The deliverable
A working playbook covering your marketing function, task by task. For each task it says where the work goes today: ready for an agent, still requiring human judgment, or part of the orchestration layer that connects the two.
The format is a 90-minute working session followed by a one-week deliverable. The standard we hold it to is that your team can use it on Monday.
Why task level
Most AI conversations inside marketing teams stall at the wrong altitude. The question gets framed as whether AI can do content, or whether AI can do campaigns. At that altitude the answer is always the same unhelpful shrug: partly.
Break the function into its actual tasks and the question becomes answerable. Drafting a first-pass campaign brief and deciding which segment to enter are both called strategy in casual conversation. One is a task an agent handles well today. The other is not close.
The reference model
We have already decomposed the modern marketing function into two hundred discrete tasks, and classified each one. One hundred and thirty are suitable for AI agents today. Fifty-five still require human judgment. Fifteen are the orchestration layer.
Your workshop starts from that reference model rather than from a blank page, which is most of why a 90-minute session is enough. The work in the room is deciding which of the two hundred apply to your business, which ones you are actually doing today, and where your classification differs from the reference.
Your version will differ. A regulated business moves tasks toward human judgment. A company with a mature data function moves more toward agents.
The orchestration layer is the part people miss
Fifteen of the two hundred tasks are neither AI work nor human work. They are the connective tissue: routing, quality gates, handoffs, deciding what happens when an agent produces something that is not good enough.
Teams that skip this layer end up with a collection of tools that each work and a function that does not. The playbook covers it explicitly, because it is the difference between having AI in the marketing team and having a marketing function that runs on AI.
What it is not
It is not a technology recommendation, and it is not a procurement exercise. We are not selling software and we are not reselling anyone else’s.
It also is not a headcount plan. The classification tells you where work can move. What you do with the capacity that frees up is a business decision, and in most engagements the answer has been to do more of the work that was never getting done rather than to remove people.
Who it suits
It works best for teams with a functioning marketing operation and a real question about where AI fits. Fixed scope, so it is also a low-commitment way to see how we work before a longer engagement.
It is a poor fit if the positioning question underneath is unsettled. Decomposing a function that is executing against an unclear strategy produces a faster version of the wrong thing.
How it connects to the rest
The workshop answers one of the three questions the practice is built around: where AI should lead. The other two are whether buyers can find you, and whether you are setting the meetings you need. Most engagements start with one and move to the others once the first is answered.
See the full task breakdown first
Two hundred marketing tasks, classified. The method and the numbers are published.