Guide · AI loops
Run Your First AI Loop: Marketing That Repeats Itself
One piece of your marketing that runs itself every week — done by an agent on a schedule, checked by you before anything ships.
Most AI advice tells you to "use ChatGPT more" — which still means you doing the work, one prompt at a time. A loop is different: you set up the work once, and it repeats on a schedule without you starting it. In TeamDay a loop is three plain things: an agent that does the work, a schedule that triggers it, and a review step where you approve the result. This page shows you how to run your first one.
How it works
- 1
Understand the three parts
An agent is an AI teammate with one job — Iris designs, Markus runs ads. A schedule is when the work happens — every Monday at 9am, without you asking. A review gate is the safety net: work stops and waits for a human decision before anything is published or spent. In TeamDay this whole package is called a mission.
- 2
Pick one small, repeating job
Start with something you already do (or guiltily skip) every week. A good first loop: every Monday, Iris makes five fresh ad image variants for your current offer. It is real work, it repeats, and a bad week costs you nothing because nothing ships without you.
- 3
See a full week in motion
Monday 9am: the mission triggers and Iris generates five ad variants from your standing brief. Monday lunch: you open the chat, approve two, reject three — thirty seconds of taps. Monday afternoon: Markus preps the approved images as campaign changes for Facebook and Instagram, and shows you the exact changes before anything touches a live campaign. Friday: you glance at what ran and tell Iris what to try next week. Your total involvement: a few minutes.
- 4
Know what a harness is (ten-second version)
Under the hood, each agent runs on a harness — the engine that connects it to an AI model, like Claude or GPT. You can see the options at /harness, and picking one is a dropdown, not a technical project. The default choice works; the point is you are never locked to a single AI company.
- 5
Why the review gate makes this safe
Loops in TeamDay can chain steps — one agent does the work, a check runs before the next step continues. The version that matters for you: publishing and spending always sit behind an approval. If you skip a week of reviews, work piles up politely in chat — it does not ship itself. That is the difference between automation you trust and automation you babysit.
Who does it for you
Real work, not demos
- Weekly SEO report — a real loop artifact, produced on schedule
- Monthly marketing readout — a recurring readout from a live loop
Frequently asked questions
What is an AI loop in plain terms?
An agent, a schedule, and your approval. The agent does one job (like making ad images), the schedule triggers it automatically (like every Monday), and the approval step means you check the result before anything is published or any money moves. TeamDay calls this a mission.
Can the AI publish things or spend money without me?
Not in the loops described here. Publishing and ad changes sit behind approval steps: Luna's posts wait in a queue for your okay, Vince uploads YouTube videos as unlisted until you flip them public, and Markus shows you exact campaign changes before touching a live ad account.
What is a good first AI loop for a small business?
Pick the smallest weekly job you keep skipping. Good starters: five fresh ad image variants every Monday (Iris), a drafted social post queue for the week (Iris + Luna), or a weekly newsletter draft (Mara + Iris). One loop, one agent or two, one approval — expand only after it has run smoothly for a few weeks.
How much time does running a loop take me?
The setup is one conversation. After that, your job is reviewing: a few minutes per cycle to approve or reject what the agents produced. The work happens on schedule whether you remember it or not — the review is the only part that needs you.
What happens if the AI produces something bad?
You reject it in chat and say why, the same way you would give feedback to a contractor. Because of the review gate, a bad batch costs you a rejection, not a public mistake. Agents also carry your feedback forward, so the batches improve week over week.
Keep going
- AI Images for Ads: Stand Out Without a Designer
- AI Visuals for Social Media: On-Brand Posts Every Day
- AI Video for YouTube: Run a Channel Without a Studio
- AI Images for Email: Newsletters People Actually Open
- Deep dive: AI Marketing Loops: Recursive Self-Improvement
One piece of your marketing that runs itself every week — done by an agent on a schedule, checked by you before anything ships.
Start free — your team is waiting →