In five years, AI may take over part of the transactions.

Written by: Boaz Sobrado

Translated by: Baihua Blockchain

Chandler Fang left Ripple to create a layer of verification and anti-fraud infrastructure to determine whether AI agents can be trusted to spend money on your behalf. Ripple has just cast a support vote for this initiative with a $5 million investment.

“When I decided to deposit $2 million into a JPMorgan Chase account, I never felt anxious, nervous, or worried.” Chandler Fang said on the podcast "On The Margin." Just before that, he had actually done it—walking into a bank branch with a portion of investors' money. But for him, the crypto world was a completely different feeling. “When I try to transfer $2 million into a crypto wallet, or maybe a cold wallet, I feel nervous to some extent. I have to check twice, and then check three times.” he said.

Fang left Ripple because of this disparity to start a business around a customer base that barely existed a few years ago: AI agents that can spend money for you. According to The Block, his startup t54 Labs has completed a $5 million seed round of financing, led by Anagram, PL Capital, and Franklin Templeton, with strategic support from Ripple, Virtuals Ventures, and Blockchain Coinvestors. The company was founded in early 2025 in San Francisco, claiming to be the "trust layer of the agent economy."

“The entire society is a trust business.” Fang said, “To some extent, you need to build your own reputation and ensure that the people around you trust your expertise, trust your experience, and trust your moral standards.”

No one has really built the infrastructure for this problem

Imagine you let ChatGPT help you buy a cheeseburger. This agent must choose the right merchant, select the right order, and pick the right price, and if any step goes wrong, someone will pay the financial price.

“It might run off and, I don’t know, buy a burger at Cheesecake Factory, where there’s actually no such thing.” Fang said. Then there’s the merchant's problem: the merchant suddenly receives a request from software claiming to act on behalf of someone, but this merchant has never seen it before.

“From the merchant's perspective, their first reaction might be: Who are you? Who do you represent? You say you represent Ivan, but how do you prove that?” Fang said. If no one can verify, then the actual human account holder will later raise a dispute. “Someone moved my money and bought something I absolutely did not need. I want my money back. All of this leads to losses.” he said.

These issues were not a problem in the past because no one had ever designed a payment system for "non-human entities." “You quickly realize that our entire financial ecosystem is fundamentally built around humans.” Fang said, “Its foundational premise is: decisions are made by people, services are provided by people, and contacts and interactions are completed by people.”

Fang said that t54 is not a payment company and is not an agency company. Funds will still flow through Visa, Ripple, or other rails. t54 sells that outer layer: it’s an identity system called 'know your agent', a real-time engine that marks suspicious activities before funds move, credit lines for agents, and products named x402-secure and XRPL x402 facilitator. The company states that it has processed over 20 million transaction records.

A crowded, noisy, and mostly still immature space

The bet on “AI agents + money” is already very crowded, and many things still do not work as of now.

Louis—a co-founder of a payments startup supported by Stripe—said on the podcast “On The Margin” that most things in this space are just “performances.” “We’ve recently seen a lot of promotional videos from so-called agent credit card companies, but not a single one of them actually works,” he said. “You’ll find that all these promotional videos get cut off before anything actually gets purchased.”

Nitya Subramanian, who runs a wallet infrastructure company called Para, also stated on the podcast “On The Margin” that delegation itself always carries costs. “An agent is fundamentally outsourcing a purchase behavior, and anyone who has outsourced purchasing knows that it necessarily comes with trade-offs.” she said. Her solution is to set hard limits. “I can create a card backed by stablecoins, give it $200 a week, and let it buy Chipotle. This way, the only thing it’s allowed to do every day is to buy that bowl of Chipotle for me.” she said.

In her view, this opportunity is still worth fighting for. “The last time we had a new financial rail was probably in the ’70s with credit cards.” Subramanian said, “If you are working in fintech or crypto, this period might be the most exciting build time in many of our—if not most of our—careers.”

Why Ripple and Franklin Templeton wrote the checks

One of the most valuable companies in the crypto industry and one of the largest asset management institutions in the world typically wouldn’t fund a startup that is still on Day 0. The reason Fang was able to secure this money is that he has been in this system for many years: first as a quantitative trader at JPMorgan Chase, then joining the bank’s blockchain payments team Onyx around 2019, and later joining Ripple to work on cross-border payments based on blockchain and stablecoins.

“I worked at Ripple for a long time, so there was already some mutual trust between us.” Fang said. Before leaving, he had already pitched this idea internally at the company. “It’s so obvious that we need some kind of financial infrastructure that enables a large number of agents to participate in touching money, so that they can perform more valuable tasks for us.” he said.

In his understanding, investing in such a company is cheaper than building one from scratch. Supporting “a fast-moving, risk-taking, and highly innovative company,” he said, is “a reasonable decision” for a mature institution.

But finance will not reward anyone who rushes in without caution. “If you want to hold anyone's money, you must have multiple licenses, such as a money transfer license, to enable the money to flow.” he said. t54 is currently employing a compliance consultant who has 18 years of experience within the SEC, FDIC, and FINRA systems.

Moreover, Fang has little patience for founders who believe that “the concept itself is the hardest part.” “It's not the idea. Ideas are cheap.” he said, “90% of successful startups ultimately do something different from their original plan.”

A five-year cycle bet

“You can imagine that in five years, perhaps 30% of the transactions that happen every day are automatically completed by AI agents.” Fang said. He envisions that you might have your car automatically replenish toilet paper, or have an agent automatically buy a birthday gift under $200 for your parents every year.

All of these transactions essentially point to the trust he felt in front of the JPMorgan Chase counter; this time, however, the trust object has become software. “Are you willing to trust them?” Fang said, “When you walk into a JPMorgan Chase, do you have the same feeling towards them as I did at that time?”

JPMorgan Chase has taken centuries to establish that almost instinctive response of trust. Fang is betting that the agent economy will not give him several centuries of time.