How Much of Your Money Would You Let an AI Control?

摘要:How much money should you trust AI to control? New research into 6.4 million AI agent payments reveals how investors should view autonomous trading.

Imagine giving an AI assistant $10 to gather research for an investment decision. It could buy access to data as it works. Whether it should also be allowed to place the trade is a separate decision.

“Giving an agent permission to spend 40 cents is very different from allowing it to manage your portfolio,” says Zoomex.

That question of responsibility runs through BeInCryptos State of AI Agent Payments 2026. Between 23 July and 26 August, researchers tracked 6.4 million x402 payments across Base and Solana.

Together, they moved $119,947.

The average payment was about 1.9 cents. More than 90% were worth less than a penny.

Source: BeInCrypto ResearchWhat are AI Agents Really Buying Today?

A separate review of the x402 catalogue showed what these small purchases could support. Among the most-requested resources were web searches and checks for available airline reward seats.

Trading platform Zoomex expects transaction counts to grow faster than average payment sizes over the next year. Buying information gives agents a useful job with relatively little money exposed on each purchase.

Moving into investment decisions requires deciding how much freedom the software should have.

Across 19 products and workflows reviewed, 14 could act within permissions agreed in advance. Some actions could therefore proceed without approval each time.

The scope of that authority varied between products.

Zoomexs own AI connection provides a different level of delegation. Through its MCP service, which connects AI applications to exchange tools, assistants such as Claude or Cursor can examine markets and prepare orders.

The customer confirms each order before submission.

Source: BeInCrypto Research

“Our MCP enables AI assistants to retrieve market information, analyse positions and prepare trading instructions, while still keeping the user in control through confirmation before execution,” Fernando Lillo Aranda told the research team.

Zoomex separately documented capacity for up to 60,000 order requests per second across approved institutional account setups.

That figure describes permitted software traffic. It does not count actual trades or decisions made by AI.

The reports readiness assessment scored micropayments at 60.7 out of 100, and trading and investment agents at 39.1.

AI Agent Payments Readiness Across Different Financial Use Cases. Source: BeInCrypto Research

“Over the next twelve months, we expect more adoption in clearly defined tasks: rebalancing within agreed limits, managing collateral or executing an order under specific conditions. Fully autonomous investment decisions will take longer,” says Zoomex.

The companys proposed measure of progress puts the account owner at the centre:

“The key metric is not just how much money agents move. It is how much responsibility people are willing to delegate to them.”

免責聲明

本文觀點僅代表作者個人觀點,不構成本平台的投資建議,本平台不對文章信息準確性、完整性和及時性作出任何保證,亦不對因使用或信賴文章信息引發的任何損失承擔責任
上一篇

Tether 支持將 USDT 支付帶入外送與銀行服務的玻利維亞金融科技公司

下一篇

OCC 完成《GENIUS 法案》穩定幣發行商與託管規則