Marketing clouds and ad platforms now sell agents: AI that can plan, build, launch and optimise parts of a campaign with limited human input.
What vendors offer
At Adobe Summit in March 2025, Adobe launched Adobe Experience Platform Agent Orchestrator to manage AI agents across Adobe and third-party systems, with agents aimed at jobs such as optimising websites, handling content production tasks like resizing, refining target audiences, running experiments and cleaning data. Other marketing clouds and ad platforms offer similar agents for audiences, journeys, content and budget.
What agents are good at today
- Repetitive, rules-based steps: building audience lists, resizing and tagging assets, setting up tests, QA checks, reporting.
- Bounded optimisation within limits you set (budget ranges, approved creative, approved audiences).
What they are not good at yet
- Strategy, positioning and creative ideas.
- Judgement calls with brand, legal or reputational risk.
- Working reliably across messy, disconnected data.
Hype check
In June 2025 Gartner predicted that over 40% of agentic AI projects will be cancelled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls. Gartner also warned of "agent washing" — rebranding existing assistants, automation or chatbots as agents — and estimated only about 130 of the thousands of agentic AI vendors are real.
How to govern a campaign agent
- Scope: what it may do (build, launch, change budget) and what it may not.
- Limits: spend caps, approved assets and audiences, blocked topics.
- Human checkpoints: approval before launch and before big budget moves.
- Logs and kill switch: every action recorded; one person can stop it.
- Measured value: a baseline and target before you start.