The Multi-Agent Organization

Every chapter of The Multi-Agent Organization with its printed epigraph, what you can do afterwards, and the moment it is built for.

1. Introduction

   The Coordination Problem

   You added agent workers faster than you added the protocol to coordinate them, and the bill is now due.

   What you can do afterwards

   You will name the coordination gap in your own team and see why protocol, not headcount, is your next lever.

   Use this chapter when

   Your agent workers piled up quietly and now the output contradicts itself.

2. Chapter 1

   Org Charts Are Obsolete

   The chart tells you who answers for the work; it never tells you who moves it.

   What you can do afterwards

   You will stop expecting the org chart to fix coordination, and start reading your work as a flow of handoffs instead of a tree of reporting lines.

   Use this chapter when

   A workflow keeps stalling between people or agent workers, and no one can name the outcome's owner even though everyone can name their own box on the chart.

3. Chapter 2

   The Multi-Agent Team

   A team is a set of named roles, each with one job, one output, and one person who answers for it, not a pile of capabilities.

   What you can do afterwards

   You will stop thinking of agents as one fuzzy "AI helper" and start treating them as specialized workers alongside the humans they support. Each worker gets a name, an output, and a known failure mode.

   Use this chapter when

   You have agent runs scattered across chat threads and shared docs, nobody can say exactly which jobs the agents own, and you cannot tell where a human must stay in the loop.

4. Chapter 3

   Charter

   A worker without a charter is not a worker; it is a guess you keep paying for.

   What you can do afterwards

   You will write one complete charter for a real role on your team, human or agent, that another person could hand work to without a meeting.

   Use this chapter when

   You are about to put an agent worker into production, onboard a human operator, or you keep getting output that is almost right but never shippable.

5. Chapter 4

   Handoff

   The work does not break at the desks; it breaks in the space between them.

   What you can do afterwards

   You will build a context packet template and run one real handoff with it, so the next worker (human or agent) never has to guess what you meant.

   Use this chapter when

   Work crosses a boundary: human to agent, agent to human, or one person to the next on a chartered task.

6. Chapter 5

   Orchestrate

   Orchestration is deciding, on purpose, what runs in sequence, what runs in parallel, what merges, and what stops for a human, not getting everyone to work at once.

   What you can do afterwards

   You can take one multi-step workflow you run today and draw it as an orchestration graph. Serial links go where order matters, fan-out where work can split, fan-in where it must merge, and escalation gates where a human steps in.

   Use this chapter when

   A workflow involves more than two handoffs, more than one agent worker, or any step where parallel work would save time but risks producing conflicting outputs.

7. Chapter 6

   Review

   A charter that promises quality without a gate that verifies it is a wish, not a system.

   What you can do afterwards

   You will build one review gate that checks agent output against written acceptance criteria, so quality stops depending on whoever happens to be paying attention.

   Use this chapter when

   An agent worker produces something that leaves your team (a client deliverable, a customer reply, a code change, a published post) and a mistake costs money, trust, or a rework loop.

8. Chapter 7

   Update

   A review that changes nothing is a review you ran for the file, not for the work.

   What you can do afterwards

   You will turn a single failed review into a permanent improvement by changing one upstream artifact and logging the change.

   Use this chapter when

   A review gate catches a defect, and you find yourself fixing the same defect more than once.

9. Chapter 8

   Sync

   A team stays aligned by reading the same surface, not by sitting in the same meeting.

   What you can do afterwards

   You will replace status meetings with a sync surface anyone can read in thirty seconds and a ritual that takes fifteen minutes.

   Use this chapter when

   Your team (human operators and agent workers) runs enough parallel work that nobody can hold the full state in their head, and you keep discovering surprises late.

10. Chapter 9

    The Small-Team Operating Model

    A coordination protocol you do not run on a rhythm is just a document; CHORUS becomes an operating model only when charter, handoff, review, and sync each land on a predictable beat.

    What you can do afterwards

    You will compose the six CHORUS skills into one lightweight weekly rhythm and capture it on a single page your whole team can read in two minutes.

    Use this chapter when

    You have learned the skills one at a time and now run them ad hoc. Charters drift, reviews get skipped under deadline, and nobody is sure what the agent workers shipped last week.

11. Chapter 10

    Failure Modes

    Multi-agent teams rarely fail from a missing tool; they fail from a missing protocol, and the failures arrive in a small, predictable set.

    What you can do afterwards

    You will recognise the five failure modes that quietly degrade multi-agent teams and match each to the CHORUS skill that fixes it.

    Use this chapter when

    Output quality is slipping, work is being redone, or you sense the team is busy without being aligned, and you cannot name why.

12. Chapter 11

    Build Your CHORUS Stack

    The protocol is the asset; the tools are rented, and you should treat them that way.

    What you can do afterwards

    You will leave with a tool mapped to each CHORUS skill, a traceability check that proves your agent runs are auditable, and a clear rule for when a spreadsheet beats a platform.

    Use this chapter when

    You are about to buy coordination software, or you already bought it and the coordination problem stayed.

13. Conclusion

    Conclusion: The Protocol Organization

    Your competitors can copy your tools by Friday; the thing they cannot copy is how cleanly your team hands work between humans and agent workers.

    What you can do afterwards

    You will leave with CHORUS as a shared vocabulary your team actually uses, and a ninety-day plan to install it without a reorg, a new platform, or a single extra meeting.

    Use this chapter when

    You have read the skills, you believe the protocol, and you are standing in the gap between knowing CHORUS and running it on Monday.