The Munich-based bank told Reuters that it is the first bank in Europe to open its platform to major AI assistants. Its new “Agentic Investing” service allows customers to connect their Scalable accounts to supported AI agents through the Model Context Protocol, or MCP. The move puts AI directly into the trading workflow.

Scalable Capital Chief Product Officer Alexander Siepp told Fortune the company sees the integration as a way for investors to begin their financial “client journey” inside an AI assistant and complete it through Scalable’s regulated banking infrastructure.

“It certainly creates a level-playing field,” he said. “Access to information and to compute and to intelligence is now available—really—literally in your pocket, 24/7.”

But Siepp noted it may take time to become fully integrated with Scalable’s client base.

“Maybe not for all client segments at the same speed,” Siepp said. “But we clearly see potential that there is a large group of clients that would be interested in doing it.”

Scalable Capital, founded in 2014, has grown into one of Europe’s largest digital investment platforms. The company also told Reuters it has more than €60 billion in client assets and more than 1 million customers, primarily in Germany and Austria, with operations also expanding across countries including Italy, Spain, France and the Netherlands. 

The company’s roots are in making investing more accessible through digital brokerage and wealth-management services. Scalable has also expanded its trading offering to include derivatives, and said in July that it would offer more than 1.8 million derivatives from seven major issuers.

Siepp said that the program can fundamentally change how investors interact with their brokerage accounts. A customer could, for example, ask an AI assistant to identify stocks that have fallen for consecutive months and monitor them. The assistant could then prepare an order based on the user’s instructions.

“It’s just a few prompts,” he said.

Is AI a reliable investor?

But the effectiveness of AI when given the reins for making trades remains to be seen. A June research report from Elm Wealth offers a mixed answer. Researchers Jerry Bell, Victor Haghani and James White tested Claude, ChatGPT, Gemini and Grok in a “Crystal Ball Challenge,” giving the models historical Wall Street Journal front pages with market-moving information but withholding the actual market outcomes.

The experiment found that Claude and ChatGPT were relatively strong at predicting market direction. Across roughly 200 sessions, Claude beat human players in 76% of sessions, while ChatGPT beat them in 63%. Gemini did so in 43% of sessions and Grok in 51%. But the models had a major weakness, the study found.

Elm found the AI systems generally took too much risk relative to the trade context. The researchers distinguished between two investment decisions—what to invest in and how much to invest. AI performed relatively well on the first factor but poorly on the second. The study found the models understood concepts such as the Kelly criterion and Merton share in theory but struggled to apply appropriate risk management when making actual simulated trading decisions.

“The US stock market has moved by over 5% on 23 days and by over 9% on seven days since the year 2000,” the study read. “Given average position sizing in stocks of 7x to 12x across the AIs, we think they were taking too much risk of a catastrophic loss of capital, given none of them had (or could reasonably expect to have) super high hit ratios.”

OpenAI, Anthropic, Google and xAI did not immediately respond to a request for comment from Fortune about using AI assistants for financial trading.

But Scalable is not handing over unrestricted control of customer accounts to AI. The company said users must approve trades and savings plans before they are executed, and its current system does not allow AI assistants to make payments or withdraw money from Scalable accounts. Siepp said the AI connection follows the same core security protocols as Scalable’s existing applications, including strong customer authentication.

“Whether you end up finding the holy grail together with your AI assistant on high returns and low risks or not, I think that remains to be seen,” he said.

He also stressed that the arrangement is not a formal partnership between the German bank and OpenAI or Anthropic. Instead, Scalable is using MCP—an open technology for connecting AI systems to external tools and services.

“We are using available resources, technologies, the MCP—the Model Context Protocol,” Siepp said. “This is really a standalone offering that comes only from Scalable users, ready-made technology integrations offered by those AI assistants.”