In 2026, global enterprise AI spending is projected to hit $665 billion — a figure that has grown faster than almost any comparable technology adoption curve in history. And yet, if you ask most executives in a candid conversation, the question that surfaces almost immediately is the same one: where is the return? According to McKinsey's 2026 Global AI Survey, 73% of enterprise AI deployments fail to achieve their projected ROI. Gartner places enterprises squarely in the "Trough of Disillusionment" of their AI Hype Cycle. The data would seem to support scepticism. But scepticism would be the wrong conclusion — because the evidence also shows that the organisations generating genuine value from AI share identifiable, repeatable characteristics. The problem is not AI. The problem is where and how most organisations are looking for the return.
The ROI Problem Is Real — But the Diagnosis Is Wrong
When AI investments fail to show returns, the instinct is to question whether the technology works. That instinct is almost always wrong. Across industries, organisations default to measuring productivity and labour cost savings — and when those signals are modest or slow to appear, momentum fades, initiatives stall, and what began with energy gets labelled a pilot and never quite scales.
The deeper problem is that most organisations have not yet built the infrastructure to measure AI's real contributions. Only 14% of CFOs report measurable ROI from AI to date — not because AI isn't generating value, but because the financial systems, attribution models, and governance frameworks needed to capture that value consistently are not in place. Meanwhile, a KPMG Global AI Pulse Survey found that 82% of self-identified AI leaders — organisations that built governance early and deployed deliberately — say AI is already delivering meaningful business value. The gap is not between the technology and reality. It is between organisations that built for scale and those that didn't.
"That's not because AI isn't delivering value. It's because many organisations are still looking for value in the wrong places — and expecting it to show up too quickly."— Enterprise AI ROI Analysis, TIME / Fortune, 2026
What AI Leaders Are Actually Measuring — And Getting Right
The organisations generating 5x to 10x ROI from AI share three identifiable characteristics that separate them from peers who are spending comparably but seeing little return. The first is that they started with strategy, not technology — identifying specific workflows where AI would deliver outsized impact before buying any platform. The second is that they treated AI transformation as an organisational change programme, not an IT project — investing in change management, restructuring roles, and bringing their workforce through the transition. The third is that they built measurement infrastructure before deployment — defining success metrics tied to business outcomes, establishing tracking systems, and reporting results back through the organisation.
IBM's 2026 goals framework for technology leaders identifies four categories where AI ROI is most reliably measured: operational efficiency (cycle time, throughput, error rates), experience and growth (customer satisfaction, conversion, retention), financial impact (cost-to-serve, gross margin), and risk and compliance (audit hours saved, policy violations avoided). Organisations measuring only labour cost savings are systematically missing most of the value AI is generating.
The Trough of Disillusionment Is Part of the Journey — Not the Destination
Gartner's placement of enterprises in the "Trough of Disillusionment" is frequently cited as evidence that AI has underdelivered. In fact, it is the opposite signal. The Trough of Disillusionment is a predictable, well-documented phase in the adoption cycle of every major general-purpose technology — from cloud computing to mobile to ERP. It follows the Peak of Inflated Expectations not because the technology failed, but because early expectations were set unrealistically high, and early deployments prioritised speed and novelty over discipline and governance.
What follows the trough — consistently, across every technology in Gartner's model — is the Slope of Enlightenment: the period where organisations that built properly begin to see compounding, defensible returns, and where the gap between leaders and laggards becomes structural rather than temporary. Ninety-three percent of senior leaders now believe that organisations that successfully scale AI in the next twelve months will achieve an insurmountable competitive lead over peers. That consensus is not unfounded optimism. It reflects a clear-eyed reading of what disciplined early movers are already seeing.
Crucially, even as executives acknowledge the ROI challenge, they are not walking away. A KPMG survey found that three out of four global leaders will prioritise AI investment despite economic uncertainty in 2026. The view is increasingly one of long-term strategic positioning, not short-term cost reduction — a shift in mindset that correlates directly with the outcomes that leading organisations are generating.
The Hidden Value Problem: AI's Contributions Are Embedded, Not Visible
One of the most practical insights from 2026's wave of enterprise AI research is that AI's most significant contributions are often embedded in existing processes rather than visible as standalone outputs. When AI shortens the time to close a support ticket, reduces the error rate in a compliance workflow, or improves the accuracy of demand forecasting, those gains show up in business outcomes — but they are rarely attributed to AI in financial reporting.
As one analyst put it: "No one calculates ROI by counting the number of Word documents produced. But ROI calculations on AI projects are not going away. If it burns cash and fails to produce tangible ROI, it will be retired." The implication is dual: AI must be measured properly, and it must be embedded into workflows in ways that make its contribution legible to the financial systems that govern capital allocation.
Deloitte's 2026 State of AI in the Enterprise report found that 54% of organisations expect to move 40% or more of their AI experiments into production within the next three to six months — yet only 25% have reached that milestone today. The aspiration is real. What stands between aspiration and achievement, consistently across the research, is governance — not technology.
"This shift in mindset by business leaders from viewing AI as something that must deliver an immediate return to one that sees AI as a long-term investment — recognising it as a strategic enabler for enterprise-wide transformation — is an important milestone. But that shouldn't translate into investing in AI blindly, without a clear strategy."— KPMG Global AI Pulse Survey, 2026
What Leaders Should Do Now: Five Practical Shifts
The research is consistent about what distinguishes enterprises that generate real AI returns from those stuck in the ROI gap. For leaders navigating this now, five practical shifts make the difference:
- →Redefine what ROI means for AI — move beyond labour cost savings to measure cycle time, quality, compliance efficiency, customer satisfaction, and margin impact
- →Build governance before you build more pilots — define who owns outcomes, establish accountability standards, and create audit trails before expanding AI's footprint
- →Send your best people — PwC's 2026 analysis found that AI front-runners assign top business talent, not only technical teams, to AI implementation — these are the people who can link AI capability to business outcome
- →Have the courage to exit — organisations that generate ROI from AI are the ones willing to stop investments that aren't working, not the ones that persist out of sunk-cost anxiety
- →Build for compounding, not delivery — AI systems designed for continuous improvement, retraining, and adaptation generate competitive advantage through iteration; one-time deployments depreciate rapidly