← Back to insights

FinOpsValue RealizationExecutive Advisory

Building an IRR model for enterprise AI investments

A defensible business-case template for AI initiatives that survives CFO scrutiny and post-implementation review.

Why most AI business cases collapse on review

I have seen hundreds of AI business cases. The two most common failures are:

  • Run cost (inference, observability, retraining) modelled as a small percentage of build cost. In production, run cost frequently exceeds build cost within eighteen months.
  • Revenue uplift modelled as a single point estimate with no probability weighting.

What a defensible model looks like

A clean enterprise AI business case has four sheets:

  1. Cost: build, run, change-management, decommissioning of legacy.
  2. Benefit: three scenarios (low/base/high) with explicit probabilities.
  3. Risk-adjusted IRR: discount rate elevated by a documented risk premium.
  4. Kill-switch: the leading indicator and threshold that triggers a stop decision.

This last sheet is the one most teams skip. It is also the one that earns CFO trust.