Fund managers may face tougher AI scrutiny after Baillie Gifford letter

While many assume investor letters discussing Generative AI are simply about portfolio positioning, Baillie Gifford’s Q4 2025 release suggests a more profound shift. It implies that within 24 months, investor demand for 'Knightian uncertainty'-resilient AI strategies will pivot fund management margins from asset allocation expertise towards proprietary, explainable AI investment models.

The letter is less about AI stocks than buyer confidence

The obvious read is that this is another active-manager letter about finding companies positioned for Generative AI. That is the consensus version: markets are changing, some companies will benefit, and long-horizon stock pickers need patience.

But the source’s own framing points to a different procurement problem. If Generative AI makes uncertainty harder to quantify, the institutional buyer’s question changes from “which companies are in the portfolio?” to “why should this manager’s process be trusted when public narratives about AI are moving faster than traditional research cycles?”

That shift matters because active management is sold as judgment under uncertainty. Baillie Gifford’s summary uses the phrase “Knightian uncertainty,” not ordinary volatility, to describe the Generative AI backdrop.

The distinction is important: volatility can be modeled from distributions; Knightian uncertainty, as invoked here, refers to unknowns that cannot be cleanly assigned probabilities. A manager can defend underperformance by saying the opportunity set is still forming, but an allocator deciding whether to renew a mandate will still need evidence that the manager has a repeatable way to separate structural growth from fashionable exposure.

The missing evidence is inside the manager’s own workflow

The letter’s load-bearing omission is not a missing company name. It is the absence, in the supplied packet, of a detailed account of how Baillie Gifford’s own investment process changes when Generative AI is the uncertainty engine.

The summary says the team focuses on structural growth companies capable of navigating that environment; it does not say whether the manager is using proprietary AI models, new research workflows, different scenario methods, or explainable tools that an LP could inspect during diligence.

That omission is not a gotcha; an investor letter is not a systems architecture document. But it is where margin pressure can enter the fund-management business.

If every active manager can write that AI creates uncertainty, the scarce product becomes demonstrable process advantage. The fee justification moves from asset allocation language toward proof that the firm has a better way to ingest filings, technical roadmaps, customer evidence, and competitive signals without becoming hostage to the same public-market AI story everyone else reads.

The counter-read is that this is just stewardship language

The strongest counter-read is simple: this may be a conventional quarterly explanation for relative performance, dressed in the language of Generative AI. The supplied summary says Baillie Gifford acknowledges “disappointing relative performance in 2025,” and investor letters often use broad structural themes to explain why a long-term strategy still deserves time.

Under that reading, “Knightian uncertainty” is rhetoric for patience, not a signal that LP diligence or manager economics are about to change.

That objection cannot be dismissed from the packet. There are no corroborating publishers, no quoted client, no due-diligence questionnaire, and no fee schedule in the supplied material.

The evidence tier is marketing-blog-level fund communication, not a filing or regulator document. The right conclusion is therefore narrower than the hype version: the letter does not prove that AI-native investment processes are winning; it shows a sophisticated manager adopting a vocabulary that will invite harder questions from sophisticated buyers.

Procurement pressure starts with due-diligence language

The second-order consequence is likely to appear first in procurement paperwork, not portfolio construction. A pension fund, endowment, consultant, or wealth platform does not need to believe that an active manager has built a fully automated investment machine.

It only needs to decide that Generative AI has changed the risk of paying active fees for a process that cannot explain how it handles faster-moving information, synthetic content, model-driven consensus trades, and company claims about AI exposure.

That turns “AI strategy” from a portfolio theme into a manager-selection question. The under-noticed middle is the investment analyst and due-diligence staff inside asset owners and consultants.

Their work could become less about asking whether a portfolio owns AI beneficiaries and more about testing whether a manager’s research process is differentiated, explainable, and resilient to narrative crowding. For fund managers, the exposed group is not necessarily those that missed a single AI rally; it is those whose internal process sounds identical to competitors once the public holdings and commentary are stripped away.

Analysis: the margin shift is from conviction to proof

My thesis is that allocator demand for strategies resilient to “Knightian uncertainty” will push fund-management margins away from broad asset-allocation expertise and toward proprietary, explainable AI investment models. That is a forecast, not something Baillie Gifford states as fact.

The observable signals would be changes in institutional due-diligence questionnaires asking about proprietary AI capabilities, investor letters that describe internal research systems rather than only external AI opportunities, consultant commentary that distinguishes AI-themed portfolios from AI-enhanced investment processes, and active managers using explainability as part of their fee defense.

The thesis would be wrong if top-performing active managers gain assets without disclosing any distinctive AI-enabled process, if AI-themed active fees fall relative to passive alternatives, or if major institutional buyers keep treating AI as a sector-exposure topic rather than a manager-process topic. For now, the Baillie Gifford letter is a thin but useful signal: when a growth manager invokes “Knightian uncertainty,” the next procurement question is not whether AI creates investable companies.

It is whether the manager can show a process advantage that survives the same uncertainty it is asking clients to underwrite.