Two Wall Street veterans launch a stock fund run by AI agents

Two Wall Street veterans are asking outside investors to put money into a stock fund where software agents do the research and propose the trades. The people still have to press the button.
Who is behind it
Poesis is a Menlo Park, California, firm founded by Alex Popa and Charles Elkan. Popa, 46, spent more than 15 years at Capital Group, where he became a partner and a portfolio manager on The Growth Fund of America, one of the largest actively managed funds in the world. Before that he was an economic adviser to Romania's prime minister and the country's secretary of state for foreign investment.
Elkan, 63, is the chief scientist. He was a managing director and global head of machine learning at Goldman Sachs, and before that led machine learning for e-commerce and cloud computing at one of the biggest US tech companies. He is a longtime professor of computer science at UC San Diego and, according to the firm, built the strategy's original signals.
The name, the firm's website says, comes from the Latin word for poetry and the Greek word for creation.
How the fund works
The first strategy is called AI Clarity, and the firm announced on November 10 that it will open to outside investors. Poesis describes an operating model in which the chief investment officer and team set the strategy, the risk framework and the parameters, and AI agents do the research, check the data feeding the models and generate trade proposals.
In practice, according to Bloomberg News, AI agents running quantitative models research US-listed large-cap companies, pick 80 to 100 of them and decide when to buy and sell, with a target holding period of about three years. When the agents detect a likely downturn, they hedge by shorting a stock index. Poesis calls this its "Dual Mandate Equity" approach: take part in market gains while trying to lose materially less in falls.
Every trade the agents propose is reviewed and approved by the investment team before it is executed. A "dynamic beta" overlay is meant to cut market exposure in stressed conditions and raise it in stronger ones.
The strategy has run on the founders' own money since August 2026 and plans to accept external capital early next year. Fees are the hedge-fund classic: 1% of assets a year plus 20% of gains.
The bigger plan
Popa and Elkan intend to add strategies and eventually build a multistrategy fund for institutions, people familiar with the plans told Bloomberg. The pitch is cost. Those people said the pair could run billions of dollars with fewer than 100 employees, far below the staffing of giants such as Citadel and Millennium Management.
The firm has already tested how far the machines can go. In a separate one-week experiment, run on separate capital, teams of agents ran two strategies with no human sign-off at all: a concentrated buy-and-hold portfolio in the style of Warren Buffett and a fast-trading, market-neutral quant fund. Both stayed within their mandates, Bloomberg reported.
On performance, there is almost nothing to judge yet. People familiar with the fund told Bloomberg that AI Clarity fell 0.3% in October, when the S&P 500 dropped almost 1%. That is one month. The firm's own risk disclosure, FinanceX notes, warns that its de-risking can trigger after losses and miss the recovery, and that models can overfit history.
Our take
Strip away the agents and this is a familiar product: a long-biased large-cap fund with a hedge overlay, charging hedge-fund fees. The novelty is the cost structure. If research, screening and trade ideas can be done by software, a firm can cover 100 stocks with a team of ten. That is a real threat to the economics of the big multistrategy shops, which pay armies of analysts and portfolio managers.
The open question is whether the agents add anything a good quant model did not already do. Quant funds have used machine learning for decades. What is new is language models that can read filings and transcripts like an analyst. That is useful. It is also exactly the kind of skill every competitor can now buy.
The human sign-off matters more than it sounds. It keeps the firm inside the rules on fiduciary duty, and it gives investors someone to call. It also means the bottleneck is still people.
For investors, the right stance is patience. A 1% and 20% fee needs years of beating the market after costs to justify itself, and AI Clarity has months. Watch what it does in a real drawdown before believing the downside protection.
This is not investment advice.
Sources
- Poesis, AI-native asset manager Poesis plans to open flagship agentic equity strategy to external investors, Business Wire (November 10, 2026)
- Poesis, firm website and founder biographies
- Bloomberg News, Goldman, Capital Group alumni debut AI-run stock-picking fund, via Financial Advisor (November 10, 2026)
- FinanceX, Poesis to open its agentic equity strategy to investors in 2027 (November 2026)
- Hedgeweek, Goldman Sachs and Capital Group veterans launch AI-driven hedge fund (November 2026)







