Ken Griffin once brushed aside the frenzy around artificial intelligence. He called much of the talk garbage. Then something inside his own hedge fund changed his mind.
At Goldman Sachs’ Apex Symposium this summer, the Citadel founder described an agentic system that his firm built. It took on the kind of assignment that used to tie up teams of young finance researchers with master’s and doctoral degrees for six to eight weeks. The task involved reading an academic paper, rebuilding its methods from scratch, confirming the published findings, and checking whether those findings survived on fresh data outside the original sample. The AI agent finished the entire sequence in two to three hours.
“AI handled a form of research that requires coding, data work, statistical validation, skepticism, and iteration.” Griffin said those words at the symposium, according to The Motley Fool. The speed stunned him. So did the quality.
Earlier this year Griffin had sounded far less optimistic. In May he told an audience that after watching advanced research tasks fall to AI he went home one Friday “fairly depressed.” The technology had become “profoundly more powerful” in just months, he explained. It opened use cases Citadel could not pursue before. The shift hit him hard because the work displaced went beyond routine coding. It struck at roles that demand advanced training.
“This is not just a white collar job. This is a master’s or PhD level job,” Griffin remarked in clips that circulated widely on X in recent days. He drew a sharp line between modest gains seen in software engineering, estimated at 15 to 25 percent, and the far steeper leap in analytical research. The latter compresses months into hours. That difference matters in a business where a single validated insight can generate outsized returns.
Yet Griffin refused to stop at the downside. He now predicts AI will trigger a surge in new business formation. Tiny teams equipped with these agents can tackle market research, generate code, test hypotheses, handle customer support, and manage operations. Barriers collapse. Costs plunge. Speed accelerates to levels he calls breathtaking.
“If a tiny team can use AI agents to research markets, write code, validate assumptions, support customers, and run operations, the cost and time required to launch a company falls sharply,” he said, as reported by The Motley Fool. Entrepreneurs gain the power to challenge established players that once relied on large staffs and deep pockets. Some traditional advantages erode when expertise turns easier to obtain through computation.
But not every edge disappears. Firms that control massive computing resources, exclusive datasets, or extensive distribution channels may grow even tougher to dislodge. Griffin sees a new divide taking shape. Productivity for skilled professionals rises dramatically. Success, however, still hinges on judgment, data ownership, and the ability to convert raw answers into viable products and loyal customers faster than rivals can respond.
The comments come at a moment when Citadel itself has moved aggressively in AI-related markets. In late July the firm bought the bulk of a stock portfolio from Leopold Aschenbrenner’s Situational Awareness hedge fund after that shop suffered heavy losses on AI bets. Citadel acquired the shares at a discount greater than 10 percent. Its stock-focused fund returned about 14 percent in July, according to The Wall Street Journal. The flagship Wellington fund rose roughly 6 percent. A tactical trading vehicle gained 11 percent. The transaction underscored both the volatility in AI investments and Griffin’s willingness to pounce when prices detach from fundamentals.
Back in January at Davos, Griffin had sounded dismissive. He questioned whether the promised returns could ever justify the hundreds of billions pouring into data centers. By spring his tone had flipped after internal experiments at Citadel delivered results no one anticipated. “Work that we would usually do with people with masters and PhDs in finance over the course of weeks or months is being done by AI agents over the course of hours or days,” he told Stanford students, per reports that resurfaced this month on LinkedIn and X.
Analysts and investors have seized on the remarks. Recent posts on X from traders and commentators highlight the same theme: agentic systems hand small groups capabilities once reserved for entire departments. One widely viewed thread noted that while talent remains vital, raw compute access now forms the primary moat. Low-margin operations may struggle to absorb the infrastructure expense. The conversation has expanded beyond productivity to questions of concentration. Who owns the clusters? Who controls the chips?
Griffin himself has warned of tail risks tied to geopolitics. In the Goldman Sachs interview he placed an 8 percent hit to U.S. gross domestic product on any serious disruption to access to advanced semiconductors from Taiwan. That assessment, shared in clips circulating this week, ties the AI story directly to national security and supply-chain fragility.
Even so, his dominant message stays bullish on creation over destruction. “There’s no reduction to headcount at Citadel on the back of this breakthrough,” he stated in one of the recent video excerpts. The firm continues to hire talent aggressively. The new tools simply let the existing staff pursue far more ideas than before. “I’ll take every single productivity gain I can get. With the talented people we have, we just have more to go after.”
That stance echoes his broader view of economic history. Periods of technological compression have often favored nimble entrants over slow-moving giants. Griffin believes the current wave could prove more potent than past examples because it attacks the cost of knowledge itself. Replicating a rigorous research process no longer demands dozens of high-paid specialists working in parallel for months. A handful of people plus the right agents can iterate at a pace that leaves legacy organizations flat-footed.
Critics point out that current AI systems still require human oversight for final validation, especially in regulated financial contexts. Hallucinations, subtle methodological errors, and the absence of true causal understanding remain real constraints. Griffin has acknowledged these limits in past remarks, stressing the need for skepticism even when the machine produces plausible output. Yet the trajectory he observes inside Citadel suggests those gaps are closing faster than many outsiders appreciate.
Fortune captured the pivot in stark terms. Months after labeling AI hype as overblown, Griffin warned the technology would reshape society in ways that left him personally unsettled. The catalyst was not some distant lab demonstration. It was performance measured in his own trading rooms and research benches. Fortune reported that the internal breakthroughs convinced him the productivity curve had bent sharply upward.
Industry observers now watch to see whether other large hedge funds disclose comparable internal advances. Quantitative shops have long invested in machine learning for alpha generation. The difference this time lies in the scope. Rather than narrow optimization of existing signals, the new agents appear capable of end-to-end scientific replication across finance literature. The volume of testable hypotheses could multiply by orders of magnitude.
Entrepreneurs outside finance have started to take note. Startup formation data remains noisy, but anecdotal evidence from venture conversations suggests founders are experimenting with AI agents to compress product development cycles in ways that mirror Citadel’s research workflow. If Griffin’s forecast holds, the next wave of challengers to incumbent banks, asset managers, and technology providers may arrive with far smaller headcounts and far faster iteration than their predecessors.
The implications stretch beyond individual firms. Labor markets for highly educated workers could face pressure even as overall economic output expands. Griffin has urged a culture of lifelong learning as one partial answer. Those who treat AI as a collaborator rather than a replacement may thrive. Those who wait for the disruption to pass risk being left behind.
Of course, hype cycles have misled before. Data-center spending has reached staggering levels on the bet that scaling laws will continue to deliver gains. Griffin himself once questioned the economic logic. His reversal, grounded in observable results at one of the world’s most successful hedge funds, carries unusual weight.
So the question now facing executives, investors, and policy makers is simple. If six to eight weeks of expert labor collapses into a single afternoon, what else becomes possible? And who will capture the value?
Griffin has placed his bet. The age of the lean, AI-augmented founder may be closer than the skeptics think. The machines are already proving it inside Citadel’s walls.