Institutional
Citadel's Ken Griffin called AI "all garbage" months ago. Now he says it is doing PhD-team work in hours instead of months
The Citadel founder's reversal is less about AI hype than about what happened when work his firm used to hand to teams of finance PhDs, tasks that took weeks or months, started coming back from AI agents in hours or days instead.
Ken Griffin runs Citadel, a roughly $68 billion multi-strategy fund that has compounded at about 19.2 percent annually after fees since 1990 on the back of 260 in-house PhDs and around 100 petabytes of data processed daily, which makes his change of view worth more scrutiny than the average executive's AI conversion story. Griffin's starting position was blunt dismissal. At a panel earlier this year he described an AI-generated report a colleague showed him this way: the first few sentences read as genuinely insightful, and then everything below that was garbage. That is not a vague scepticism about hype, it is a specific, technical complaint about output quality falling apart past the surface layer, from someone whose firm has the infrastructure to actually stress-test the claim.
What changed his mind was not a demo or a benchmark, it was watching Citadel's own workflow shift under him. Work that the firm used to route to teams holding master's degrees and PhDs in finance, analysis that previously took weeks or months to turn around, is now coming back from AI agents in hours or days. Griffin has said he went home one Friday actually fairly depressed after registering what that meant, which is a more interesting data point than the eventual optimism, because it reads as someone updating on capability he had personally underestimated rather than someone talking his book.
The more useful part of Griffin's account for anyone trying to separate signal from noise in this space is his critique of how other executives talk about AI. He asked business leaders to share their AI transformation stories and got four or five accounts of real productivity gains. When he dug into what had actually driven them, not one involved AI in the sense people meant, the improvements traced back to data optimisation and plain digitisation that had been mislabelled. His point, that the nuance between AI specifically and technology broadly gets lost in the C-suite, is a useful filter to apply to almost every other AI-adoption claim currently circulating in finance, including some of the more triumphant hedge fund performance narratives making the rounds this year.
The caveat is that Griffin's read is still a single, well-resourced practitioner's account of internal workflow change at one firm with unusually deep infrastructure and talent to make AI tooling work well, not evidence that the same gains generalise to smaller shops running thinner data pipelines and less rigorous validation. His own critique of mislabelled digitisation cuts both ways: without seeing Citadel's internal benchmarks, there is no independent way to confirm the hours-not-months claim is genuinely AI-driven rather than partly the same kind of process cleanup he says fooled other executives. Read it as a credible practitioner signal that the capability curve moved faster than a sophisticated sceptic expected, not as a case study anyone can safely copy without their own infrastructure to match.
Read the original: The Motley Fool - Citadel's Ken Griffin called AI "all garbage" months ago. Now he says it is doing PhD-team work in hours instead of months. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.