Will AI Replace Fund Managers?

The question is framed as a prediction. It shouldn’t be. The evidence is already in.

89.5% of active fund managers underperform their benchmark over 15 years. Zero of 22 equity categories beat their index net of fees. Their selling decisions are measurably worse than random selection — not occasionally, not in bad years, but structurally, across all experience levels, with no improvement over careers. (SPIVA Year-End 2024; Akbas et al., Journal of Finance 2023.)

$60 trillion in capital is allocated based on this record.

The case for AI replacing fund managers is not a technology thesis about what will be possible. It is an investment thesis about what the evidence already demands.

What Is Actually Happening

The financial industry has already automated most of what it does. Ninety percent of trade execution is algorithmic. Market making, index replication, options pricing, risk monitoring — all systematised. The human layer was peeled back everywhere it could be automated without threatening the people at the top of the fee structure.

The research and alpha-generation layer — where active managers claim their edge, and where they charge 2% and 20% — has not been automated. That is where the $60 trillion sits.

This is not because it cannot be automated. It is because it is the most complex part, and because the people responsible for it control the organisations making the decision.

An AI-native fund is not a fund that automated the easy parts. It is a fund that started with the hard part — removing the human investment decision entirely — and built everything else around that premise.

Why Human Judgment Fails Systematically

The underperformance of active managers is not a distribution problem. It is not the case that some managers are excellent and drag down the average. It is that the distribution of human investment outcomes, at scale, produces chronic underperformance across virtually every category measured.

The failure modes are well-documented:

Career risk distorts position sizing. Fund managers hold positions that are safe to hold, not positions that are best to hold. Losing money in a stock everyone else owned is acceptable. Losing money in a contrarian position is career-ending. This produces systematic herding that destroys alpha.

Recency bias dominates. Human pattern recognition overweights recent events. Markets that have gone up feel safe to be long. Markets that have fallen feel dangerous. The cognitive pull is toward the current environment rather than the base rate.

Attachment to positions corrupts selling. The evidence from Akbas et al. is striking: buying decisions in institutional portfolios show genuine skill. But selling decisions are worse than random selection, losing an estimated 50 basis points per month compared to selling at random. Managers hold positions they should exit because they have publicly defended them, built models around them, presented them to investors.

Organisational politics compound everything. In a firm with a CIO and an investment committee, the best investment decisions are not the ones the evidence supports — they are the ones that survive internal debate.

You cannot hire your way out of these failure modes. They are not personal flaws. They are features of human cognition under the conditions that professional fund management creates.

What AI Actually Changes

An AI-native systematic manager removes each of these failure modes structurally:

No career risk — positions are sized according to validated signals and portfolio risk constraints, not according to what is defensible to a committee.

No recency bias — the evidence is weighted across long market history, not across the current quarter’s narrative.

No attachment to positions — exit decisions are made by the same rules that made entry decisions, applied continuously without the emotional cost of admitting a mistake.

No organisational politics — there is no committee. There is a validation framework, and what passes it gets deployed.

This is not an incremental improvement on the human fund management model. It is the removal of the structural constraints that produce chronic underperformance.

The Transition Is Already Underway

The question is not whether this transition happens. It is how long it takes.

$60 trillion moves slowly. Institutional capital has multi-year investment processes, consultant relationships, career risk at the allocator level that mirrors career risk at the fund manager level. The same cognitive biases that make fund managers underperform make allocators slow to move capital away from underperforming managers.

But the pressure is building. Passive indexing — the simplest possible expression of “human stock-picking doesn’t add value” — now accounts for roughly half of all equity fund assets in the US. Every year that passes without active management delivering on its performance claims accelerates the move to systematic approaches.

The funds that will capture the next phase of this transition are not the ones that added AI to existing human structures. They are the ones built, from scratch, as AI-native systems — where the removal of human judgment is the design principle, not an aspiration.

---

AvB Capital is an AI-native systematic manager. We did not add AI to a human fund. We built a fund for AI. Track record available to qualified investors on request.