Key takeaways
- Most enterprise AI projects fail to show positive returns not because the technology underperforms, but because CEOs deploy it without a clear value measurement framework.
- HBS research found that AI actively rewrites internal corporate power dynamics, meaning middle managers with headcount-based authority will quietly block initiatives that threaten their reporting lines.
- Kearney’s work published in Harvard Business Review challenges the prevailing “spend-scale-speed” approach to AI, arguing that leaders must pause to reframe before they accelerate.
- Companies that tie AI mastery explicitly to competency models and promotion cycles see faster adoption and fewer internal blockers, according to HBS research.
- Governance frameworks such as the NIST AI Risk Management Framework and ISO 42001 are no longer optional compliance exercises; they are the operational guardrails that protect return on investment.
Most CEOs know their organization is spending heavily on AI. Few can say exactly what they are getting back for it.
That gap is not a technology problem. It is a leadership problem. The decisions that determine whether AI delivers real financial return sit squarely at the CEO level, and the specifics of those decisions are more precise than most executives expect.
Why AI spending so often becomes a cost center
AI investment becomes a cost center when spending is decoupled from measurable business outcomes from the start. A project without a pre-agreed financial metric is, by definition, a project that cannot prove its worth.
Kearney’s analysis, published via Harvard Business Review, identifies “spend-scale-speed” as the dominant but flawed corporate reflex. Organizations race to deploy tools, announce projects, and report AI initiatives to boards before anyone has established what success looks like in dollar terms. The result is a portfolio of pilots that never graduate to production value.
The fix is not slower deployment. It is earlier definition of the metric. Before any AI project receives budget approval, the CEO should demand one number: the specific, time-bound financial outcome the project is expected to move, and by how much.
That number anchors everything else. It determines which team owns accountability, which data needs to be clean before launch, and when the organization should stop funding a project that is not moving toward its target.
The internal politics that kill returns before they start
AI threatens existing power structures in ways that financial projections rarely capture. HBS research highlights that AI actively rewrites internal corporate power dynamics, and that dynamic plays out in budget meetings, roadmap reviews, and hiring decisions long before any model goes live.
The mechanism is straightforward. Middle managers whose influence and compensation are historically tied to headcount have a direct financial incentive to slow or block any AI initiative that reduces the number of people reporting to them. They rarely do this openly. They raise technical objections, question data quality, or request additional pilots.
CEOs who are unaware of this pattern misread the resistance as legitimate concern. Those who recognize it can respond structurally. Progressive enterprises are explicitly baking AI mastery into competency models and accelerating promotion cycles for those who use the technology to drive efficiency, according to HBS research. That structural signal changes the incentive calculation for every manager watching it happen.
The implication is direct: the CEO must make AI proficiency a visible criterion for advancement, not just a talking point in an all-hands meeting.
Measuring what actually matters
Meaningful AI ROI measurement requires three layers: operational efficiency gains, revenue impact, and risk cost reduction. Each needs its own baseline.
JpMorgan Chase provides one of the clearest documented cases of AI applied to legal contract analysis, with the firm’s tools reported to have eliminated a substantial volume of attorney hours on work that previously required manual review. That outcome matters not because the scale is impressive in isolation, but because it was measurable against a known baseline: the hours lawyers previously spent on the same task. The CEO-level lesson is that ROI is only calculable where a pre-AI baseline exists.
Deloitte research on AI investment ROI points to the gap between organizations that measure AI value systematically and those that treat it as a general overhead. The organizations that build formal tracking mechanisms, covering both the cost of compute and the value of output freed, consistently report clearer returns and make faster decisions about when to scale or discontinue projects.
CEOs should also ensure cloud costs are tracked at the workload level. Data from cloudzero.com shows that AI infrastructure costs can grow faster than anticipated because teams provision capacity speculatively, without tying consumption to production value. That pattern is a direct drain on ROI and requires the CEO to make cost visibility a non-negotiable operating standard.
Governance as a return-protection mechanism
Governance is not a constraint on AI ROI. It is what protects AI ROI from regulatory, reputational, and operational risk. This distinction matters because many leadership teams treat compliance work as an overhead to be minimized rather than an investment in durable returns.
The NIST AI Risk Management Framework, published by nist.gov, provides a structured approach to identifying, assessing, and managing the risks that sit inside AI systems before they reach customers or regulators. ISO 42001, covered in detail by deloitte.com and konfirmity.com, establishes an auditable management system for AI governance that is increasingly being referenced by institutional buyers and regulators in procurement and disclosure requirements.
The connection to ROI is concrete. blog.promise.legal and thecaq.org report that SEC disclosure requirements for AI are tightening, and that companies which cannot articulate their AI governance posture face both regulatory exposure and investor scrutiny. A governance gap does not just carry compliance risk. It erodes the credibility of the return numbers the CEO is reporting to the board.
CEOs who build governance structures early convert them into competitive signaling. Organizations that can demonstrate ISO 42001 alignment or NIST RMF implementation tell institutional counterparts that their AI returns are auditable, not just asserted.
The “stop, reflect, reframe” discipline
Speed is not the enemy of AI ROI. Unreflective speed is. Kearney’s Ramyani Basu, writing for Harvard Business Review, challenges the “spend-scale-speed” mindset directly by arguing that leaders need a deliberate pause to reframe their AI portfolio before they commit to the next wave of investment.
That pause has a practical structure. The CEO should ask three questions of every active AI initiative: Is the project solving a problem the business would pay to solve if it were solved by humans? Has the baseline against which we measure improvement been formally recorded? Who owns the number, and what happens to their career if the number is missed?
None of these questions require a consultant. They require a CEO who insists on evidence before the next budget cycle opens. The organizations that build this discipline consistently are the ones that convert AI from a line item into a return.
Connecting AI strategy to board-level accountability
The CEO is the person who decides whether AI appears on the board agenda as a cost report or as a value report. That choice shapes how the board allocates capital over the next several years.
Terminal-x.ai’s analysis of enterprise AI outcomes identifies a consistent pattern: organizations where the CEO personally owns the AI value narrative produce cleaner ROI data than those where AI is delegated entirely to the CTO or a Chief AI Officer. The issue is not technical expertise. It is signal clarity. When the board hears the AI story from the CEO, it carries strategic weight.
MIT Sloan Management Review’s leadership and AI insights for 2025, available at mitsloan.mit.edu, reinforce this point by noting that the CEO’s visible engagement with AI governance and measurement sets the cultural tone that determines whether middle management treats AI accountability as real or performative. That cultural signal is the CEO’s exclusive responsibility. No one else in the organization can send it.
FAQ
What is the single most important action a CEO can take to improve AI ROI?
Require a specific, time-bound financial baseline before approving any AI project budget. Without a pre-agreed metric tied to a known pre-AI baseline, the project has no mechanism to prove or disprove value.
How should a CEO handle internal resistance to AI automation from middle management?
Restructure the incentive system rather than argue with the resistance. HBS research shows that explicitly tying AI proficiency to competency models and promotion decisions shifts the calculus for managers who would otherwise protect headcount-based authority quietly.
Is governance spending a drag on AI ROI or a contributor to it?
Governance is a contributor. Frameworks such as the NIST AI Risk Management Framework and ISO 42001 protect returns by reducing regulatory exposure, meeting increasing SEC disclosure expectations, and providing the auditability that institutional buyers and investors require before treating AI returns as credible.
Does the CEO need deep technical knowledge to manage AI ROI?
No. The CEO needs clarity on financial baselines, ownership accountability, and governance posture. Technical depth sits with the engineering and data science teams. The CEO’s role is to insist that every technical decision connects to a measurable business outcome.
How should AI ROI be presented to the board?
Present it as a value report with specific baselines, actual outcomes, and variance explanations, not as a capability update. When the CEO owns the narrative directly, the board treats AI investment as a strategic matter rather than a technology overhead.