AI Is Everywhere in Investment Research — Except in Asset Allocation
Every layer of the investment process is changing. Except the one that matters most.
Asset allocation is the boardroom at the top of the building. AI has entered through the front door, but it is still working its way up the floors.
Almost every recent article about AI in investing focuses on research, tools or workflow. Almost none address allocation itself. That gap is not accidental. It reflects how technology moves through investment organisations: first into instruments, then into process, and only later into policy. The boardroom is always last.
Governance first, models second
The CFA Institute report Explainable AI in Finance identifies a problem every allocator recognises immediately. AI models can be powerful, but investment committees will not use them if they cannot explain the output. Transparency is not a technical preference. It is a governance requirement.
For allocators this is not abstract. Strategic Asset Allocation, ALM and risk budgeting depend on models that must survive scrutiny from boards, regulators and clients. A model that cannot be explained will not survive an investment committee, regardless of its predictive accuracy. Explainability is the price of entry.
What AI cannot carry
The Human Advantage in an AI World (Asia Asset Management) argues that the real edge in an AI-driven world may still be human judgement. Not because humans process data better. They do not. But because investment decisions involve context, responsibility and accountability that cannot be cleanly delegated.
This is precisely where allocation differs from the rest of the investment process. Security selection can be automated. Rebalancing rules can be codified. But a policy decision about how much risk to take, against which liabilities, under which assumptions about the future, requires someone to own the answer. AI can inform that decision. It cannot carry it.
The risk hiding in the workflow
A Forbes piece by the Finance Council takes the broadest view: AI as both opportunity and threat. Firms are deploying AI for research, monitoring, portfolio analytics and client reporting. The same technology introduces model risk, operational risk and governance challenges.
For asset allocators, this tension is familiar. Every new model changes the workflow before it changes the portfolio. The danger is not that AI will get allocation wrong. The danger is that it reshapes the surrounding process so gradually that allocation assumptions go unexamined through an entire ALM cycle, across multiple policy reviews, before anyone notices the ground has shifted.
The door at the top
Taken together, the message is consistent and slightly uncomfortable.
AI adoption inside investment management is accelerating. The change is real, the tools are improving and the workflow is shifting. But allocation itself remains largely untouched. Not because it will stay that way, but because organisations move policy last: after the tools are proven, after the research is trusted, after the committees are ready.
Models move first. Processes follow. Policy waits.
AI has entered the building. It has improved the research, sharpened the tools and restructured the workflow. Floor by floor, it is working its way up. But the boardroom, where allocation policy is made, defended and owned, remains untouched. Not forever. Just last. And in investment management, last is not a detail. Last is a decision.