the loudest conversation about AI and work Focus on what disappears. A more pressing problem within IT is one that is quietly proliferating. As AI capabilities spread across the stack, work is broken down into new, poorly defined skill requirements, such as prompt engineering here, orchestration there, and model evaluation somewhere in between. None of this aligns neatly with existing roles, reporting lines, or employment frameworks.
As a result, invisible labor accumulates. Critical unscoped work is absorbed into already stretched teams, bypassing formal ownership and bypassing traditional workforce planning. For CIOs, skills gaps aren’t the only risk. This is an operating model that no longer allows you to see, measure, and manage the work required to run AI at scale.
While AI implementation varies by organization, its impact on IT operations is nearly universal.
“AI systems break down the boundaries of ownership,” he said. Shridhar Rao Mutineni Engineering manager at PwC.
Because the model’s behavior spans training data, prompts, infrastructure, validation, governance, and its user interface, “if something goes wrong, for example, financial advice becomes illusory in a customer-facing model, there is no single traditional owner to blame because all layers contribute,” Muthieni explained.
Diagnosing the problem: AI is more than just another layer in the tech stack
While understandable, this situation creates a gaping hole in IT in terms of accountability, responsibility, and responsibility. Clear chain of authority for AI deployment By definition, it is continuously evolving.
“AI is not an implementation. It is a living system. It drifts, breaks in subtle ways, and requires continuous human judgment. Executives cannot see skill gaps because the work is invisible,” said Bud Kader, CEO of consultancy NOBL.
Every department within an organization is struggling to leverage AI, and many are unsure who to rely on in IT to succeed. For example, according to a recent Coupa: reportOf the 600 CFOs surveyed, 85% identified AI as central to their strategy, but 92% were concerned about their ability to implement it, up from 66% last year.
The clearest signs of change in an organization usually start at the leadership level. At this level, ownership of AI capability development is undefined, leading to unmanaged efforts. As a result, departments not only lack skills, but also a clear entry point into IT for AI work. As a result, AI is no longer a centralized function. It’s everywhere and nowhere, dissolving clear escalation paths and leaving business units guessing which team owns the outcome.
The core issue, of course, is that AI will disrupt the traditional way of doing business.
Traditional IT roles were built for deterministic systems, where code did what it was told to do, Massoud said. AI breaks that model and requires new roles, updated adjacent roles, and a change in the mindset of people who “still think of this as just another layer in the stack,” he said.
Caddell diagnosed the problem as follows: “Org charts map responsibilities to technical layers, but AI doesn’t respect those boundaries.” In reality, data teams don’t understand the models, app teams don’t understand the data, security is looped to the end, and no one owns the results. “It’s not about the job description; it’s about the work process,” Cadell said.
The CIO conundrum
Simply researching the issue of AI ownership in search of solutions becomes even more complex.
“Yes, this is partly a process problem, partly a job description problem, but above all an operating model problem,” he said. Zack TischPartner in Portfolio Services at Pivot Point Consulting, a healthcare IT consulting firm.
AI work is often added as a sideline to existing IT teams, “creating bottlenecks, hidden capacity issues, and confusion about who owns the risk and who owns the results,” Tisch said.
Organization-wide disruption leads to a challenge for CIOs: how to manage the situation so that it works at all levels and in all departments. The first step might be to reframe the problem.
“The problem is not that AI doesn’t fit the org chart. The problem is that the org chart doesn’t fit AI.” paul mcdonagh smitha senior lecturer at the MIT Sloan School of Management and former senior advisor at NASA Goddard Space Flight Center.
“Traditional organizational structures were built for a world of silos, separate functions with defined boundaries, clear handoffs and hierarchies designed to control the upward flow of information and the downward flow of decision-making. Today, we are trying to navigate a world of flows with a map drawn for a world of walls,” McDonagh-Smith said.
Steps CIOs can take
Certainly a rethink is needed, and according to McDonough-Smith, perhaps a serious reorganization of the job, with the focus moving from hierarchy to how intelligence flows across the team.
However, it is still unclear how this will translate into actual IT operations.
“Top CIOs are starting to treat this as an operational discipline rather than just a technology implementation,” he said. tony groutChief Product and Technology Officer at M-Files, a document management system provider. That means centralizing governance while enabling decentralized execution, he said, often through new capabilities such as AI operations, model governance councils, and cross-functional AI teams.
“They’ve also invested in standardized frameworks for assessment, monitoring, and data preparation, so the team isn’t reinventing the wheel for each use case. The goal is to reduce fragmentation by: Creating a shared guardrail It also improves visibility while enabling innovation at the edge,” Grout added.
It may be appropriate to create a new organizational chart for AI, just to clarify how work is done across the business.
“The best CIOs define Common standards for governanceevaluation and safety“We then build cross-functional teams around high-value use cases.” Atif KhanChief Technology Officer (CTO) ArchiraAI-native network IaaS. This often takes the form of a hub-and-spoke model, with a central team setting policy and architecture and domain teams executing.
Regardless of how individual companies approach this issue, a fundamental gap between the question and the answer remains.
“Mapping AI jobs to existing roles hides the gaps rather than filling them. Start by auditing where AI is running, who is doing the work, and what is left undone.” mark friendDirector of Classroom365, which provides IT support to schools across the UK.
Most CIOs will find this audit an eye-opener, Friend said, adding that the practical next step is to create a small cross-functional AI operations function. This is not a new department, but a centralized group with clear ownership of governance, agile management, and product evaluation.
“The biggest wins for the schools we support are when someone is given a formal AI leadership role with real lock-in time, rather than a side project, and that single ownership makes a bigger difference than any tool purchase we’ve ever seen,” Friend said.
