Why AI governance funding is consistently underprioritised
AI governance investment competes for budget against AI capability investment — and capability investment consistently wins. The reason is how the investment cases are framed. AI capability investment — a new model, a new tool, a new capability — has a visible, describable return: this much efficiency improvement, this much revenue uplift, this much customer experience enhancement. AI governance investment is framed as cost avoidance: spend this much now to avoid spending more later on something that may never happen.
Cost avoidance arguments fail in budget competition because they require the decision-maker to believe both that the avoided cost is real and that the probability of incurring it is significant. Both beliefs are easier to dismiss than they should be, particularly for governance failures that feel hypothetical to executives whose organisations have not yet experienced one.
The solution is not to make the cost avoidance argument more persuasively. It is to change the framing entirely. AI governance investment is not cost avoidance — it is risk-adjusted return on investment. The analysis is explicit: here are the specific scenarios we are trying to reduce the probability of, here is what each scenario would cost if it occurred, here is the probability reduction that the governance investment achieves, and here is the net present value of that probability reduction. This is capital allocation analysis, not compliance budgeting, and it is evaluated differently.
The three-scenario model
Every organisation's AI governance investment case should be built around three scenarios — not comprehensive risk catalogues, but three specific, credible scenarios that represent the material risks for that organisation. The scenarios should be specific enough to be costed: not "a data protection enforcement action" but "an enforcement action by [specific regulator] for [specific AI use case] resulting in [penalty range] plus [remediation cost]".
For a financial services firm using AI in credit decisions, the three scenarios might be: a discriminatory lending enforcement action (ACMA/FCA/CFPB), a data protection enforcement action for GDPR Article 22 violations, and a class action from customers denied credit on the basis of a flawed AI model. Each scenario has a cost estimate (penalties, legal, remediation, compensation) and a probability estimate. The governance investment reduces the probability of each scenario. The expected value of that probability reduction, over a governance investment horizon, is the return on the governance investment.
