You do not remove humans from anything. You move specific kinds of work to AI teammates, so the humans you already have stop spending their week on it. That distinction matters: the version of this pitch that promises you will never hire again is selling something, and you should not buy it from us either.
What a small team can actually do is change which work requires a new hire. Some functions can be staffed by an AI teammate, a named, persistent worker with a role, a memory, and standing routines, long before they would justify a salary. Some cannot, and knowing the difference up front is the whole game.
The work that moves first is recurring and well-scoped
Three tests tell you whether a function is ready to be staffed by a teammate. It happens on a schedule, weekly reports, daily triage, a standing research pass. It has a clear definition of done, a list built, a draft written, a queue sorted. And its output is inspectable, meaning you can look at the result and judge it in a minute without redoing the work.
By those tests, the usual first functions are unglamorous and real. Outreach research and drafting, where the list-building and first drafts are the grind and the judgment call is the send. Support triage and response drafting, where sorting and summarizing eat the day but the hard conversations still need you. And reporting, the numbers someone assembles every Monday that nobody was actually hired to assemble. Each of these has a walkthrough of its own: sales outreach, customer support, and weekly reporting. The same three tests admit the next functions in line, social content, website operations, and the hiring pipeline, not long after.
The work that stays human is the reason your company exists
Strategy stays human. Deciding what to build, when to change course, which market to walk away from, those are bets, and an AI teammate should not be placing bets for you. Relationships stay human. Your biggest customer, your investors, the candidate you are trying to close, all of them can tell the difference, and the difference is the point. And judgment under ambiguity stays human: the refund that is really an apology, the deal term that is really a trust question.
The question is not which humans to replace. It is which Tuesday afternoons to get back.
Teammates in Dock are built around this line rather than against it. When one hits a decision it should not make alone, it escalates upward to whoever it reports to instead of guessing. The system assumes a human is in the loop for the calls that matter, because that assumption is correct.
Start with one function, not a reorganization
The failure mode is trying to staff everything at once. The working pattern is narrower. Your first teammate in Dock is a Chief of Staff. It arrives with routines already running, learns you and the company, and becomes the place your operational load starts to land. From there, you hire specialists one function at a time, each with a single well-scoped job: one for outreach, one for support, one for reporting.
Each specialist starts small, with one routine, and takes on more as the work proves out. Hiring one costs you a decision, not a recruiting pipeline, a notice period, or a replacement search when it does not work out. That is the actual economics here: the cost of trying is low enough that you can staff functions years before they would justify a human hire.
What this looks like at five people
A five-person company running this way does not look like a five-person company with fewer meetings. It looks like a bigger company's org chart with the recurring boxes filled differently. The founders still sell, still decide, still handle every conversation where trust is on the table. But the Monday report writes itself before the meeting. The support queue is triaged before anyone opens it. The outreach list for next week is researched and drafted, waiting for a human to approve the sends.
Nobody was replaced in that picture. The company just stopped needing a sixth and seventh person for work that never deserved a whole person in the first place. Whether that means you hire more slowly, or hire the same number of people into better jobs, is your call. The tool does not make that promise. It just moves the work.