The curve of the agentic web: What the last 10 years tell us about the next 5

Unlocking the Agentic Web: How AI Payments are Revolutionizing Automation and Commerce

Co-founder and CEO of Circuit & Chisel, Louis Amira, opened his talk by referencing Marc Andreessen’s 2011 assertion that “software is eating the world.” Mr. Amira highlighted a crucial limitation: despite its intelligence, software has never possessed its own wallet. For a decade, AI could perform complex tasks like image recognition, novel writing, and cancer diagnosis, yet it remained dependent on humans for basic transactions, such as buying a cup of coffee.

Mr. Amira emphasized that real-world actions and transactions have historically been human-controlled. However, this paradigm is shifting as AI payments and their underlying infrastructure are now being developed. Drawing on his 15 years of experience at the intersection of AI and payments, first at Google and then at Stripe, he shared insights into the evolution of these fields.

At Google, Mr. Amira studied under Ray Kurzweil, witnessing the company’s growing investment in AI, culminating in the acquisition of DeepMind. He recalled using a DeepMind API that was remarkably superior to existing discovery tools. Later, at Stripe, as the first external crypto hire, he questioned conventional payment methods, encountering initial resistance to both AI/ML and real-time crypto payments.

Significant turning points occurred at both companies. For Google, it was AlphaGo’s famous move 37, demonstrating AI’s intuitive capabilities. At Stripe, a demo built by Mr. Amira’s co-founder prompted John Collison to recognize it as the future of money movement, leading to Stripe’s acquisition of Bridge. These events underscored the transformative potential of both AI and advanced payment systems.

Mr. Amira’s personal “aha!” moment connecting AI and crypto came after leaving Stripe in late 2024. He observed Large Language Models (LLMs) evolving from simple question-answerers to potential economic actors. In 2025, agents learned to browse, negotiate, and research, but the true breakthrough was when they transacted independently. His co-founder demonstrated this by giving an LLM a crypto wallet and instructing it to pay its own bill, which it successfully executed.

This pioneering demonstration was soon followed by major industry players. Coinbase launched its X402 protocol, and within months, Google, Stripe, Visa, Mastercard, and others introduced similar solutions. Mr. Amira noted that while his team might not have been the absolute first, they were among the first to bet on the idea that these transactions would eventually cost almost nothing, thereby unlocking an economy of sub-$5 transactions.

Many companies are investing heavily in this space, anticipating that it could add trillions of dollars through previously unviable 2-cent and 20-cent transactions. Mr. Amira predicted three major transformations in agent transactions over the next five years: “do it with me” (agents assisting humans), “do it for me” (agents executing tasks autonomously), and “work together to figure it out” (agents trading with other agents).

The “work together” stage, where agents buy compute, purchase data, and hire sub-agents, will unlock a new market, enabling more complex and economically viable tasks. McKinsey estimates agent-to-agent commerce could reach $3 to $5 trillion by 2030, a figure Mr. Amira believes is conservative. This would make agent-to-agent transactions the fastest-growing peaceful economic movement in history.

Crucially, none of this progress is possible without a free, neutral, and unbiased payments layer. Unlike traditional variable fees, which average around 3%, the new fixed-fee model costs less than a penny per transaction. This shift has profound implications for individuals, akin to the evolution from human travel agents to self-service platforms like Expedia, and now back to AI agents that work for a fraction of the cost, 24/7, constantly seeking better deals.

A decade ago, machines learned to interpret images, play games, and translate. A year ago, they learned to think. Today, they are learning to transact. Mr. Amira emphasized that the next five years are not a spectator sport; the agentic web requires builders, developers creating agents, enterprises deploying them, and protocols ensuring seamless collaboration. He encouraged attendees to visit ATXP.AI to deploy an agent and experience the future of autonomous transactions firsthand.