The CFO's 2026 Reckoning: How Finance Leaders Are Rethinking Capital Strategy in an AI-Driven Economy

The annual finance leadership survey that Alvarez & Marsal's Global CFO Practice published in April carries a headline that would have seemed implausible three years ago: 68% of the 600 CFOs polled are actively reallocating capital expenditure toward AI infrastructure in 2026, with a median budget shift of 14 percentage points away from traditional IT categories. For a profession that has historically prized financial discipline and measurable returns, the scale of this reallocation, in the absence of established ROI frameworks, represents a significant departure from standard capital allocation doctrine.

"We are making multi-hundred-million-dollar bets on technology platforms whose long-run economic returns we genuinely cannot model with precision," said Claire Ashford, CFO of a Fortune 200 industrial manufacturer who participated in the survey. "That is uncomfortable for finance leaders trained to demand evidence. But the alternative, waiting for certainty while competitors move, feels far more dangerous in the current environment."

The Build-Buy-Partner Question Evolves

For decades, the build vs. buy vs. partner framework has been the organising logic of corporate technology investment decisions. AI is destabilising all three legs of that stool simultaneously. Building bespoke AI capabilities requires talent and infrastructure that few companies can sustain at competitive scale. Buying off-the-shelf solutions risks embedding dependencies on vendors whose own competitive position may shift dramatically within 18 months. Partnering with hyperscalers creates access to cutting-edge capability but raises questions about data sovereignty and negotiating leverage as dependency deepens.

The Alvarez & Marsal data suggests the most sophisticated finance leaders are abandoning the framework as a static decision and replacing it with a dynamic portfolio approach, regularly reassessing the optimal mix across a rolling 18-month horizon rather than making single large bets. The firms executing this approach tend to have one characteristic in common: a CFO who has personal technical literacy in AI systems, not just financial literacy about AI spending.

The ROI Measurement Gap

The survey's most sobering finding may be the measurement gap. Less than 30% of respondents said they have a robust, outcome-based ROI framework for AI investments currently in place. The majority are still tracking inputs, headcount redirected, compute spend, model training costs, rather than outputs like revenue attributed, cost avoided, or cycle time reduced. That matters enormously for capital allocation discipline: without outcome measurement, the budget approval process for AI becomes a faith-based exercise rather than an analytical one.

"We spent 18 months approving AI projects based on vendor-supplied productivity estimates that turned out to be directionally right but operationally useless. Building our own measurement infrastructure was the most important thing we did in 2025." , CFO, Fortune 500 Financial Services firm (survey respondent)

• Tier 1 metrics (most commonly tracked): Compute spend, headcount redeployment, implementation timelines
• Tier 2 metrics (emerging practice): Process cycle time reduction, error rate improvement, analyst hours freed per week
• Tier 3 metrics (best-in-class): Revenue directly attributable to AI-assisted decisions, cost-per-outcome vs. pre-AI baseline, decision latency reduction

The Evolving CFO Role

Beyond the capital allocation question, the survey paints a picture of a finance function undergoing a profound role shift. CFOs at the highest-performing organisations, defined by revenue growth and EBITDA expansion over the prior three years, are now spending an average of 40% of their time on strategy and forward-looking analysis, up from 22% in 2022. The corresponding reduction is coming from financial reporting, compliance oversight, and operational finance, as automation and AI-assisted analytics absorb much of the traditional workload of the finance function.

That shift is not without tension. Several survey respondents flagged concerns about governance and oversight in a world where AI systems are producing the analysis that CFOs are presenting to boards. "My job used to be knowing the numbers cold," noted one CFO of a mid-cap technology company. "Now my job is knowing which AI outputs to trust, which to question, and how to build the right human review checkpoints into a largely automated process. That requires different skills, and frankly, different people."

The talent implication of that observation is not lost on the finance leaders who participated in the research. Building a finance function capable of operating effectively in an AI-augmented environment, while maintaining the governance standards that regulators and boards expect, is the defining organisational challenge facing CFOs heading into the second half of 2026.