A colleague at a major utility is releasing 200 AI agents next quarter. Most of them are being built outside of IT, in places like marketing. When he was asked who owns the agents after they ship -- who handles maintenance, model updates, and the day the model is deprecated -- he did not have a good answer. He is not unusual. The Kana research from May showed 70 percent of large enterprises already run custom AI agents on real marketing work. Kana also found that about 40 percent of senior leaders expect the chief AI officer to own those agents, while marketing executives tend to point at their own function or a shared model. In most companies right now, two capable groups are each assuming the other one has it.

The 43-point ownership gap nobody is closing

Ivanti surveyed 1,500 IT professionals in February and March about agent ownership. 85 percent said a named owner exists for every AI agent in their organization. Only 42 percent said ownership is actually clear. That 43-point gap is the defining governance problem of agentic marketing in 2026, and it is not closing on its own. The same research found 65 percent of organizations run a review before an agent is deployed. Governance shows up for the launch. Then oversight shifts to a quarterly rhythm while the agent keeps working every day.

There is a related habit worth checking in your own stack. Ivanti found that permission sprawl starts on day one, because organizations spin up agents by cloning a human user profile. In marketing, that means your campaign agent may be walking around with the CRM access of whoever set it up. That was fine when it launched. The question is whether it is still fine now, and who would notice if it was not.

Split the question or it stays unanswered

Both camps have a real argument. The centralizers will tell you that agents handle customer data, carry regulatory exposure, and that no single function can be trusted to police its own output. The marketers will tell you that nobody in a central AI group knows whether an agent tone is right, whether the offer is current, or whether the segment logic still matches how you go to market. Both are true. The way through is to split the question. Central owns access, data, and the model layer. Marketing owns instructions, tone, and whether the agent is still saying something true. Both sides need names attached, and right now most organizations are leaving it fuzzy.

What the software lifecycle already taught us

From the 1940s into the 1960s, code was treated as something you wrote once and then put on a shelf. At the 1968 NATO conference in Garmisch, engineers from a dozen countries compared notes and found they were all stuck on the same problem -- software gone out of date. What emerged was the lifecycle we still teach today: requirements, design, build, test, deploy, maintain, retire. Maintain and retire were added because the industry learned the hard way that software does not maintain itself. Bennet Lientz and Burton Swanson studied 487 organizations in 1980 and found maintenance consumed roughly half the software budget. The largest category was perfective work, meaning the requirements had changed. Adaptive work, meaning the environment around the software had changed, came next. Fixing defects was the smallest of the three. Most of that work came from the world moving around software that was written correctly the first time.

Now read it with an AI agent in mind. Your content agent was written in March against a March offer and positioning. Your SDR agent learned an ICP that predates the pricing change. Your brand voice guardrails were tuned against a model version that got deprecated over the summer. None of that is a defect. It is maintenance, and someone needs to say what changed and what to do about it.

What to do this quarter

Three things, in order. First, take the inventory you have been avoiding. You do not need a fancy tool for this. Ask every team that built an agent in the last 12 months to fill out a one-page form with the agent name, who built it, who maintains it, what data it touches, and what its retirement trigger is. The form will be harder to fill than you think, and that is the point. Second, draw the line between central and function ownership before the next agent ships. Central gets access, data, and the model layer. The function gets instructions, tone, and the right to retire the agent when it stops being useful. Write it down. Third, give every agent an owner with the word "responsible" in their job description, not just the word "familiar with." If nobody is on the hook for the agent drifting, the agent will drift.

The 200 agents your colleague is shipping next quarter are going to do real work. Some of them will touch real customers, real spend, and real regulated data. The question is not whether your organization has owners on a slide deck. It is whether someone on the org chart is responsible for the agent on a Tuesday in November when the model underneath it changes and nobody notices.

Sources