Executive Summary
An AI proposal can promise a credible improvement and still depend on commitments no executive has secured.
For leadership teams preparing 2027 budgets, an AI proposal may promise a credible operational improvement yet depend on commitments no executive has secured. The benefit may require changes to staffing, workflows, data or customer delivery across several functions. A budget review needs to establish whether the organization can make those changes and who has the authority to resolve competing priorities.
We recommend that each material request identify the business result, the executive accountable for it, the full cost and the evidence required for further funding. Ask affected leaders to confirm their commitments and examine any conflicting accounts of progress. Where uncertainty remains, fund a test only if resolving it could support a worthwhile investment.
What to require in each funding request
Ask each team for a one-page decision summary, supported by the underlying analysis. Finance should review the economics, and the business owner should confirm the changes needed to deliver the result.
| Requirement | Evidence needed for the decision |
|---|---|
| Business result and owner | The outcome, baseline, target and measurement date; the executive accountable for delivering it. |
| Benefit and financial effect | The evidence behind the estimate and the action that converts an operational improvement into savings, additional margin or another agreed outcome. |
| Full cost and timing | Implementation, integration, data preparation, licenses, usage, training, human review and support, with a downside case. |
| Organizational changes | The changes to roles, workflows, incentives and decision rights, and who must approve them. |
| Dependencies and safeguards | The data, technology and shared services required, their owners and readiness. |
| Funding and review conditions | The amount requested now, the total expected commitment and the next decision date. |
| Results to date | For renewals, 2026 year-to-date cost and results compared with the original case. |
Match the funding decision to the evidence
An experiment, a shared data investment and a proven application require different approval conditions. Applying the same immediate profit test to all three can exclude worthwhile preparatory work. Funding all three on projected benefits leaves too much uncertainty unresolved.
A worked example
The paper follows a hypothetical $90,000 license request presented as a $117,000 annual net benefit. After finance and the business owner identify the spending that can actually fall and complete the cost estimate, the same case shows a negative $130,000 result in year one. The paper sets out what the team would need to show before requesting a test.
Topics
The full paper covers the seven funding requirements, approval conditions by investment type, the Change, Grow and Run allocation, a worked example and considerations for government contractors. Free, no gate.