CoworkersGet started

The Loop · A new operating system for work

Work isn’t a ladder
anymore. It’s a loop.

For a century, companies scaled by adding people to an org chart. The next decade belongs to teams that scale by adding loops — humans and AI digital workers passing work back and forth until it ships. This is the field guide.

Get startedSee the four loops ↓

The org chart is being replaced by the loop.

The old unit of work was a role: a person, a job description, a box on a chart. You grew by adding boxes. But a box is static — it does the same work whether there’s one task or a thousand, and it forgets everything the day its occupant leaves.

The new unit of work is the loop: a human and a digital worker, each doing what they’re best at, passing work between them continuously. Loops don’t sit in a hierarchy. They run — measuring, deciding, learning and shipping — and they get better every time they go around.

Run a company on loops and four of them show up again and again. Here they are.

01

The Work Loop

Request → prioritise → build → ship → repeat.

Work stops being a project with a start and an end and becomes a queue that never empties. You add requests; the highest-priority one becomes active; a digital worker builds it; a human reviews and ships it; the loop immediately pulls the next item. No re-quoting, no re-scoping, no waiting for the next planning cycle — just a continuous line of shipped work.

Humans own
  • Decide what matters most
  • Review and approve
  • Set the standard for “done”
Agents own
  • Build the active item
  • Draft, test, prepare to ship
  • Pick up the next one
HumanAgentWorkLoop
02

The Decision Loop

Autonomy is a dial, not a switch.

The hard question isn’t “can the agent do it?” — it’s “who decides?” Every task sits somewhere on an autonomy dial. Low-stakes, reversible work (tag the ticket, update the record) runs unattended. High-stakes, irreversible work (send the money, email the customer) keeps a human in the loop. The dial moves toward autonomy as trust is earned, task by task — never all at once.

Humans own
  • Set the autonomy dial
  • Own irreversible calls
  • Widen the remit as trust grows
Agents own
  • Act within its boundaries
  • Escalate when unsure
  • Leave an audit trail
HumanAgentDecisionLoop
03

The Knowledge Loop

Context in, a smarter company out.

A digital worker is only as good as what it knows. Your docs, data, past decisions and brand voice flow in as context; every task it completes becomes a new example the next one learns from. Done well, knowledge stops living in one person’s head and starts compounding — the company gets sharper every week instead of forgetting every time someone leaves.

Humans own
  • Curate the source of truth
  • Capture the “why” behind decisions
  • Correct what’s wrong
Agents own
  • Ground every action in your data
  • Cite its sources
  • Surface gaps in the knowledge
HumanAgentKnowledgeLoop
04

The Feedback Loop

Trust is earned in public, one correction at a time.

Nothing runs unattended on day one. The worker proposes, a human checks, and every correction tightens the next result. Accuracy is measured, not assumed; mistakes are caught early and cheaply while a person is still reviewing. Over weeks, the review burden shrinks from “check everything” to “check the edge cases” — and that shrinking is the proof the loop is working.

Humans own
  • Review and correct
  • Measure accuracy honestly
  • Decide when to let go
Agents own
  • Show its reasoning
  • Improve from feedback
  • Flag its own low-confidence calls
HumanAgentFeedbackLoop

Where the humans go

Agents take the doing. People take the deciding.

The fear is that digital workers replace people. What actually happens is that people move up the value chain — from executing tasks to directing them.

The Orchestrator

Designs the loops, sets priorities, and decides what work belongs on a queue in the first place.

The Decider

Owns the irreversible calls — the spend, the hire, the customer promise — that agents are deliberately not allowed to make alone.

The Reviewer

Holds the standard for “done”, catches the edge cases, and turns every correction into better future output.

The Relationship-holder

Does the irreducibly human work: trust, judgement, taste, and the conversations that no agent should have.

The principles

Six rules for a company that runs on loops

01

Outcomes, not seats

You don’t add headcount to grow capacity. You add loops. The question shifts from “who do we hire?” to “what work should run on a loop?”

02

Humans move up, not out

Agents take the doing; people take the deciding, reviewing and relationship-holding. The most human work becomes the most valuable work.

03

Trust is earned, never assumed

Every loop starts supervised. Autonomy is granted task by task, backed by logs and approvals — the same way you’d trust a new hire.

04

Knowledge compounds

Context flows in, learning flows out. A company built on loops gets smarter every week instead of resetting with every departure.

05

Reversible by default

Anything an agent does unattended should be cheap to undo. Irreversible actions keep a human in the loop, on purpose.

06

One queue, always moving

The backlog is the heartbeat. Work is continuously prioritised, built and shipped — never frozen waiting for the next quarter.

Go deeper

The Loop, one chapter at a time

Each guide takes one piece of this and makes it concrete — what the words mean, how it works, and how to put it to work in your business.

Stop scaling by headcount. Start scaling by loops.

Coworkers builds the digital workers and the software that runs your loops — unlimited development, one feature at a time, as a flat monthly subscription. No lock-in, pause anytime, Australian timezone.

Get started today