Beyond the Headline ROI

Some of us are looking at substantial funding requests for AI initiatives, platform modernisation, and broader enterprise development. Whether they arrive during the budget period or otherwise, we typically expect clearly articulated ROI figures. The underlying decisions, however, remain complex. Such investments compete with mandatory capital expenditure, compliance requirements, ongoing operational costs, and projects that typically offer more predictable near-term returns. While boards continue to anticipate both growth and earnings stability, this is where disciplined ROI analysis earns its keep.

Conventional ROI analysis tends to favour lower-risk options and leaves larger capability-building investments under-resourced. Approving the full request increases pressure on the overall capital budget directly tied to the firm’s competitive advantage. Neither strict cost containment nor unconstrained strategic ambition adequately eases the tension.

NPV, IRR, and payback periods remain essential tools as they instil discipline on cash-flow projections and timing. For AI and enterprise development initiatives, however, they turn inadequate. Many benefits materialise only indirectly or after complementary changes in processes, data quality, and organisational behaviour. Productivity improvements, risk reduction, and the creation of future options rarely appear cleanly within a single project’s cash-flow model.

Deloitte research notes that satisfactory ROI on typical AI use cases often takes two to four years — well beyond the seven-to-twelve-month payback periods expected for conventional technology investments.1 When the budget cycle plans no further than the next financial year, it structurally excludes the estimated time window in which AI returns actually materialise—which means the budget proposal can never fully articulate the benefit case for investment.

Raising the discount rate to compensate for risk is common practice, yet the adjustment is often arbitrary and can eliminate projects that still merit consideration. Scenario-based analysis offers a more transparent alternative. Organisations also rarely evaluate these requests within a genuine portfolio context. Ranking proposals solely by standalone IRR or NPV systematically privileges short-payback efficiency projects at the expense of longer-horizon capability building. McKinsey research shows that companies treating the budget as a true capital-allocation process and deliberately shifting approximately 8 percent of capital each year toward higher-value opportunities achieve superior growth and returns.2 KPMG data further highlights the measurement gap: only a small minority of leaders report having established clear AI ROI, underscoring the need for stronger visibility and consistent tracking.3 BCG’s analysis of capital allocation reaches a parallel conclusion — higher-performing firms allocate resources according to the strategic roles of businesses rather than on a purely project-by-project basis. They apply differentiated evaluation criteria and maintain deliberate portfolio balance. Uniform treatment of all requests tends to dilute impact.4 Harvard Business Review (HBR) has similarly emphasised the importance of prioritising AI investments according to where value is created in an industry rather than spreading resources evenly across experiments.5

A more robust approach preserves financial rigour while addressing these gaps. Begin by classifying the investment.

1. Distinguish among Operate (maintaining resilience), Expand (modernising and scaling existing capabilities), and Innovate (building new sources of advantage). Each category requires different standards of evidence and different time horizons. Most organisations still allocate disproportionately to Operate; deliberate rebalancing is typically required.6

2. Require driver-based business cases constructed around explicit scenarios. Replace high-level assertions with specific value drivers — hours saved multiplied by value per hour, conversion improvements multiplied by volume, or quantified risk reduction. Present base, upside, and downside cases.

3. Include full cost of ownership and a clear baseline of outcomes if the investment is deferred. At this stage, the presence of individuals who can translate technical AI claims into credible financial drivers proves particularly valuable. Domain expertise spanning both technology and finance strengthens the quality of the analysis. Deloitte findings reinforce that greater cross-functional C-suite involvement in digital investment decisions correlates with stronger profitability outcomes.7

4. Calculate core financial metrics without distortion. Apply NPV at the organisation’s standard cost of capital rather than an inflated rate. Report IRR, payback, and sensitivity on the variables that most influence results. McKinsey research indicates that supplementing NPV with complementary factors such as resilience, adaptability, and sustainability helps surface assumptions that pure financial analysis often leaves unexamined.8

5. Make non-financial considerations explicit through metrics established in advance. A straightforward scoring approach that incorporates financial profile, strategic contribution, risk, and resource intensity renders trade-offs visible. BCG emphasises that IRR alone is insufficient; evaluation criteria should reflect the nature of the investment.9

6. Insist on portfolio-level discussion. Rank the request against competing claims on capital and identify specifically which activities will be slowed or stopped if funding proceeds. Treat reallocation as standard practice. Apply stage gates so that larger programmes continue to receive funding only when interim results materialise. McKinsey analysis of technology budgets demonstrates that directing a larger share of early investment toward platforms and enabling tools can improve overall returns over time.10

Real-world examples, such as those reported by finance leaders at companies including General Mills and Hewlett Packard Enterprise, illustrate how linking AI initiatives directly to quantifiable efficiencies and cost reductions strengthens both internal discipline and external credibility.11

Finance function’s contribution lies in providing consistent comparison of material requests, clear articulation of trade-offs, and a method for assessing subsequent delivery. With this discipline in place, CFOs can devote less attention to gatekeeping and more to directing capital toward priority uses.

A shared vocabulary with technology and business leaders also ensures that proposals reach the board in terms that support rigorous evaluation. Including team members who bridge AI technical detail and financial assessment further improves the quality of recommendations. KPMG research similarly underscores that alignment between CFOs and CIOs on the definition and measurement of technology ROI remains a frequent point of friction—and a critical opportunity.12 Uncertainty cannot be eliminated. It can, however, be made visible, compared across the portfolio, and used as the basis for reallocating capital when evidence changes. Given that the benefits on typical AI use cases could take a few years to materialise, the approval decision cannot be the last point of scrutiny. Each investment warrants a fixed reassessment point, ideally aligned to the interim stage gates built into the portfolio process. As AI and enterprise development budgets increase and external scrutiny intensifies, this capability itself becomes a source of advantage.

The central questions on any large request therefore extend beyond the headline ROI. They concern the investment’s position in the broader portfolio, the assumptions that must hold for value creation, the activities that will be reduced or deferred to fund it, and the means by which progress will be measured. Addressing these questions rigorously converts budget approval into effective capital allocation.