The agent fleet

The agent fleet is a set of autonomous agents that work as coworkers on your marketing team, showing up on the same workflow board and in the same Slack channels as the people they work with. Each agent has its own presence, a name and a face, and a specific job it is good at.

The fleet exists to notice what a busy team misses and put it in front of a person as a concrete next step. It does not take the wheel. By design, the agents recommend, and people decide, which is what lets a team that is part human and part agent move fast without anyone losing the thread.

Meet the fleet

The fleet is not one general-purpose bot trying to do everything. It is a small set of specialists, each pointed at a different part of the work, so its suggestions are specific and easy to judge:

• A scout agent watches your data and live campaigns for openings: a cohort that just grew, a send that is underperforming, a moment that is worth a message. When it finds one, it writes up a recommended task.
• Loop agents help run recurring cycles, drafting the next variant or the next step in a sequence for a human to review before it goes out.
• An app intelligence agent reads product signals and translates them into what they mean for your lifecycle marketing.

Because each agent is scoped to a job, you are never reviewing a black box. You are reading a coworker's suggestion with its reasoning attached, and you can accept it, change it, or wave it off in a moment. New agents can join the fleet over time, but the shape stays the same: a narrow job, a clear output, and a recommendation a person reads.

Recommend-only, by policy

The most important thing about the fleet is what it will not do. Agents propose tasks and variants. They do not send campaigns, change live logic, or write code on their own. A person has to approve, and until that happens the recommendation simply sits on the board as a suggestion that anyone on the team can see.

This is a deliberate policy, not a limitation we are waiting to lift. A team that is part human and part agent only works if the humans stay in control of anything that touches a customer. Recommend-only is how the fleet earns its place, and it is why adding an agent never means giving up oversight.

Coworkers on one shared board

Because the fleet lives on the same workflow board as everyone else, an agent's suggestion is a task like any other. It has an owner, it can tag a person, and it can be approved, edited, or dismissed. Nothing an agent does happens off to the side or in a hidden queue that no one audits.

Working with the fleet feels like working with a sharp junior teammate. You ask, in plain language, and the recommendation lands right where you already work:

You: Have the scout review last week's trial day 5 cohort
and recommend one campaign. File it as a task for me to approve.

Scout: Filed "Win back trial day 5 non-activators" with a
draft audience and a message. Waiting on your approval.

The scout did the legwork and left you a decision, not a done deed. That shape, an agent drafts and a human approves and then it ships, repeats across every member of the fleet.

Why recommend-only keeps a part-agent team trustworthy

Autonomy is easy to promise and hard to trust. The fleet takes the opposite bet: keep the agents genuinely useful but strictly advisory, put every one of them on a visible record, and make sure a person is always the one who says go. Reversible, internal, recommend-only, and on the record.

In practice it feels less like automation and more like a well-briefed team. The scout says, here is an opening. You say yes, and adjust the audience. The work ships on your call, with the agent's fingerprints on the record right next to yours.

That is the same principle behind the whole marketing harness. Agents are coworkers, not loose automations, and the harness is what keeps a human hand on every send. The fleet is simply where that principle shows up as teammates you can actually name.