Step 1: Create a detailed taxonomy of key marketing activities Creating tomorrow’s agent-driven workflow, of course, cannot be done without first developing a granular understanding of the way work gets done today. An important first step in that process is to break down priority workflows into the full chain of key activities involved. This mapping must include the underlying systems—customer relationship management, content management systems, digital asset management, analytics, and data pipelines—that support each activity, since system constraints often shape how agentic workflows can be designed. This will serve as the foundational current state that eventually will be translated into the future- state “clean sheet” agentic workflow. This is how many companies across industries have begun. Take, for example, one leading consumer brand that sought to redesign the process of creative ideation and production. Historically, this was an often complex undertaking that could take months of effort, with numerous stakeholders, both internal staffers and outside agencies, engaged in iterative cycles of feedback and rework. To determine how AI agents might help, the organization first created a comprehensive list of activities involved in the process, encompassing ideation, concept creation and testing, content production, content versioning, content optimization, and agency management. Those activities were then further broken down into hundreds of individual microtasks. Within concept creation and testing, for example, the team identified subtasks like concept image generation, pretesting with focus groups, assessing risk, and more. This detailed taxonomy provided executives with a more comprehensive understanding of its workflows—an understanding that later informed the build-ready specifications for agents. This taxonomy should also include the insights function within marketing—activities such as synthesizing data, generating hypotheses, interpreting consumer signals, and translating findings into action. These activities form a critical part of the marketing process, and many can be augmented or accelerated through agentic workflows without replacing the human judgment required to make meaning from them. Step 2: Define agent archetypes After establishing a baseline understanding of organization-wide tasks, the next step is to classify these tasks into agentic archetypes, which will serve as reusable blueprints to guide where and how agents are deployed within workflows. In marketing organizations, some of those archetypes might include “extracting knowledge to build context and reasoning,” “analyzing data to define outputs,” and “generating materials across mediums with variations.” Leaders at the consumer brand above, for example, classified scores of marketing tasks into six agentic archetypes—content generator, knowledge, localization, analyzer, planner, and operator—which were subsequently used to define the modular, scalable individual agents to be deployed and reused across the marketing process (Exhibit 3). Reinventing marketing workflows with agentic AI 5