Ken Griffin Says AI Agents Now Do PhD-Level Finance Work in Days
Overview
What this article covers
A summary of Citadel CEO Ken Griffin’s shift from AI skepticism to recognizing AI’s powerful impact on finance and society, and the implications for high‑skill finance work.
Who it’s for
Readers interested in finance, hedge‑fund operations, or the broader business impact of artificial intelligence—especially those unfamiliar with Griffin’s recent statements.
Prerequisites
None required, though a basic understanding of hedge‑fund terminology (e.g., “alpha,” “research throughput”) helps.
Core Concepts
| Term | Explanation (as used in the sources) |
|---|---|
| AI skepticism | Griffin’s earlier view that AI was “all garbage” when examined beyond surface‑level demos (January 2026, Davos panel). |
| Agentic AI | Autonomous AI agents that can perform complex, high‑skill tasks previously done by people with master’s or PhD credentials. |
| Productivity step‑change | Griffin’s description of a rapid increase in AI capability over the past nine months, making the technology “profoundly more powerful.” |
| High‑skill finance work | Research and analysis tasks that traditionally require advanced academic training and take weeks or months to complete. |
| Adaptability | Griffin’s message that lifelong learning will become the key career advantage as AI reshapes work. |
Detailed Explanation
1. Griffin’s Early Skepticism
-
In January 2026 at the World Economic Forum in Davos, Griffin called AI “impressive on the surface” but “all garbage” once examined more deeply.
-
He warned that hype around AI (e.g., predictions of massive job loss) was being used to justify data‑center spending.
2. The Turning Point
-
A few months later, during a conversation with Stanford Business School professors, Griffin described a personal reckoning: “I went home one Friday, actually fairly depressed” after seeing AI’s potential impact on society.
-
He noted that AI had become “profoundly more powerful” than nine months earlier, allowing Citadel to “unleash” a broader range of use cases.
3. From Software Gains to Research Disruption
-
Software engineering: Griffin cited productivity boosts of roughly 15‑25 % from AI‑assisted coding tools.
-
Financial research: More strikingly, tasks that previously required teams with master’s or PhD qualifications and took weeks or months are now completed by AI agents in hours or days. He called this “extraordinarily high‑skilled jobs being… automated by agentic AI.”
4. Operational Impact at Citadel
-
In December 2025, Citadel CTO Umesh Subramanian said at the Reuters Next conference that the firm had rolled out the Citadel AI Assistant, a chatbot trained on licensed third-party content including transcripts, regulatory filings, brokerage research reports, and Citadel's own investment strategies.
-
The tool surfaces risks and generates customized research and reading lists based on an investor's portfolio, producing relevant material in a fraction of the time a human researcher would need. It is now used regularly by nearly all of Citadel's equities investors.
-
Subramanian said investment judgment still remains in human hands: "I don't think just by using AI you're going to become a much better investor." Separately, at a JPMorgan conference in October 2025, Griffin said generative AI is not helping hedge funds produce market-beating returns and has yet to meaningfully impact the industry.
5. Broader Implications
-
Griffin said the automation he described affects "extraordinarily high-skilled jobs," not mid-tier white-collar roles, and that watching it inside Citadel left him "fairly depressed" about the potential impact on society.
-
Griffin said career success will depend on being a lifelong learner: in his words, AI "will just make this all the more important."
Common Questions
Q: Is the AI impact limited to low‑skill, clerical tasks?
A: No. Griffin highlights that AI is now handling “extraordinarily high‑skilled jobs” previously performed by highly trained finance professionals.
Q: How does Griffin quantify AI’s productivity gains in software engineering?
A: He mentions improvements of about 15‑25 % in software development speed.
Q: What is Griffin’s advice to workers?
A: He stresses that career success will depend on being a “lifelong learner” because AI will make adaptability crucial.
Q: Did Griffin’s view change completely?
A: His stance shifted clearly between January and May 2026, from calling AI "all garbage" beneath the surface at Davos in January to saying at Stanford in May that AI had become "profoundly more powerful" and was automating high-skilled finance work. That said, his January skepticism was specifically about hype outpacing results: at Davos he argued large language models, while powerful, were still far from being deployed at scale across companies. His May remarks, as reported, did not explicitly revisit that point.
All information in this article is drawn directly from the provided source highlights.