Sapiom raises $15M so AI agents can start spending your money

Imagine your virtual assistant doesn't just write code, but also buys access to a cloud server, pays for an API and renews a subscription to a service it needs. Until recently this was a technical dead end: an AI has no passport and no credit card, and it can't pass the "I'm not a robot" CAPTCHA (ironic, given that it is one). The startup Sapiom decided it was time to hand algorithms a checkbook, and raised $15 million for it with backing from heavyweights at Accel. It is an important step toward what we call the agent economy, in which programs become full participants in the market.

We have drifted into the era of AI agents: programs that don't just answer questions but carry out complex chains of actions. But as soon as an agent leaves the cozy sandbox of its developer, it hits a solid wall. The modern internet was built by people, for people: everywhere you need logins, passwords and two-factor authentication by SMS. If you ask an AI to book a hotel or buy data to train a model, you will still have to enter your card details by hand. That kills the very idea of autonomy and turns us into a "gasket" between two digital interfaces.

Technically, Sapiom's solution works as an abstraction layer between the AI and the outside world. The startup takes on the dirty work: authorization and processing of microtransactions. Investors at Accel are betting that within the next couple of years the number of transactions made by machines will exceed human traffic several times over. This is not just another payment gateway; it is a complex trust system. If an agent suddenly decides to burn your entire monthly budget on dubious tokens or endless queries to an expensive model, the system must stop it. Sapiom promises flexible limits and controls so that digital freedom doesn't turn into a financial hole.

Why does this matter right now? Because the language model market is oversaturated, and the competition among the giants is shifting toward "action." OpenAI, Anthropic and Google are racing to build the bot that best controls a computer and a browser. But without the ability to pay legally, those bots stay locked in a cage of free demos. Sapiom is building rails for a new economy in which the participants will be not only people and companies but also autonomous pieces of code. That fundamentally changes how software is consumed: why would you hold a personal subscription to an expensive graphics editor if your agent can rent it for five minutes to complete one specific task?

Of course, this raises a huge set of security and ethics questions. Giving artificial intelligence direct access to money sounds to many like the start of a cyberpunk horror story. But the alternative is stagnation. If we truly want agents to save us time, we will have to delegate to them the right to manage resources. Sapiom is trying to make that process transparent and controllable. We will probably soon see a whole class of "banks for bots" emerge, and the current investment is only the first installment on the foundation of a new financial reality.

The bottom line: AI agents are ceasing to be mere chatbots and becoming economic actors. If Sapiom solves the problem of trust and authorization, we will see explosive growth in autonomous services that run entirely without human involvement. Are you ready to trust your bank account to an algorithm if it promises to do all the work for you?