What is agentic marketing?
Agentic marketing uses AI systems that can take steps towards a goal and choose how to use tools within their available permissions.
An agent differs from a workflow whose sequence is fixed in advance. Anthropic’s engineering guidance makes that distinction between model-directed activity and predefined orchestration. In practice, a marketing process can combine both. Anthropic’s explanation of agents and workflows
The label does not establish how autonomous a system is. An agent that drafts a plan has different authority from one that can alter a live campaign or commit expenditure.
Which parts of a campaign can agents handle?
Agents can support research, drafting, asset adaptation, campaign setup and reporting when the required information and tool connections are available.
The scope depends on the system’s capabilities and permissions. Access to an advertising account does not by itself establish that every action in that account is supported or authorised.
Campaign work also spans different timescales. Source information, approvals and results can change between stages. A continuing process therefore involves more than a long conversation with a model: it includes the state of the work and the decisions already made.
How does our agentic campaign model work?
Our campaign development model comprises human brief, agentic work, agentic review and human finalisation.
The human brief defines the business problem and relevant constraints. Agentic work develops the analysis and creative material. Agentic review examines sources, reasoning and weaknesses before people finalise the work and make the consequential decisions.
This is our campaign method, not the definition of every agentic campaign. The depth and placement of review depend on the assignment. A model reviewing another model can help identify problems, but cannot establish correctness simply through agreement.
Who is responsible when AI agents produce advertising?
Responsibility remains with the relevant people and organisations under the applicable rules and contracts. An AI system is not a transfer of accountability.
CAP’s June 2026 guidance explains that advertisers remain primarily responsible for compliance when automated AI platforms generate or distribute their ads. CAP’s guidance
Operationally, permissions determine which actions an agent can take and which require approval. Records of inputs, changes and approvals help establish what happened. The appropriate controls depend on the consequences of error, including incorrect claims, unauthorised publication or unexpected spend.
Legal note: This answer provides general information, not legal advice. Seek advice from qualified legal counsel for your circumstances.