Franklin Templeton: Proxy AI will become the next core application of blockchain

Franklin Templeton executives: AI agents will become the core driver of blockchain network demand

Sandy Kaul, head of Franklin Templeton's digital assets and innovation department, recently said that AI agents are expected to become an important demand driver for blockchain networks-especially those protocols that can support fast and low-cost payments between machines. In a lengthy article released on Platform X on Wednesday, Kaul pointed out that the "agent-based AI" economy will require settlement speeds and fee structures that are difficult to achieve with traditional card payment systems. He believes blockchain ecosystems such as Aptos, Solana and BNB Chain are better able to adapt to this demand-oriented shift.

Core Points

Franklin Templeton's Sandy Kaul linked AI proxies to the growing demand for blockchain protocols, arguing that these protocols need to support micropayments between machines. Kaul pointed out that compared with the finality of blockchain transactions, traditional payment cards are not suitable for agent-based payments in terms of fees and settlement speed. A joint report by Visa and Artemis also believes that card-based payment infrastructure cannot meet the needs of AI agents for near-zero fees and fast settlements. According to the report, the x402 payment protocol developed by Coinbase has processed more than 109 million adjusted transactions since its launch in May 2025, with adjusted transaction volume reaching US$15 million.

How AI Agents Change Payment Requirements

The core argument is that proxy systems-software that can act autonomously on behalf of users or other systems-will create a very different form of business from today's human-driven transactions. Kaul sees this opportunity as an evolution of investors 'traditional approach to AI: Instead of focusing only on companies that are "AI-associated", he suggests that the market may also reward infrastructure designed to automate execution and continuous micro-interactions. In his view, if the payment layer cannot support high-frequency and small-amount transfers, it will become a bottleneck. Proxy micropayments are highly sensitive to time and cost, which means that even minor frictions-such as higher fees or slower settlements-can make repetitive machine payments economically unattractive.

Traditional card payment versus settlement speeds

Kaul's view is not that card payment networks are outdated, but rather that they are designed for different usage patterns: relatively low-frequency human transactions, where settlement delays are rarely a major constraint. He emphasized that Visa network settlement may take one to three working days, while certain blockchain networks can complete transactions in seconds. This time gap is likely to become important as agents continue to coordinate work, as delays can affect the entire workflow and reduce the feasibility of quick settlements. Kaul also pointed out that "high fees and settlement times" are key factors making traditional payment tracks unsuitable for agent-based micropayments.

Visa and Artemis: Infrastructure gaps in "agent" commerce

The Franklin Templeton executive's views are consistent with a joint report released last Wednesday by Visa and investment theory platform Artemis. The report believes that traditional card payment systems built for human-scale payments are not designed for the needs of AI agents. Visa and Artemis particularly emphasized that proxy payments require infrastructure with near-zero fees and faster settlements to make micropayments commercially viable. The report's framework strengthens Kaul's argument that the real area of competition lies in payment throughput and cost efficiency, not just application-level AI capabilities. It is worth noting that this is not a purely conceptual idea; the report also points to existing machine payment experiments and early adoption signals, including x402-related activities.

Status of "Early Adoption": Activity Data for x402

In the Visa-Artemis report, the x402 payment protocol was highlighted as an example of a machine payment track with measurable usage. The report said that since its launch in May 2025, x402 developed by Coinbase has processed more than 109 million adjusted transactions, with adjusted transaction volume reaching US$15 million. For investors and builders, the value of this statistic lies not in a specific number, but in the direction it points: machine payment protocols are beginning to attract usage within a framework designed for frequent transfers. However, it is worth noting that this indicator is reported in the form of "adjusted transaction volume" and "adjusted transactions", so readers should regard it as an operational indicator in the report rather than directly translating into end-user revenue or broader market share.

Signals from payment service providers

Although Visa-Artemis's analysis criticizes card-based infrastructure as insufficient to meet agent-type needs, these companies are also actively exploring how the broader payment ecosystem can support agent-type behavior. According to reports from the same background, Visa's cryptocurrency-related arm and Stripe-backed Tempo launched AI tools in March. Visa's products are described as supporting same-day payments-an effort to address the speed limits on which agent-based micropayments rely. At the same time, Kaul's remarks directed readers to the blockchain environment where settlement speeds are structurally faster, implying a practical mismatch: Even if card payment providers add features to speed up payments, their fees and settlement models may still not be able to match the economics of high-frequency machine-to-machine transactions.

Looking ahead, the key is to see whether agent-based payment demand will be realized in a way that drives the continued use of a low-fee, fast-settlement track-especially as protocols such as x402 and newer types of infrastructure compete to serve repetitive micropayment streams. An open question is how quickly mainstream agent deployments need to reach sufficient scale to turn settlement and fee limits from theoretical issues into decisive factors.