If the leadership team cannot say who owns the AI decision, there is no AI strategy. There is a technology project with a budget.
I have watched organizations hand the AI agenda to the technology group and then wonder why the operating model did not change. IT did its job. Leadership did not. The tools arrived, and the work, the decision rights, and the measures of success stayed where they were. That is not an implementation lag. That is what happens when a company treats an organizational choice as a systems purchase, and it happens more often that one would think.
A January 2026 IBM Institute for Business Value study of 2,000 C-suite executives found that 68% were concerned their AI efforts could fail because AI was not integrated with core business activities (IBM Institute for Business Value, 2026). I am less interested in the percentage than in the failure it describes.
An organization can have capable technology, substantial investment, and a great deal of AI activity without having decided who owns the choice.
Framing AI primarily as an IT issue is a common trap. Though it depends on heavy technical foundations like data and cybersecurity, treating it through a standard, linear technology lens misses the broader organizational shift. Implementing the technology and determining what the organization should do with it are two different responsibilities.
Start With the Problem, Not the Technology
When a new technology develops as quickly as AI has, it is natural for conversations to begin with the tools. Which platform should we use? Which model performs best? Should we build something ourselves or purchase it? How will it connect with our existing systems?
Those questions matter, but I believe they should come later. The first question should be much simpler as in What problem are we trying to solve?
The problem might be customer response times that are too slow, or employees buried in routine administrative work, or managers who cannot get useful information in time to decide. It might be errors, delays, and handoffs that nobody owns. Those are operating problems. They are not yet a reason to pick a platform.
AI may be an excellent way to address any of them. It may also not be. Starting with the problem forces leadership to define the outcome before becoming invested in a particular solution. I have seen enough technology initiatives over the years to be wary of reversing that sequence. Once an organization becomes enthusiastic about a tool, there is a temptation to search for places to use it. That often generates activity without necessarily creating much value.
The more disciplined sequence is to identify something consequential that needs to improve, understand why the current process is not producing the desired result, and then determine whether AI offers a better way of accomplishing it. Identifying the problem and optimizing the process should come before deciding on an AI platform.
AI Changes Work, Not Just Systems
The strongest reason AI cannot be owned primarily as a technology initiative is that meaningful AI adoption eventually changes work. Once a system begins classifying information, analyzing data, preparing content, recommending decisions, routing requests, or executing tasks, the organization is no longer simply installing software. It is changing how work moves through the enterprise.
A role may contain ten different activities, five of which AI can perform very well. What should happen to the role? Should the employee spend more time with customers, analyze more complex problems, exercise greater judgment, or simply handle more volume? If AI produces a recommendation, who has authority to act on it? When does a routine transaction become an exception requiring human review?
These questions become especially important because automation often occurs incrementally. An organization may automate one task, then another, and then another without ever reconsidering the job or process as a whole. I have watched a workflow pick up a new tool every few months until the steps no longer matched the job anyone had been hired to do. Each change had a sponsor and a reasonable explanation. What nobody did was step back and redesign the work those tools were now doing. The system looked current. The organization had not caught up to it.
That is why the leadership team must look beyond individual use cases. The objective is not simply to determine which tasks AI can perform. It is to determine how work should be organized when people and intelligent systems can each perform different parts of it. The objective is to strive for optimality.
That requires input from IT, certainly. It also requires operations, finance, human resources, risk and compliance, and, most importantly, the people who understand the work itself.
Adoption Is Not the Same as Value
There is another distinction I think executive teams need to protect carefully. AI activity is not necessarily AI value. Organizations can count users, licenses, pilots, automated processes, prompts, or hours saved. Those measures tell us something about adoption. They do not necessarily tell us whether the organization is performing better.
I recommend seeking to understand something more fundamental. Did customer response improve? Did cycle time decrease? Did errors decline? Did employee capacity move toward more valuable work? Did risk decrease? Did revenue improve? Are decisions being made faster or with better information?
Those are organizational outcomes. A technically successful implementation can still be a poor business investment. The technology may perform exactly as expected while producing little meaningful improvement in the outcome that originally mattered.
This is one place where senior leadership has to resist the excitement that naturally accompanies a rapidly developing technology. I am exceedingly optimistic about the potential of AI, but optimism should not lower the standard by which investments are evaluated. If anything, the amount of attention and capital moving toward AI should raise that standard.
The relevant question should not be, Did we successfully implement AI? It should be, What became measurably better because we did?
Governance Is Part of the Strategy
Risk and governance are sometimes treated as activities that follow the strategy. First, determine what the organization wants to do with AI, then establish the controls. That sequence is how a pilot becomes a practice before anyone has decided who can stop it.
The value an organization hopes to create and the risks it is willing to accept should be considered concurrently. A routine, low-consequence process may justify substantial automation. A decision involving significant financial exposure, regulatory obligations, employment consequences, sensitive information, or customer harm may warrant a very different level of oversight.
That does not mean the organization is being less innovative. It means leadership is distinguishing among different kinds of risk rather than treating every AI application the same way. If governance shows up only after the model is already in the workflow, it usually arrives as a memo. Leaders still need to know where the system may act on its own, where a person has to review the work, what information may be shared, and who remains accountable when the output is wrong.
The questions that matters are who can stop it, and whether that person knows the authority is theirs. Those decisions are not simply controls around the technology. They help define how the organization intends to use it.
Leadership Has to Own the Agenda
IT should be one of the central partners in any serious AI initiative. In many organizations, technology leaders will have a broader view than anyone else of the available tools, architecture, data environment, security requirements, and integration challenges. That expertise should carry considerable weight. But it should not require IT to answer questions that belong to the leadership team.
I have sat in meetings where the CIO was the most prepared person at the table and still should not have been the owner. The question on the table was not which platform to buy. It was which work would change, which jobs would be redesigned, and who would be accountable when the model was confident and the outcome was wrong. Those are not IT questions. Leaving them with IT is how a company gets very good at pilots and no better at the business.
IT should answer whether the system can be deployed safely, whether the data can support it, and whether it will work with what the organization already runs. Leadership has to answer which work should change, who is accountable for the result, and what improvement would justify the investment. Those are both necessary. They are not the same decision.
Where can AI create the greatest organizational value? Which processes deserve to be redesigned? What should employees do differently as routine work is automated? Which decisions should remain human? What level of risk is acceptable? Where should the organization invest first? How will success be measured? Those are choices about the organization, not merely its technology.
My concern with describing AI as an IT strategy is therefore not semantic. The label can influence who owns the initiative, which questions get asked, and how success is ultimately defined. If AI is treated primarily as a technology program, organizations may become very good at deploying tools. If it is treated as an organizational strategy, leadership is forced to confront the harder questions about work, value, people, risk, and performance.
The Institute of Strategic Leadership & Innovation is constructed for this kind of question. We are not here to select the system. The focus is to put the ownership question in front of the people responsible for answering it, before a purchase is treated as a strategy.
Technology teams will remain indispensable to putting the tools to work. They should not be asked to invent the reason for using them.
If the CEO is waiting on IT to invent the strategy, the organization already has its answer, and it is the wrong one.
References
IBM Institute for Business Value. (2026, January 19). The enterprise in 2030. IBM. https://newsroom.ibm.com/2026-01-19-ibm-study-ai-poised-to-drive-smarter-business-growth-through-2030