AI trading agents move closer as brokers test 24/7 investing

AI trading agents are starting to move from an investing thought experiment into products that brokerages, startups and some retail investors are testing now. For everyday investors, the shift could mean software that watches a portfolio around the clock, suggests moves and, in some cases, carries out investing tasks that used to require a person at the keyboard.

CNBC reported that firms are building tools around “agentic trading,” a term for artificial intelligence that can do more than answer questions or produce research. In investing, that means an AI system may eventually be able to monitor accounts, connect actions to a user’s goals and execute parts of an investment strategy.

Devin Ryan, head of financial technology research at Citizens, told CNBC the idea is approaching faster than a distant technology cycle. He said the model could give individuals something closer to a personal finance team that works continuously, including while they sleep.

How would AI trading agents work?

An AI trading agent is software designed to take instructions, evaluate financial information and complete defined investing tasks. A cautious version might summarize a portfolio and recommend changes, while a more advanced version could rebalance holdings or place trades after a user approves the plan.

Ryan said he expects these systems to grow beyond stock trading. He told CNBC that future agents could help manage taxes, cash balances, borrowing, mortgages and portfolios based on a person’s financial goals. Fully hands-off investing is still being developed, and companies are taking different approaches to how much power the tools should have.

Podium Markets AI, a startup focused on investing tools, has built an assistant called Ivy. According to Dirk Mueller-Ingrand, the company’s co-founder and CEO, Ivy reviews a customer’s holdings across multiple brokerage accounts and produces recommendations tied to goals and risk tolerance. The user still chooses whether to act and must place the trade.

Mueller-Ingrand told CNBC that the company sees AI as an adviser that informs the investor, rather than a system that replaces the investor’s final call.

Larger platforms are also testing the category. CNBC reported that Robinhood introduced tools in May that let third-party AI agents connect with customer accounts. Public, the brokerage firm, is building its own AI agents to automate investing workflows inside its platform.

Leif Abraham, Public’s co-founder and co-CEO, told CNBC that agentic systems could move brokerage apps beyond research and manual trading by letting AI agents execute investment strategies for users.

Why brokers care about agentic trading

For brokerages, the commercial appeal is activity. Ryan estimated to CNBC that agentic finance could lift transaction volumes by at least ten times. He said an investor who now trades about twice a month could eventually make 20 trades a day under an agent-driven setup.

Ryan also said he expects agents to account for most transactions by number of trades on some platforms by the end of next year. That is his estimate, not a confirmed industry outcome.

Retail investors have already been experimenting with general-purpose AI tools since ChatGPT became widely used in late 2022. CNBC reported that investors have used ChatGPT and Anthropic’s Claude to read earnings reports, research companies and generate stock ideas, with mixed results.

Obioha Okereke, a 29-year-old technology consultant in Georgia and founder of College Money Habits, told CNBC he built a Claude-based agent to look for undervalued stocks and options opportunities. He said he reviewed each suggestion before trading and views AI as a tool rather than a replacement.

Thomas Schlossmacher, a 31-year-old retail investor whose company Specialty Tokens builds AI systems for businesses, told CNBC he tested a trading agent after seeing claims that AI could find profitable market patterns. He said the system kept losing money and cautioned against blindly asking an agent to make money.

What guardrails are companies adding?

The central risk is translation. If a user asks an agent to grow a portfolio aggressively, the system still has to understand whether that means more volatility, fewer holdings, options trading or a higher risk of losses.

Public requires customers to review and approve an agent’s workflow before it completes investing tasks, according to CNBC. Abraham said the user keeps the final say and that the AI will execute only within approved instructions.

Ryan said firms also have to make sure customer interests remain central. If an agent behaves differently than expected, he told CNBC, that becomes a risk for the company offering it.

This story draws on original reporting from CNBC.