1
Pick the workflow
The same one you scored on the canvas. Never the company, never the department.
Fit
One cell, blown up.
The canvas asks twelve questions about one workflow. This sheet takes the fourth, Composition, and splits it in two: what the work is made of, and what each carrier can actually hold. Fit is where the halves match. Where they do not, the misfit gets a name and a move, not a score.
Zooms intoCell 04, Composition
OutputA named misfit list, never a score
BasisThe Scaffold judgment taxonomy on O*NET 30.3 Primary source
RulesEight, this site's own Advisory
Rests onPrinciples two and three
I / The sheet
Mark the work, then mark the carriers
Sixteen tiles. The orange column is the work itself, by component. The three columns beside it are the human, the agent, and the system, and each asks the same question: how much of this component can you carry today. Misfits appear under the sheet as you mark.
Composition Fit Canvas · v1.0One cell, blown up · four components · three carriers
Workflow
Scored by
Date
Version
Component
The work
How much of the workflow is this kind of work.
The human
A named person, today, with the time they actually have.
The agent
The loop as it runs now, not as the roadmap describes it.
The system
The platform and the workflow rules: routing, records, guardrails. Not an actor.
Pattern
Work that repeats and can be recognized: the same input shape, the same acceptable output, again.
PThe work
How much of this workflow is Pattern work?
PThe human
How much of the Pattern work can the human carry today?
PThe agent
How much of the Pattern work can the agent carry today?
PThe system
How much of the Pattern work can the system carry today?
Judgment
Decisions under uncertainty that a named person has to answer for afterward.
JThe work
How much of this workflow is Judgment work?
JThe human
How much of the Judgment work can the human carry today?
JThe agent
How much of the Judgment work can the agent carry today?
JThe system
How much of the Judgment work can the system carry today?
Relationship
Work whose value sits in the other party's trust, and moves when the face changes.
RThe work
How much of this workflow is Relationship work?
RThe human
How much of the Relationship work can the human carry today?
RThe agent
How much of the Relationship work can the agent carry today?
RThe system
How much of the Relationship work can the system carry today?
Creative
Work that produces something that did not exist before, where the brief itself is part of the output.
CThe work
How much of this workflow is Creative work?
CThe human
How much of the Creative work can the human carry today?
CThe agent
How much of the Creative work can the agent carry today?
CThe system
How much of the Creative work can the system carry today?
0of 16 marked0misfits
A misfit is a finding with a name, not a score. Mark what is true today, not the roadmap.workcomposition.org · v1.0 · CC BY 4.0
Mark the work column first; misfits appear here.
II / Five moves
Twenty minutes, inside the ninety
The sheet runs inside the second move of the protocol, where the room is already deconstructing the work. It does not need its own meeting.
2
Mark the work
Four components, one tick each: how much of the workflow is Pattern, Judgment, Relationship, Creative.
3
Mark the carriers
Twelve tiles: how much of each component the human, the agent, and the system can carry today. Today, not the roadmap.
4
Read the misfits
Each one has a name and a move. The list is the output; there is no score to argue about instead.
5
Take it back to the canvas
A workflow with a live misfit cannot score Composition as solid. Copy the assessment and attach it to cell 04.
The sheet is a zoom, not a second method. Everything it finds is filed against one cell of the canvas.The five moves of the protocol
III / The misfits
Eight named misfits, each with a move
The two headline misfits are Judgment routed to the agent and Pattern work left with the human. The other six were named during the build. Every one links to the shape, anti-pattern, gate item, or record that resolves it. The rules are written as testable claims so that a room can disagree with a specific one rather than with the sheet.
Judgment routed to the agent
Fires when Most of the work is Judgment, the agent carries some of it, and no human is marked as fully able to carry it.
Judgment is the component that carries accountability. When it sits with the agent and nobody is marked fully able to carry it, the loop has no qualified halter, and the human oversight duty is being met on paper only.
Pattern work left with the human
Fires when Most of the work is Pattern, the human is marked Fully, and neither the agent nor the system is.
Pattern is the one component every executor can carry. When only the human is marked Fully, the most delegable work is sitting with the most expensive carrier, usually because nobody has tested the agent on it.
Relationship carried by the system
Fires when There is Relationship work and the system is marked as fully carrying it.
A system can route a relationship. It cannot hold one. The counterparty experiences a queue, and the trust the work was producing goes to whichever competitor still answers.
Relationship work with no face
Fires when Most of the work is Relationship and no human is marked as able to carry any of it.
The workflow's value is in the other party's trust and nobody on the human side is available to hold it. This is the misfit that shows up as churn six months later, after the efficiency figures were reported.
Creative work with no owner
Fires when There is Creative work and no human is marked as able to carry any of it.
Creative output is a choice somebody has to stand behind. If no person can carry the component, the choice is being made by the model's defaults and the Authority cell has a hole in it.
Judgment with no reviewer
Fires when There is Judgment work, the agent carries some of it, and the human is marked Cannot.
The room has said out loud that nobody available can review the decision. That fails the fifth item of the gate before the agent ships, and it fails it in writing.
Capacity with no work
Fires when A component is marked None in the work column and some carrier is marked Fully on it.
Somebody has bought or built capacity for work this workflow does not contain. It is the demo-attached cell in miniature: the capability was funded because it could be shown, not because the work called for it.
Work nobody can carry
Fires when A component is present in the work and every carrier is marked Cannot.
The work exists and the room cannot name who does it. In practice it is being done by whoever is nearest, unrecorded, which is how a delayering turns coordination into delay.
IV / Method
What the sheet claims, and what it does not
The four components come from the Scaffold judgment taxonomy, built on the O*NET occupational database at version 30.3. That taxonomy has been independently reviewed and found defensible but novel, with one cluster flagged as its weakest; the review is cited on the library entry rather than hidden here.
The eight misfit rules are this site's own position. They are not drawn from the taxonomy and they have not been validated against outcomes. Each is stated as a condition on the sixteen marks so that it can be tested, argued with, and replaced by a better one.
The sheet does not score the workflow and does not rank the carriers. A workflow with no misfits is not thereby well composed; it has only avoided the eight mistakes named here. Task deconstruction is still the move that decides what the work is made of, and this sheet only records the result.
Argue with a ruleRead next