Yet behind that surface momentum lies a striking paradox: capital spending is soaring, experimentation is spreading, and technical capabilities are advancing rapidly, but organizations that have genuinely restructured the way they operate around AI remain the exception rather than the rule. This suggests that the central question of the current phase is no longer how powerful the models are, but how ready enterprises are to place AI at the core of their operating system.
What is especially notable is that recent international reports and executive discussions are converging on the same conclusion. AI does not produce transformation simply by being attached to legacy systems. Instead, it tends to amplify the strengths and weaknesses already present inside an organization. Where a company has clean data, disciplined processes, clear governance, and a workforce capable of learning and adapting, AI can become a force multiplier. But where data is fragmented, project ownership is ambiguous, software architecture is patchy, and decision-making remains heavily manual, AI often stalls at the “promising pilot” stage and fails to create enterprise-scale value.
Recent work by the World Economic Forum and McKinsey points to the scale of this gap. Global AI spending in 2025 reached roughly $1.5 trillion. The share of organizations using AI in at least one business function has risen sharply. Yet only about one-third say they have begun scaling AI at the enterprise level, and only a very small minority consider themselves truly mature.
These figures matter because they confirm that the bottleneck is no longer a lack of enthusiasm or funding. The bottleneck is the ability to move from experimentation to operations, and from technological excitement to institutional discipline.
When the Constraint Lies Inside the Enterprise
For years, discussions about AI revolved around model performance, compute power, and the race for chips. At scale, however, the center of gravity shifts. The hardest problem turns out to be internal governance. To move beyond a pilot, an AI program forces an enterprise to answer foundational questions: Which data can actually be trusted? Who owns the outcome? Which processes can be automated without undermining control? Where does human intervention sit when the model is wrong? And who is responsible for measuring economic impact after deployment?
These questions may sound less glamorous than the latest model release, but they determine the fate of most AI programs. Many organizations invest aggressively in tools while underinvesting in data standardization, process redesign, and workforce capability. As a result, AI may deliver quick wins in document handling, office support, or marketing, yet fail to reach the operational backbone: demand planning, production scheduling, inventory optimization, maintenance, customer service workflows, or capital allocation. In that sense, AI is much easier to deploy at the edges than at the core - unless the organization is willing to redesign itself.
Viewed this way, one recurring executive insight becomes highly revealing: AI does not first replace organizational structure; it exposes its hidden flaws. A company that looks data-rich on presentation slides but lacks consistency across systems quickly runs into trouble when AI requires operational-grade data. A company that speaks fluently about digital transformation but has never clarified ownership of key workflows will hesitate the moment a machine is allowed to recommend or execute actions. AI, in other words, is an unusually honest test of managerial quality.
From Pilot Purgatory to Real Economic Value
The phrase “pilot purgatory” has become an accurate description of many enterprises: countless promising pilots, but no broad operational rollout. This is not outright failure, yet it is a form of strategic waste. The organization proves that AI can work, but lacks the consistency required to turn it into a money-generating system. That is why real economic value emerges primarily in the relatively small number of organizations that treat AI as a long-cycle industrial investment program rather than a short-cycle IT upgrade.
Saudi Aramco is often cited as a clear example. At the World Economic Forum in 2026, CEO Amin Nasser said the company had expanded its portfolio from roughly 400 to 500 AI use cases, moved about 100 into real deployment, and realized around $6 billion in technology value across 2023 and 2024, with AI contributing nearly half.
The important point is not the existence of a magical proprietary algorithm. It is the depth of preparation: decades of operational data, a disciplined process for selecting use cases, large-scale retraining of engineering talent, and a management system capable of measuring economic outcomes with rigor.
The lesson is straightforward. Companies that generate return on investment from AI do not begin by asking which model is fashionable. They begin by asking which business bottleneck is large enough to justify redesign. They do not chase the number of pilots. They build a selective investment portfolio in which many ideas can be tested, but only those that prove operational, financial, and governance value are scaled. That logic resembles internal venture capital far more than conventional software procurement.
What Leaders Should Reassess Before Talking About AI
Taken together, these developments make one thing clear: the next phase of AI is fundamentally a test of institutional capability at the enterprise level. Saying that a company “has deployed AI” no longer means very much. The more relevant question is whether AI has changed any operating metric, in which function, and whether that improvement can survive at scale. If those questions cannot be answered, the organization is probably still in demonstration mode.
For business leaders, that means reversing the order of priorities. Instead of rushing to adopt every new tool, companies need to invest more heavily in four foundations: data integrity, a clear operating model, workforce capability, and governance trust. These are the invisible infrastructures that determine whether AI produces real value or merely creates the appearance of modernity.
AI is therefore not simply a new layer of technology placed on top of an old enterprise. It is a stress test that forces an organization to become more orderly, more transparent, more measurable, and more willing to relearn how decisions are made. The companies that understand this are beginning to pull ahead. Those that continue to treat AI as little more than an IT enhancement may accumulate many interesting pilots - yet remain outside the trajectory of genuine transformation.
---------------
Main reference sources: World Economic Forum, McKinsey, Accenture, remarks by Saudi Aramco at WEF 2026.
