Marketing teams urged to document context before deploying AI agents

src-61f867b2

Marketing teams looking to use AI agents for research, reporting and drafting need to do more than automate individual tasks. They also need to document the reasoning behind brand decisions, map the full workflow and build in human review, speakers said during a session at a November 2026 marketing conference.

The discussion brought together Stacey Ackerman, an independent consultant and fractional CMO, who moderated, with Ed Poppe, founder and fractional marketing leader at Poppe Marketing; Sean O’Connell, head of customer success at Screendragon; and Lisa Shaw, founder of Devon Point Group. An audience poll found that 51% of respondents said their organizations were actively piloting AI, while none said they were avoiding it.

The speakers focused on a practical challenge for marketing operations: AI can help teams handle more work, but it does not automatically know the history, constraints or strategic intent that shape a good decision. Without that context, organizations risk producing faster work that still requires substantial correction—or overlooks important brand and customer considerations.

Capture the reasoning behind the work

Shaw said she uses AI agents to coordinate tasks and compile reporting data, allowing her to devote more attention to strategy and relationships. But the usefulness of those agents depends on information that may otherwise remain in employees’ heads: brand nuances, past customer friction and the rationale behind earlier choices.

She documents that context and incorporates it into her agents’ instructions. The distinction matters because a tool may be able to follow a task description without understanding why a particular phrase is unsuitable or why a customer situation calls for care. As Shaw put it, “I think the why is just as important as the what.”

Poppe connected the practice to a familiar marketing discipline: writing a strong creative brief. Teams should set out the desired deliverables, explain brand guardrails and identify checkpoints for review, he said. Those materials give AI systems direction while preserving the human decisions that determine whether the work is appropriate.

That documentation also addresses the risk of institutional knowledge disappearing when employees leave or roles are not replaced. In lean teams, recording decision context can help remaining colleagues and AI tools work from a more consistent understanding of the organization.

Redesign workflows before automating tasks

O’Connell cautioned against treating an agent or automated task as a complete workflow. A workflow describes the end-to-end route work takes; agents can handle repeatable steps within it. Automating one step without examining what comes before and after may simply shift a bottleneck to another team.

Organizations are experimenting at different levels, from individual users to enterprise teams with agents in production. A central challenge, O’Connell said, is connecting isolated pilots into a coherent operating system. His advice was to identify the operational bottleneck and define what success means before selecting a tool: “Start with the process, not the technology.”

Ackerman made a related point: adding AI to a flawed manual process can magnify existing friction rather than resolve it. Teams should understand how work currently moves, where it gets delayed and which decisions require human judgment before deciding where automation belongs.

Cross-functional involvement is another part of making pilots useful beyond the team that starts them. Poppe encouraged marketers to share a working pilot with colleagues in sales, operations or creative functions. A concrete example can help partners understand how the approach works and encourage collaboration, rather than leaving experiments disconnected from the rest of the organization.

Keep judgment and governance in the workflow

The speakers’ recommendations point to a measured approach: use agents for defined, repeatable work, but make the surrounding process and decision criteria explicit. Briefs, documented brand rules, historical context and review checkpoints can help teams maintain quality as they increase output.

For marketing leaders, the operational question is not simply which tasks an agent can perform. It is whether the organization has captured enough context to delegate those tasks responsibly—and whether the workflow gives people a clear role in checking the result. The poll suggests AI pilots are already widespread among the audience, making process design and knowledge capture immediate considerations for teams seeking to turn experimentation into sustainable practice.