In short
Deciding what runs, in what order, with what rules, and who gets told. The constraint now that intelligence is cheap.
When capability was scarce, the hard part was building the thing. Now a capable model is a call away, and the hard part is orchestration: which agent owns which job, what runs on what schedule, what the rules are, how work is handed between agents, and when a human is told.
Orchestration is mostly writing. Instruction files, doctrine, schedules, boundaries, a board where work is tracked. Done well, a company's operations run with a small number of people setting direction and a set of agents doing the repeating.
This is the skill we sell and teach. Not prompting. Knowing what should run, and being able to say it in a way that holds.
Example
What it looks like.
One board, four agents. A card lands, gets a brief and an owner. The owner runs it, leaves proof, marks it done. A sweep every fifteen minutes finds anything stuck. A human sees only what is blocked on them.Related
Read next.
- doctrineThe set of written rules every agent in a company follows. Voice, safety, reporting, git. Written once, applied everywhere.Read the definition
- kanbanA shared board of work in columns: to do, ready, running, blocked, done. Agents pick up cards and leave proof. The board is the report.Read the definition
- agentSoftware that reads its instructions, looks at the situation, takes actions with real tools, and reports back. Runs on a schedule or on a trigger, without being prompted.Read the definition
- workflowA piece of work from trigger to done, run by an agent under instructions. Not a flowchart. The agent handles the branches the chart never drew.Read the definition