The best first AI employee is boring. It does one job you already do every week, well enough that you stop doing it yourself. Here is how to get there in an afternoon.

Step 1: choose an employee

Start from an available template or create a role for a job you already have in mind. A good first job has three properties:

• You do it on a schedule (weekly is ideal).
• The input is something you already have written down: notes, a doc, a spreadsheet, an inbox.
• You can tell in under a minute whether the result is good.

A weekly update from your project notes fits all three. So does a competitor round-up or a first draft of a newsletter.

Step 2: make it yours

Add your instructions, select a model, and connect the tools the job needs. Keep the brief short and specific. This one works:

Prepare a weekly update from my project notes. Highlight decisions, open questions, and next steps.

Connect the tool that holds the notes, whether that is Notion, Google Drive, or a Slack channel. Pick the model you already pay for. Nothing else is needed for the first run.

Step 3: review and repeat

Start a conversation and hand over the first assignment. Here is the kind of reply to expect:

Decision: start with the onboarding flow. Open question: who owns the content review? Next step: confirm an owner before scheduling the work.

Now review it the way you would review a new hire's first week. Are the decisions real decisions? Did it invent an owner? Confirm the follow-ups yourself the first few times. When the update is right more often than not, set a schedule and let it become a routine.

What to hand over next

Once the first routine is running, the next candidates are usually obvious: the numbers you check every morning, the support replies you draft from the same three paragraphs, the research you do before every sales call. Each one becomes its own employee with its own brief, on the same canvas.

One job, one employee, one useful assignment. Build from there.