AI Capital Allocation and Governance

A Portfolio Model
James E. Murphy · ORCID 0009-0001-2217-4306 · Working draft for discussion · v0.70 · July 2026

A model for allocating AI investment as a portfolio: steering capital to its highest return, subject to an agreed risk limit.

Most AI governance asks whether a system is safe enough to run. This paper asks whether a workload is worth funding, how much risk it adds to the firm’s AI portfolio, and whether the whole book stays inside a limit the firm has agreed to carry. It sets out a model that treats AI activity as a portfolio of distinct investments, each valued against a named financial driver, costed on a fully burdened basis, and judged by its marginal, correlation-aware contribution to a risk budget grounded in the board’s appetite and the firm’s capacity to absorb loss, with an absolute floor no expected return can buy through. The portfolio framing is now widely urged; the machinery underneath it is the contribution.

Download the working draft · PDF · 43 pages · v0.70 Also on SSRN · abstract and mirror of this draft

Status

This is a living draft, developed in the open. Comments and criticism are welcome: contact@jamesemurphy.com.

Cite as

James E. Murphy, “AI Capital Allocation and Governance: A Portfolio Model,” working draft v0.70, July 2026. Available at https://jamesemurphy.com/ai-capital-allocation/portfolio-model/ and https://ssrn.com/abstract=7095978.