Tools, tools and more artificial intelligence (AI) tools.

2025 was the year of AI adoption

in technology, licenses and training. Yet, as 2026 unfolded, many executives found themselves asking a surprisingly difficult question: Where’s the return on investment (ROI)?

AI adoption and AI value creation turn out to be two very different things. The speed of AI is intoxicating, but speed without discernment scales costly mistakes.

Fortunately, an enormous opportunity is hiding in plain sight and it has less to do with AI than with human intelligence.

The 3 Traps of Enterprise AI Adoption

Most companies fell into one of three traps:

1. The AI Adoption Trap: These organizations invested heavily in tools, licenses and training, then measured success by how many employees used a tool (e.g., Microsoft Copilot or ChatGPT) or completed a course. But adoption is not impact. The result was activity without measurable ROI.

2. The AI Acceleration Trap: These organizations told everyone to experiment with AI without clear governance, decision rights or guardrails. Employees moved faster, but some exposed sensitive information, relied on inaccurate outputs or made poorly informed decisions. Experimentation wasn’t the problem. Accelerating activity without strengthening judgment was. The result was faster mistakes and greater exposure.

3. The AI Paralysis Trap: These organizations waited for AI to mature or for competitors to prove the returns. Now they feel behind, and that anxiety can push them into rushed investments. Their mistake was treating AI as an all-or-nothing decision rather than a series of deliberate, measurable experiments.

All three share the same root problem: Organizations built AI capability before decision capability. The missing ingredient isn’t another tool. It’s better human judgment, and judgment can be developed.

The Missing, Trainable Muscle: Human Judgment

No airline would put a pilot in a cockpit without hundreds of hours of training, simulations and demonstrated decision-making proficiency. Yet organizations routinely entrust employees with consequential business decisions without systematically developing or assessing their judgment.

Judgment is often treated as something people either have or don’t. It is better understood as a trainable capability, built through critical thinking, analytical reasoning, risk assessment, adaptability and reflection. Educational researcher David Jonassen classified decision-making as one of 11 distinct types of problem solving, and like the others, it can be taught. Furthermore, it improves with three components: a proven framework, challenging repetitions and meaningful feedback. Aviation, medicine and the military have understood this for decades. Business can, too.

Developing Decision Mastery

Decision mastery is how organizations turn AI proficiency into better decisions, and better decisions into measurable business value. It requires three complementary capabilities:

• Decision intelligence: framing problems, evaluating alternatives, assessing risk, challenging assumptions, and defending recommendations.
• AI fluency: understanding what AI can and can’t do, when to use it, when to trust its outputs and how to use it to improve decisions.
• Human intelligence: applying experience, critical thinking, creativity, ethical reasoning, collaboration and cooperation, as well as accountability, individually and as a team.

As an equation: DM = DI (AI + HI) = V^x, where V is value and x reflects how that value compounds as better decisions build on one another.

Sound decisions weigh value, feasibility, scalability, readiness, risk and governance rather than simply choosing the fastest or most appealing option. The result is a workforce that doesn’t merely use AI faster but makes better decisions with it.

The Consequences of Poor Decision-Making

The stakes are real. And sometimes the smartest AI decision is knowing when to stop.

In 2024, McDonald’s ended its IBM-powered automated drive-thru ordering test after trying it at about 100 restaurants. Viral videos showed the system misreading orders, piling Chicken McNuggets onto a single order and adding ice cream topped with ketchup and butter.

Ending the pilot wasn’t necessarily a failure. It may have been exactly the right call. Testing, learning and knowing when not to scale are essential elements of sound judgment. AI doesn’t eliminate the need for human decision-making; it raises the stakes. Whether leaders pilot, scale, redesign, defer or reject an AI initiative, each pathway carries consequences, and those consequences can be quantified.

The Economics of a Decision

ROI is a simple ratio. It has no time dimension, no risk adjustment, and it is not additive across choices. Two projects can show identical ROI while one pays back in eight months and the other in six years. One can be a coin flip while the other is near certain.

Net present value (NPV) handles all three. It compares the present value of expected cash inflows with the present value of outflows, reflecting the fact that a dollar today is worth more than a dollar tomorrow. It discounts for time, carries risk in the discount rate and it is additive, so competing options can be ranked against each other as well as against doing nothing.

Consider a hypothetical company evaluating AI for its sales function. Suppose the risk-adjusted NPV of scaling immediately is $3 million, while piloting first, after added costs and delays, is $4.5 million. Rejecting the initiative has an incremental NPV of zero.

Same technology, different decisions and a potential $1.5 million difference in value. And rejecting the initiative means foregoing $4.5 million in expected value.

These are projections, not guarantees, which is precisely the point. Better judgment means weighing the economic consequences of each path before committing resources. Now multiply that across thousands of decisions.

Measuring Real Impact

Organizations have measured the return on training and technology for decades. Far fewer measure the quality of the decisions those investments are meant to improve.

In “Moneyball,” the Oakland A’s learned to look beyond conventional statistics to overlooked indicators of performance, such as on-base percentage. The same discipline can apply to decisions: assessing how well employees frame problems, challenge assumptions, evaluate alternatives, anticipate risks and defend recommendations, then tracking improvement across teams, functions and leadership levels.

That offers something far more valuable than another AI adoption dashboard: evidence of whether decision-making capability is improving and whether that improvement shows up in business results.

From Individual Mastery to Organizational Decision Readiness

Enterprise value is the cumulative consequence of thousands of decisions made every day. Developing, practicing and measuring decision-making across the entire workforce, not just among leaders, does more than avoid costly mistakes. The benefits compound.

Organizations have spent enormous resources measuring whether they are AI-ready. As 2027 approaches, the more consequential question may be whether they are decision-ready.

In the AI era, the ultimate competitive advantage won’t be how intelligently the technology performs. It will be how intelligently the organization decides.