and programmable assets were intended to provide an alternative structure in which users could exercise more control over their assets and participate in the economics of digital networks.
In a recent essay, Alliance co-founder Imran Khan argues that this vision may have been missing a sufficiently strong source of demand. His thesis is that the internet is moving from a human-centered environment toward one in which AI Agents become increasingly important participants. If that shift continues, crypto infrastructure could encounter a class of users—or software entities—with needs that are different from those of ordinary consumers.
Khan does not reject Chris Dixon’s argument that decentralization and ownership are central to crypto. Instead, he suggests that the original picture was incomplete. For many years, the industry focused on building new forms of identity, markets, social graphs, storage and financial applications. Yet most people ultimately prioritized convenience over the principles that motivated crypto’s early communities. If a centralized platform offered a smoother product, users were often willing to entrust it with their data, relationships and economic activity.
That gap between ideological appeal and everyday usability has been one of the sector’s persistent problems. AI Agents, in Khan’s view, could change the demand side of the equation.
Why Agents create a different infrastructure problem
An AI Agent is not simply another human user with a faster interface. It is generally envisioned as software that can carry out a sequence of tasks, call multiple services, retrieve information, make decisions within defined limits and coordinate with other systems. Instead of waiting for a person to open an application and complete each step, an Agent may initiate interactions across the web on behalf of a user, company or another software process.
The source essay cites a 2026 disclosure from Cloudflare saying that automated traffic had, for the first time, surpassed human activity and accounted for approximately 57% of all web requests. Khan argues that a significant portion of this activity may be related to AI Agents. The figure and its attribution should be understood as the source’s characterization rather than a final accounting of all automated traffic. Automated requests can include crawlers, monitoring systems, software updates and other processes unrelated to generative AI.
Even so, the broader question is important. If internet activity increasingly involves software entities operating at machine scale, the account and service-access models designed for human users may become inefficient. A person can tolerate creating a separate account, accepting terms and linking a payment method for a limited number of services. An Agent that must interact with thousands of providers cannot easily establish a bespoke relationship with every counterparty.
That is where portability, open access, shared state and interoperability could acquire practical economic value. The argument is not that decentralization should be pursued for its own sake. Rather, when the number of machine interactions expands dramatically, common standards and reusable permissions may reduce coordination costs. An Agent may need to prove what it is allowed to do, access a service without a custom integration, and preserve a usable identity or reputation across different environments.
Why payments are returning to the infrastructure debate
Khan’s essay places particular emphasis on payments. The early web included HTTP status code 402, meaning “Payment Required,” as a placeholder for a future payment mechanism. Coinbase’s x402 takes its name from that status code and is presented in the essay as a reference to earlier efforts to make financial capabilities native to the internet.
Netscape reportedly worked with Visa on an attempt to integrate payments directly into the browser, but that effort did not become the standard foundation for online commerce. SSL helped establish the security infrastructure for modern web commerce, while money itself remained largely external to the underlying protocols. As digital commerce expanded, credit cards and platforms such as PayPal and Stripe became the primary ways online payments were processed.
The growth of AI Agents raises a similar question in a new form: can software entities access and use payment systems without relying on a human to complete every transaction step? An Agent may need to pay for data, computing resources, APIs or other digital services within a set of preapproved constraints. That requires more than a simple checkout screen. It requires machine-readable terms, verifiable authorization, spending limits, settlement rules and a clear way to assign responsibility when something goes wrong.
Crypto networks may offer several relevant building blocks, including programmable settlement, assets controlled through cryptographic keys and payment rules that can be executed by software. Open identity and authorization systems could also reduce dependence on a single platform’s account database. These capabilities are potentially relevant to institutional wallet and custody infrastructure as well, particularly where organizations need policy controls, audit trails and clearly separated permissions for automated processes. The need for those controls becomes greater, not smaller, when software can act continuously and at scale.
However, the existence of a technical capability does not demonstrate product-market fit. Any Agent-facing payment system would still need to address key recovery, fraud detection, privacy, sanctions and other compliance obligations, operational resilience and dispute resolution. A system that is technically open but difficult to supervise may not be acceptable to businesses or regulated institutions.
Complementarity rather than automatic convergence
Khan describes AI and crypto as complementary. AI could become a new large-scale demand source for open economic infrastructure, while crypto could offer AI Agents mechanisms for identity, payments, data access and asset coordination. The pairing is attractive because each side appears to address a weakness in the other. AI can produce more autonomous activity, but it needs reliable ways to interact with external systems. Crypto has proposed open economic rails, but it has struggled to attract ordinary users when the experience is less convenient than centralized alternatives.
Agents may have a different set of priorities. They could value standardized interfaces, portable permissions, predictable settlement and cross-platform compatibility more than human users do. An Agent does not necessarily prefer decentralization as a principle, but an open protocol may make it easier to switch providers, preserve authorization across services or coordinate with an unfamiliar counterparty.
That does not mean crypto will automatically become the default infrastructure. Centralized platforms can provide agent accounts, hosted permissions, proprietary APIs and internal settlement systems. For many use cases, those arrangements may be simpler, faster and easier to control. Crypto-based infrastructure will need to demonstrate advantages in reliability, cost, composability, transparency and compliance—not merely offer a different technical architecture.
The unresolved questions around scale and responsibility
The transition to Agent-driven activity also creates difficult questions about accountability. Who does an Agent represent? What decisions is it authorized to make? Who bears responsibility for an erroneous payment or an abusive request? How should a service provider distinguish a legitimate automated task from an attack? These issues span software design, contract terms, cybersecurity and regulation.
Open access without strong permissioning could increase the risk of automated abuse, data exposure and fraud. Excessive controls could, on the other hand, make cross-service coordination too cumbersome to be useful. The challenge is to create systems that are open enough to support interoperability while still providing identity assurance, policy enforcement and meaningful auditability.
The growth of automated traffic should also not be treated as evidence that crypto usage has already reached an inflection point. Even if a portion of web activity is generated by AI Agents, those Agents may continue to rely on centralized APIs, platform accounts and conventional payment processors. The transition from automated activity to crypto-native economic activity remains an empirical question.
AI Agents therefore represent a plausible new entry point for crypto demand, not a completed validation of the industry’s original thesis. Their rise brings the debate back to foundational questions about payments, ownership, identity and open protocols. The decisive test will not be whether Agents can technically connect to crypto networks. It will be whether those networks can provide a secure, auditable, compliant and sufficiently simple way for machines to coordinate at scale.
Source: link