AI and the Future of Payments

The rapid development of artificial intelligence could significantly reshape how payments work on the internet. Increasingly, experts are discussing a future in which many transactions are performed not by humans, but by autonomous software programs known as AI agents.

According to Brian Armstrong, the founder of Coinbase, the number of AI agents making payments could eventually surpass the number of humans doing so. Changpeng Zhao, former CEO of Binance, has gone even further, suggesting that automated programs might one day execute up to a million times more transactions than internet users.

One of the main reasons lies in how traditional banking systems operate. Software cannot easily open a bank account because banks require identity verification. Cryptocurrency wallets, however, work differently. They can be created using only a private key, without the need for identity checks or KYC procedures.

Micropayments are another challenge for traditional financial systems. While performing a single task, an AI agent may need to access multiple online services such as real-time data, computing power, or specialized AI models. Each of these operations may cost only a fraction of a cent.

The problem is that card-based payment systems were never designed for such small transactions. For example, the minimum fee for a single transaction on Stripe is around $0.30, which makes handling micropayments economically inefficient.

This is why new solutions are emerging. Coinbase is developing a protocol called x402, which allows stablecoin payments to be embedded directly into HTTP requests. In practice, this means that an AI agent can automatically pay for access to a service and immediately continue its task without any human involvement.

Such systems could be used across many industries. In healthcare, an AI agent could pay for access to medical records. In logistics, it could automatically choose and pay for transportation services. In media, it could purchase access to individual articles or datasets.

For now, however, the scale remains relatively small. The x402 protocol currently processes about $28,000 in daily volume, and part of that activity consists of testing transactions.

At the same time, traditional financial companies are developing their own approaches. Visa has introduced the Trusted Agent Protocol, while Mastercard has completed the first bank payment executed by an AI agent within Santander’s system.

The most likely outcome is a divided market. Traditional commerce will probably remain within card-based payment networks, while machine-to-machine micropayments may move toward stablecoins, which are better suited for extremely small and frequent transactions.