Jefferies Warns AI Can Trap Investors in Stale Theses, Compressing 6-18 Month Cycles Into Weeks
Updated
Updated · Financial Times · Jun 9
Jefferies Warns AI Can Trap Investors in Stale Theses, Compressing 6-18 Month Cycles Into Weeks
1 articles · Updated · Financial Times · Jun 9
Summary
Jefferies’ chemicals team said AI-augmented quant research can speed narrative investing from genesis to adoption in weeks, but a “thesis lock-in” flaw can keep bad ideas alive after they should have expired.
LLM-based systems tend to reuse confirming sources, reinterpret negative evidence and avoid relabeling old data, the bank said, making them weak at judging whether a cycle is ending or merely pausing.
Jefferies mapped six phases—Genesis, Validation, Adoption, Euphoria, First Cracks and Exhaustion—and argued AI makes euphoria shorter but more violent while causing first cracks to barely register.
Using Aristotle’s Poetics as a framework for narrative structure, the team said AI is useful for tracking trends within an existing frame, but human analysts remain better at recognizing when the frame itself must change.
The bank said the cost of arriving late is rising: investors not in by Phase 2, Validation, risk becoming exit liquidity for AI-augmented funds.