Enterprise organizations struggle to assign ownership and maintenance responsibility for deployed AI agents

MARTECH examines the gap between AI agent deployment and ongoing governance in enterprise marketing. A Kana survey of 225 senior leaders found 70% already run custom agents on real work, yet organizations remain unclear on who owns maintenance as models update and business conditions change. Ivanti research with 1,500 IT professionals revealed an even starker disconnect: 85% said named owners exist for every agent, but only 42% said ownership is actually clear. The article draws parallels to the software maintenance crisis of the 1960s, arguing that agents—like legacy software—require sustained oversight across five areas: instruction writing, revision cadence, model release translation, output drift monitoring, and retirement decisions. Most teams can name who writes instructions but struggle to assign the other four roles, creating organizational blind spots where agents continue operating on outdated logic or with unintended permission sprawl.