Most of my marketing reading this week described the same workflow, and it kept stopping at the same place. AI does the research, the drafting, the routing, the follow-up. Then a person takes over for the last step.
A solo operator writing at The Indiepreneur laid this out plainly. He runs a six-figure business with an orchestrator agent and about ten channel agents underneath it, covering SEO, paid media, and lifecycle email, each tied to a share of a monthly target of 900 signups and 200 customers. The agents draft work and route it onto a shared board. The one rule he does not break is that they never send. He approves, and he sends. That is the whole design. The machine gets everything up to the moment the work goes out under his name.
Kieran Flanagan described the same line from the org-chart side. He sees marketing teams reorganizing into what he calls 0-1 marketers, people who take an idea to shipped output with fewer handoffs, working in small pods that own an outcome and ship weekly. His argument is that as AI handles more of the production, depth of craft gets more valuable rather than less, because someone still has to decide whether the output is any good. The production gets cheap and the judgment does not.
There was also a story going the other way, and it is worth sitting with. 404 Media reported that Meta tested an AI calling feature that was, in part, humans in a call center. Internally it was described as a "human agent layer," and at least one tester only realized after the call that a person had placed it. We have gotten so used to labeling things AI that a company hid the human even when the human did the work. I think that is the tell. The question is no longer whether AI is in the workflow. It is where the person is, and whether anyone can find them.
The send button keeps coming up because it is not really a technical limit. Agents can send email. The reason the operators I read hold that step back is that sending is the moment of accountability. Approving is a decision that this is good enough to go out under our name, to this client, this list, this reporter. You can hand a machine the drafting and keep the deciding, and most of the people who are actually running on AI seem to be drawing the line in exactly that place.
None of this comes from a survey, so I would not read it as a trend with a number attached. It is three people describing how their own work is arranged, which is a smaller and more useful thing. This fall I am talking with a set of Vancouver marketing practitioners about where AI has actually landed in their workflows, and this is one of the questions I want to put to them. Not what they have automated, but what they have decided to keep.
Sources
- The Indiepreneur, "I built a growth team out of agents": theindiepreneur.substack.com
- Kieran Flanagan, "Is the traditional marketing team dying, shrinking, or turning into something else?": kieranflanagan.io
- 404 Media, "Meta tests Muse AI agent calls that are actually made by humans in a call center": 404media.co