Marketing Organizations Rethink AI Agent Ownership As More Brand Work Changes Hands

An AI agent can now process a week of reporting, read through a year of customer reviews, or rewrite product copy across an entire catalog. The harder questions for a marketing team are which of those jobs to hand over, what an agent does once it has one, and which function owns the result. At many companies, decisions about AI sit with the head of technology, and marketing has little say in the ones that shape its own work.

In one survey of senior marketing leaders, Gartner found that 65% expected AI to dramatically change the CMO role within two years. The marketers already running agents describe a short list of jobs those agents handle well, and a clear line around the work that stays with people.

From reporting to product copy

Reporting is usually the first job to move, since the data an agent needs already sits in a system it can read. Carla J. Donahue, Sr. Director of Lifecycle at WITHIN, has her agency stand up a separate model for each client so the reporting runs against that client's full history. "We really use AI as a third business partner," says Donahue. "We set up individual LLMs for each of our clients so that we have a full history and data. It can process all the reporting."

Customer insight is harder to hand over, since the raw material arrives as loose text and someone still has to decide which themes are worth acting on. Caitlin Berzok, SVP, Digital at Therabody, stitched review data and survey responses together and set mention thresholds so a model reports how deep a theme runs. "It'll save me what would have taken hours and hours to go through and seep through all these customer reviews," Berzok explains. "I can then run it through Claude or a different model and get a synthesis and set parameters of it, of if it's been mentioned 10 times or if it's been mentioned 20 times, what's the depth of these trends?"

Product copy is where volume tends to settle the argument, since a large catalog needs refreshing faster than a writing team can keep up with. Alicia Littleton, Director of Ecommerce at Gerber Childrenswear, kept that work in-house for years before the catalog outgrew what her internal writers could refresh. "There's a lot of layers on top of content creation that we have, for the longest time, said we're going to do internally, with our own copy and content creation process and internal writers," Littleton notes. "It's not realistic anymore, not at the rate that we need to update content at the scale of the amount of products that we have."

Beyond the head of technology

Once an agent writes copy or messages that customers read, the brand carries whatever it gets wrong. Liz Villani, Founder of #BeYourselfAtWork, treats that risk as the reason ownership has to move. "With the head of AI or head of technology is passively where this work sits," says Villani. "But the minute that AI represents you, everything changes there."

Assigning an agent to one team gets harder when it does work that used to pass through several. Chris Bradley, Chief Marketing Officer and Head of AI Transformation at Veritiv, had separation of duties built into his processes as a safeguard. "We always had a lot of separation of duties, almost for audit reasons, where you have to do this handoff to this handoff to this handoff," Bradley says. "But when we have complete audits of what the agents are doing, we have the same agent do processes that would happen across multiple functions and groups."

One answer is to stop assigning the work to a single function. Rohit Prabhakar, Global Head of Digital Experiences and Capabilities at Visa, argues for a shared mandate with one measure behind it. "Experience needs to become a mandate that only marketing should not own, or a technology organization should not own," Prabhakar adds. "It has to be a mandate that the commercial organization, marketing, data and technology need to own together, with one single KPI that they all have to deliver towards."

Briefs, builds, and freed hours

Deciding which agents a team needs also means drawing a line around the work that stays with people. Athena Kasvikis, Founder and CEO of Behave Bras and most recently CMO of iRobot, names two jobs she won't hand over. "I'm not going to use AI to write a creative brief for my lead agency to do my lead creative," says Kasvikis. "I'm not going to use it to make the decision whether we move forward with a product or not."

Building an agent takes someone who understands the business problem and knows the technology well enough to build for it, and most marketing teams don't have that person on staff. Steve Blackford, Chief Marketing Officer at PrivacyHawk, describes a role that doesn't exist on most teams yet. "It's the person who can understand a business need and understands AI enough to say, 'I can go build you an automated flow for that,'" Blackford explains. "It's almost that go-between of, 'I can understand what the business needs, but I understand how to go make the tools do it.'"

Every job an agent takes over frees up hours the team then has to spend on something else. At Veritiv, Bradley has put a number on how many one agent gives back. "We're going after a million orders that we manually key every year, that takes about six minutes an order on average," Bradley concludes. "That's 6 million minutes of human time that we're going to repurpose to serving our customers better. We're trying to take the manual work out of all of these functional jobs and put humans on to higher-level work."