German bank Scalable Capital has opened its investment platform to outside AI assistants, allowing customers to connect ChatGPT, Claude, or Grok to live brokerage accounts. This move, known as “Agentic Investing,” lets AI analyze portfolios, calculate savings plans, and adjust limit orders, with trades requiring explicit customer confirmation. The feature, launched on Aug. 25, works without any in‑house assistant and keeps deposits and withdrawals confined to Scalable’s own app and website, ensuring that the core flow of funds remains under the bank’s direct control.
Co‑CEO Erik Podzuweit told Reuters that this is a first step towards deeper AI integration, and the feature runs through the Model Context Protocol, an open technical standard for AI systems connecting to outside services. He emphasized that future updates will embed AI more tightly inside Scalable’s own app, extending the same plain‑language interaction model to a broader set of functions. The German neobroker, which holds more than €60 billion for over 1 million customers, thus becomes the first bank in Europe to expose a live brokerage environment to third‑party large‑language models.
U.S. Brokers’ Approaches to AI Integration
U.S. brokers are taking different paths to integrating AI. Robinhood launched Agentic Trading and an Agentic Credit Card on May 27, letting an outside AI agent connect to a dedicated sub-account with its own budget and place real trades, while still allowing the agent to view a customer’s full portfolio across every account they hold. Interactive Brokers introduced Ask IBKR in October, an AI tool that answers client questions about their own portfolio rather than generating trading instructions or linking to external models like Claude or ChatGPT. Public went further by building its own agent from scratch, becoming the first brokerage to introduce AI agents for investing and enabling customers to set standing instructions that let an AI execute trades, move cash, and manage risk automatically inside Public’s app; the same platform also supports a direct Claude Desktop connection via the Model Context Protocol, spanning stocks, ETFs, options and cryptocurrencies.
These launches target power users, who represent 10% of consumers and 19% of millennials, using AI for 27 or more distinct tasks a month, including investment management. According to PYMNTS Intelligence, power users’ average monthly task count was 25 in September and 27 in December.
Related: OpenAI ends Cursor deal over Musk distrust
This trend signals a shift in AI integration in retail investing. As AI capabilities and adoption grow, more brokerages are exploring ways to incorporate AI into their services. The varying approaches by U.S. brokers suggest a competitive setting where each firm is seeking to differentiate itself through unique AI offerings. Mainstream users, by contrast, average around eight tasks a month and tend toward lower-complexity activities such as comparison shopping.
For instance, Robinhood’s dedication of a separate sub-account for AI agents provides an extra layer of security, isolating budgetary exposure while still granting the agent a panoramic view of the client’s holdings. Public’s AI agents can act across multiple asset classes, automatically reallocating between equities, exchange-traded funds, options contracts, and digital currencies based on pre-set risk parameters. Interactive Brokers’ advisory-only approach maintains a safe distance from direct AI-driven trades, positioning the tool as a conversational aide rather than an execution engine.
As AI assistants become more prevalent and capable, it’s likely that we’ll see more innovation and competition in AI-driven investment services. The potential for AI to democratize finance and make investing more accessible is vast, and the race to harness this potential is on.
