Named AI Clarity, the strategy introduces the Poesis Dual Mandate Equity approach to pursuing equity upside with downside protection in bear markets

The strategy has been trading with proprietary capital since August 2026 and plans to accept external capital in early 2027

Poesis, the AI-native asset manager founded by former Capital Group partner Alex Popa and former Goldman Sachs Global Head of Machine Learning Charles Elkan, today announced its plan to open AI Clarity, its flagship agentic institutional equity strategy, to external capital in early 2027.

AI Clarity introduces Dual Mandate Equity, Poesis's investment approach with two objectives: participating in equity-market upside while seeking to materially reduce losses in market downturns.

AI Clarity is an agentic equity strategy focused on U.S. listed large-cap companies. Its operating model puts machine intelligence under human command. The CIO and team are responsible for setting the strategy, risk framework and parameters within which agents conduct research and operate. Agents generate trade proposals by using quantitative models and tools, with each trade subject to human review and approval by the investment team before execution.

The same research, forecasting and risk process is designed to be applied systematically across market cycles and sectors. Agents are tasked to validate incoming data used by models. Machine intelligence informs stock selection and a dynamic beta overlay is designed to reduce market exposure in stressed markets and increase it in stronger ones.

“At Capital Group, I learned the value of a scalable, repeatable investment process and well-defined risk controls,” said Alex Popa, founder, CEO and CIO of Poesis. “AI now lets us apply that discipline across the full investment cycle with greater consistency, while keeping judgment and accountability with people.”

This fall, Poesis partnered with ASUS and NVIDIA to run a one-week live agentic investing experiment, separate from AI Clarity. Running on a different system with separate capital, autonomous AI agents stayed on track while conducting investment research, managing risk and executing trades within predefined mandates. [Link]

Popa spent more than 15 years at Capital Group, where he became a partner and served as a portfolio manager for The Growth Fund of America, focused on U.S. large-cap equities. He previously served as an economic advisor to Romania’s Prime Minister and as Secretary of State for Foreign Investments. He earned a Harvard MBA as a Baker Scholar, after graduating first in his engineering class at Stanford.

Elkan, Poesis’s Chief Scientist, was Managing Director and Global Head of Machine Learning at Goldman Sachs. He also led machine learning across e-commerce and cloud at Amazon, where he became the first Amazon Fellow. He is a longtime professor of computer science at UC San Diego, holds a PhD from Cornell, and built AI Clarity’s original signals.

About Poesis

Poesis is an AI-native asset manager developing agentic systems that research, manage risk and trade, deploying investment strategies in live markets. Poesis is based in Menlo Park, California and was founded by Alex Popa, previously a partner and portfolio manager at Capital Group, and Charles Elkan, former global head of machine learning at Goldman Sachs.

Compliance disclaimer

An investment in the Poesis strategy is speculative and involves substantial risk, including possible loss of the entire investment. There is no assurance that the strategy will achieve its objectives, outperform a benchmark, or avoid losses. Risks include equity-market, issuer, concentration, liquidity, derivatives, leverage, counterparty, operational, cybersecurity, and regulatory risks.

Risk controls, diversification, and human oversight cannot eliminate losses. De-risking may occur after losses, miss recoveries, or generate losses in reversing markets. Historical correlations and downside characteristics may not persist. Artificial intelligence and quantitative models may rely on inaccurate data, produce erroneous outputs, overfit historical relationships, or fail as market conditions change. Testing and review may not detect these failures. Planned capabilities are not currently implemented unless expressly identified otherwise.

View source version on businesswire.com: https://www.businesswire.com/news/home/20261109779700/en/