For about a decade, CIOs have been asking for a seat at the table. In 2027, many will finally get one, along with a stack of AI promises somebody else made and a few hundred agents nobody remembers hiring. The next phase of AI comes with a less glamorous problem: figuring out who owns what when something goes right, and especially when it doesn’t. Who owns the data? Who owns the agent? Who owns the return? Most organizations assume those questions have already been answered. Then a deal stalls, an agent makes a decision nobody expected, or a CFO asks where the value went, and everyone discovers the ownership conversation was still waiting to happen. Our Predictions 2027: Tech Leadership report explores where those conversations are headed. Here are three of the five predictions that stood out for me. 

  • A major enterprise will publicly walk away from an acquisition because of data debt. Bad data used to be a post-close problem. The deal closed, the integration team opened the hood, and somebody spent 18 months reconciling customer records that couldn’t agree on what a customer was. AI politely insists you discover the problem before wiring the money. If the growth story depends on AI, buyers need to understand who owns the data, whether it can legally train models, and what it will cost to make it useful. Then, somebody prices the cleanup (usually after the board has picked a font for the press release), and the math falls over. With only 22% of AI decision-makers  reporting an EBITDA lift from AI in the past 12 months, paying a premium for an AI story with a cleanup bill stapled to it gets hard to defend. 
  • One in four CIOs will create an agent operating model office. Agents are multiplying faster than most organizations can update the org chart. Ask how many agents are running and you’ll often get three different numbers and one person asking what you mean by “running.” Governance boards, enterprise architecture, and AI centers of excellence each own part of the problem. None was designed to supervise a workforce built on agentic systems that never takes PTO and occasionally approves a purchase order at 3 a.m. Eventually somebody has to own decision rights, autonomy levels, and escalation paths, because “the committee” has never once appeared on an incident report. Start with a plain inventory of every agent, its owner, and what it can touch. If producing that list turns into a workshop, you already have your answer. 
  • More than half of CIOs will miss their AI-related ROI targets in 2027. 45% percent of AI decision-makers expect a positive return within a year. Yet fewer than 30% offer AI architecture training, and fewer than 30% of data and analytics leaders plan to focus on AI-related change management in the next 12 months. The hard part is doing the organizational work around the technology (not the tech itself), including governance and scenario planning. So, the money buys models and platforms while process redesign and skills development get penciled in for “phase two,” which is corporate speak for never. The pilots will keep crushing it in the demo because that’s what pilots tend to do. The conversation changes when the CFO asks for the firmwide number and the room suddenly gets very interested in its coffee. 

The Seat Came with Fine Print 

CIOs are getting pulled into decisions that used to happen somewhere else: whether a deal closes, how much autonomy an agent receives, and whether AI investments continue to earn funding. In many organizations, the accountability arrived before the authority, leaving technology leaders responsible for outcomes without always having a clear path to influence the decisions that drive them. 

When ownership questions remain unresolved, organizations eventually answer them the same way they answer most operational problems: whoever is standing closest gets handed the issue. A failed acquisition, an agent incident, or an uncomfortable ROI review tends to settle governance debates quickly, even if nobody likes the answer. 

Data ownership, agent ownership, and value ownership all sound straightforward until the organization has to prove it knows the answer. That’s when the discussion moves from planning documents and steering committee updates to real decisions about accountability, risk, and investment. By that point, somebody usually discovers they owned the problem all along. 

Forrester clients can read the full Predictions 2027: Tech Leadership report and schedule a guidance session with me to discuss what these predictions could mean for their technology strategy. If you’re not yet a client, visit the Predictions 2027 Hub for complimentary guides, blogs, and webinars. And if your current AI governance plan still includes the phrase “we’ll sort that out later,” there’s a good chance you’re already living one of next year’s predictions. 

Related Links 

Related Forrester Content 

Share