AI is rapidly becoming an enterprise-wide capability, but many organizations still struggle to answer the basic question: “Who owns the AI?”
Unlike previous technology initiatives that naturally belonged to IT, AI is impacting nearly every business function. Contact centers, security teams, data organizations, line-of-business leaders, and IT departments are all claiming rightful ownership and creating governance models that often fragment AI programs before they reach scale.
The result is an organizational challenge that goes beyond technology. How companies divide responsibility for AI strategy, governance, and execution will influence whether AI becomes a coordinated business function or another layer of operational complexity.
Business units drive AI
One reason ownership has become so difficult is because AI has fundamentally changed who can build technology.
“Unlike previous waves of enterprise technology, AI is designed to be accessible,” said Juan Jaysingh, CEO of agent workflow platform Zingtree. “Everyone will be able to use natural language to write, analyze, build, code, and automate tasks that once required specialized technical expertise.”
Accessibility means that AI adoption is no longer focused on the IT department.
“Marketing, sales, finance, customer service, human resources, and operations are all deploying AI to solve business problems, blurring traditional ownership lines,” Jasin says. “The challenge is not deciding who owns AI, but recognizing that everyone owns it.”
Alessandro Perilli, vice president of enterprise AI strategy at IDC, says he’s seeing similar changes, although the impact on governance is increasing.
“Ownership of AI initiatives remains a challenge as the CIO’s office and business units within an organization are driven by very different incentives and move at different speeds,” Perilli said.
He explains that while business units need to achieve financial results, the CIO office needs to manage risk, and the disconnect is becoming increasingly measurable.
According to IDC’s June 2026 CIO Survey, 68% of agent AI investment decisions are currently led by business leaders, with only 4% owned by AI centers of excellence.
Perilli said that as many business units build their own AI agents and deploy third-party tools outside of traditional IT oversight, some organizations are adopting a zero trust approach to controlling who can build and deploy AI.
Governance must not become a bottleneck
As AI spreads across the enterprise, organizations are looking for governance models that enable innovation without slowing it down.
Jasin believes the most successful organizations reposition their governance teams from gatekeepers to enablers.
“The most effective model is one that puts AI governance at the center, whether it’s a dedicated department or shared core values and responsibilities across the organization,” he said. “The goal is for teams that traditionally owned technology and compliance functions to become enablers rather than gatekeepers.”
It starts with expanding AI literacy across the workforce before enabling widespread adoption.
“This starts with upskilling employees and improving the baseline of AI knowledge, including capabilities, risks, and limitations across the organization,” Jasin said.
Organizations must methodically provision systems and data access to help teams securely deploy AI solutions.
Perilli says there is no single organizational model that will emerge as the standard, as each company balances innovation and risk differently.
“It depends on the company’s priorities: speed of innovation or risk mitigation,” he said.
Organizations that prioritize speed may give business units responsibility for AI budgeting and delivery, while leaving governance to the CIO organization and best practices to an AI center of excellence.
According to Perilli, 68% of organizations still don’t have a fully operational AI center of excellence, but those that do are moving significantly more AI pilots into production.
Standardization enables scale
This does not indicate that centralized control over all AI initiatives is complete. Instead, governance should be seen as creating common operating rules while allowing individual business units to innovate within them.
Perilli says organizations can reduce friction by establishing pre-approved budgets and low-risk AI use cases that don’t require additional review.
“If a company decides to implement centralized governance, the best way to alleviate bottlenecks is to pre-approve budgets and low-risk usage patterns,” he said.
Once the business value is validated, “a resident automation team can refine and scale the initial implementation and take it to production scale in stages.”
Jaysingh points to another governance challenge that is becoming increasingly important: cost visibility.
“Governance committees will also need to closely monitor consumption as enterprise AI moves to pay-as-you-go and flat-rate pricing in 2026,” he said.
Whether organizations access frontier models directly or leverage AI through applications, usage-based pricing is flashing back to the skyrocketing cloud prices of the past decade.
Without visibility into AI usage, organizations risk compliance and security issues as well as financial surprises.
“Building visibility and discipline now is what separates organizations that sustainably scale AI from those that are blindsided by high bills, compliance violations, or security failures,” Jasin said.
Coordination is more important than ownership
As AI moves from experimentation to everyday practice, it may become less important to identify a single owner and more to share responsibility among all parties.
Perilli argues that divided ownership is not necessarily problematic.
“Ownership of AI initiatives can remain ‘fragmented’ in the sense that there is a clear separation of duties and all stakeholders are aligned with the same priorities,” he said.
The real danger emerges when that alignment is lost and unaligned ownership fragmentation leads to slowed innovation, alienation of the workforce, loss of talent, uncontrollable risk profiles, and an inability to scale.
Still, he cautions against assuming that today’s organizational structures will remain appropriate as AI continues to evolve.
“The jury is still out on the best organizational model,” Perilli said. “It is too early to declare an appropriate role for accountability in AI strategy, governance, and execution.”
He points out that AI is changing and changing the way organizations work in many ways.
“It is dangerous to assume that the organizational models that have worked for the past 20 years will still work in this new era,” Perilli says.
