The Self Driving Portfolio

How Multi Agent AI Could Rebuild Institutional Asset Management from the Investment Policy Statement Up

Institutional asset management has spent decades trying to solve the wrong bottleneck. Data became cheaper. Compute became faster. Models became more sophisticated. Yet the true constraint never disappeared. The hard ceiling sat with human bandwidth. A CIO can only supervise a limited number of teams with real depth. An analyst can only track a finite number of securities before signal quality starts to decay. Investment committees still gather in episodic meetings, absorb thick decks, debate under time pressure, and often deliver decisions after markets have already moved. That is the backdrop behind The Self Driving Portfolio: Agentic Architecture for Institutional Asset Management by Andrew Ang, Nazym Azimbayev, and Andrey Kim. The paper argues that autonomous, tool using, LLM powered agents can shift institutional investing from a labor constrained workflow into a scalable system of parallel reasoning, structured debate, and governed delegation.