AI has arrived. The value still hasn't added up.

Adoption has advanced. Value, not always. The problem is no longer using AI. It is transforming usage into a reliable economic consequence.

This question matters because the market debate has shifted. Adoption is no longer the exception. McKinsey reports that 88% of organizations regularly use AI in at least one business function. Yet, the same research notes that most have not embedded it deeply enough into workflows and processes to generate material, enterprise-level benefits. Following this trend, only 39% report meaningful EBIT impact at scale.

This changes the conversational pivot. The problem is no longer “how to put AI somewhere.” The problem is different: how to transform technological presence into economic consequence. Gartner has been insisting that reaching value requires more discipline. In March 2026, they stated that without strong foundations, AI tends to remain an “expensive experiment” for most organizations. Along the same lines, Gartner also points out that the ROI debate in AI cannot be treated as a simple and uniform problem, because value creation depends on fundamentals, context, and portfolio management.

AI adoption is no longer the exception

The most common mistake may be a category error. Many companies confuse adoption with value capture.

Having AI in copilots, analytics, service desks, content, enterprise search, or automations can elevate capacity. But raising capacity is not the same as adding up the numbers in the business. Distributed capacity without process redesign, reliable context, prioritization criteria, and execution discipline tends to produce localized gains, not compounded impact. McKinsey itself points out that one of the factors most associated with relevant impact is the intentional redesign of workflows, not merely the availability of the technology.

It is no coincidence that Gartner warned that more than 40% of agentic AI projects will be canceled by the end of 2027 due to rising costs, unclear value, or inadequate risk controls.

What the board actually buys when investing in AI

This point also helps explain why the conversation is moving. In May 2025, Gartner was direct: “data-driven is a means to an end”. The real objective is to make better decisions. Therefore, they recommended that D&A advance from data-driven to decision-centric, applying Decision Intelligence to make decision-making more contextual, continuous, and connected.

From data-driven to decision-centric

This shift matters a lot. It suggests that value is not simply in seeing better, but in deciding better under new conditions of speed, automation, and ambiguity.

But here enters a distinction that must be made carefully.

In theory, an organization can declare itself decision-centric without formally adopting Decision Intelligence. That is true.

The problem is different: as scale, speed, and AI autonomy increase, along with the pressure for ROI and the need for accountability, the decision-centric discourse tends to become fragile if it is not accompanied by an equivalent mechanism of decision discipline.

In other words: decision-centric is the axis; Decision Intelligence is one of the most consistent answers to transform this axis into real organizational capability.

This connection becomes even stronger when we look at an idea the market has accepted for a long time. For years, we have learned — correctly — to say that data is a strategic asset. Gartner itself speaks of using data as an asset for differentiation and growth. DAMA, in turn, treats data as a valuable strategic asset.

Data accumulates context, memory, evidence, and potential. But it is the decision that commits resources, allocates risk, chooses paths, moves execution, and converts potential into economic consequence. Put more directly: data are information assets; decisions are the assets that transform information into action and potential into value.

This point is especially important because it protects against an increasingly common confusion: the belief that analytical maturity, by itself, already equals decision-making maturity.

It does not.

A company can have a good data platform, sophisticated analytics, embedded AI, mature data governance, and still continue making fragile, poorly framed, poorly prioritized, or poorly governed decisions. When decision architecture does not keep up with the technological one, clear symptoms emerge in operations:

In other words: AI entered through the technology door, but the problem it opened up is one of management. It is a problem of context, mandate, criteria, coordination, risk, prioritization, and consequence. The more AI spreads, the less sufficient a conversation centered merely on models, tools, or features becomes. What begins to matter is the organization’s ability to repeatedly bridge technical capacity with economic outcome.

There is another distinction here that warrants care. Saying that the value hasn’t added up yet does not mean denying that real gains exist. They do. McKinsey shows growing use across multiple functions and benefits in specific cases. The point is more demanding: localized benefit is not the same as material and sustainable enterprise-wide impact. It is perfectly possible to experience productivity, time reduction, and even sporadic revenue gains without these having yet been converted into a consistent value architecture at the enterprise level.

Nor does it mean saying that “only governance is missing.” Governance matters a lot, and Gartner has even highlighted it as a value accelerator. But governance alone does not choose where to apply AI, it does not solve operational debt, it does not create quality context, it does not redefine flows, it does not make trade-offs explicit, and it does not bridge the gap between insight and consequence. It reduces risk and increases integrity; it does not substitute managerial discipline.

This is exactly why the transition from decision-centric to Decision Intelligence is relevant. Not because every company needs to adopt this vocabulary immediately, but because the market has already begun demanding solutions for the problem this vocabulary tries to solve.

When Gartner recommends advancing from data-driven to decision-centric and applying Decision Intelligence, what it is saying, in practice, is that the company needs to leave the comfort of the analytical layer and start treating the decision as an explicit object of operations.

The executive question that truly matters now

For this reason, perhaps the most mature executive question today is no longer “where can we use AI?”.

Perhaps it is this one:

Until this question is treated more seriously, many organizations will continue holding AI everywhere and value in few places.

And perhaps this is the true inflection point of this market phase: no longer discussing merely the entry of AI, but rather the discipline needed for it to finally add up.