Use Cases
Workflow Automation
Recipes, agents, and daily rhythms that run themselves.
The boring-but-vital ops — research, briefings, content, deploys, reporting — wired into agent recipes that run on a cadence. Same outputs, 24 hours a day, no human in the typing seat.
- Recipe-based agent orchestration
- Daily / hourly / on-trigger cadences
- Human-in-the-loop at the steps that matter
Why
Every team has 20 hours a week of repeatable work that nobody owns. Pinning each one to an agent recipe converts that loss into a daily output the business can build a plan around.
How
- Map the 20 most-repeated tasks; rank by frequency × pain
- Wrap each in a recipe: trigger → context → action → log
- Roll out the cheapest ones first; let savings fund the rest
Proof
- Recipes in production
- 30+
- Daily automated outputs
- PAD · LAD · IAD · AAD · VAD · BAD
- Always-on agents
- 13
Workflow automation — the universal pattern
Trigger · Context · Action · Log
Hover or tap a node to see details.
FAQ
- What kinds of triggers does this support?
- Schedule (cron), webhook (HTTP POST), queue message (any broker), and filesystem changes. New trigger types are pluggable — write the adapter, register it, done.
- Can I add manual approval steps?
- Yes — the action step accepts a human tool. The agent pauses, the operator UI surfaces the request, the run resumes on approval or stops on rejection. Audit-traced both ways.
- What if the agent makes a wrong call?
- The audit log captures the full input, decision, and output. Replay with a corrected prompt or constraint, or roll back the downstream effect if the action wrote to a system.
In production
- Daily content cron
Six daily workflows running on AIOS with one operator — PAD, LAD, IAD, AAD, VAD, BAD.
See it - xmas + worldcupai automation
Per-site automation jobs (aa, images, lad, ai-summary, social) running on their own crons.
- Audit-traced replay
Every run capturable, replayable, comparable. Tune by inspecting the failures, not by guessing.