across tasks like creative-content development, sales-collateral development, e- commerce/web optimization, and comarketing with business partners. Some marketing technology platforms, including Adobe and HubSpot, now offer AI agents that can be embedded directly into creative workflows. These agents can generate and refine copy and design variations, tailor assets to audience segments, and update content across channels based on real-time behavioral signals. Marketers remain responsible for brand integrity and strategic guidance, but the agents orchestrate much of the ongoing production work. Early pilots show shorter production cycles and an increased ability to respond quickly to changing market conditions. Step 4. Define future-state workflows with clear roles for humans in the loop Of course, as AI agents are increasingly inserted into workflows, human roles will need to change. In marketing, that will mean focusing more time on tasks like developing marketing strategies based on qualitative factors like “taste” that are not prone to automation; developing a deeper understanding of what will resonate with audiences; sustaining and building relationships with stakeholders; and engaging on tasks best handled in person, such as marketing activations. Marketers will also need to oversee the technology infrastructure powering these workflows: data quality and schemas, content metadata, orchestration rules, and API governance that ensures agents operate safely and consistently. This will require brands to invest in talent capable of fine-tuning off-the-shelf foundation models to brand context and upskilling human employees to redefine ways of working. Among the new skills humans will need to master: — prompt engineering : knowing how to structure instructions so agents can produce desired outputs — collaborating with agents : understanding handoffs between agents and marketers, and steering agents to formulate new strategies — quality monitoring : ability to monitor agent activity, spot anomalies in quality, compliance, and so on, and track agent tasks — refining ideas with human expertise : assessing and enhancing AI outputs with human instinct and experience — data and AI fluency : ability to prep and clean datasets and validate AI-generated insights against real-world performance — machine learning modeling : knowledge of applied machine learning, data engineering, experimentation, and workflow orchestration Consider the concept generation and testing workflow at the consumer brand cited above. The future-state agentic process the team created includes squads of agents that collaborate with Reinventing marketing workflows with agentic AI 7