Macro 3D Playbook court with an AI agent seated in a calibration rig.
The Delegation Practice

Try an AI workflow before connecting your tools.

Use a practice workflow to decide what AI can do, when it needs approval, and how to check its work.

Map one workflow Examine the evidence
  • Owner Accountable operator
  • Authority Observe one workflow
  • Evidence Draft
Practice steps

Find the problem, then practice the fix.

Start by checking the information and rules. Then walk through a sample task and review the result.

01 / 04 Database → automation → judgment

Diagnose the system before changing the policy.

Check three things in order: the information available, whether the task ran correctly, and whether it followed the right rules.

Delegated Work Control

Check the information, the task, and the rules.

This order helps you find the cause before changing the system.

01 / Database Is the information correct?

Check the records, their dates, and the rules the task is allowed to use.

02 / Automation Did the task run correctly?

Check the tools, approvals, retries, and results. A completed run does not always mean the result is correct.

03 / Judgment Did it follow the right rules?

Identify who decides, which rule applies, and when the agent should ask for help or stop.

02 / 04 Run, wait, or stop

See where work may run, wait, or stop.

This example shows a Marketplace review queue. It is for learning: it cannot access live systems and is not evidence of client results.

Map story canvas

Map story for Submitted asset packet

The same Map graph can teach the workflow to a person and preserve the typed contract an agent needs to act safely.

1. Map Submitted asset packet before execution. Marketplace review owner owns the operating path. The canvas makes the workflow, handoffs, and next decision legible before an agent or system acts. 2. Validator and queue sync can run when the rule is clear. Validator and queue sync coordinates with Supplemental reviewer brief; the map keeps AI assistance bounded to the work it can safely support. 3. Reviewer approval decision stays with a person. Human decides approve, reject, request changes, or escalate policy ambiguity. 4. No ungrounded approval is the stop condition. Stop before approval, rejection, security claims, or timeline promises without evidence. 5. Reviewer dashboard receipt shows the receipt. Shows validation evidence, reviewer state, creator-facing notes, and policy flags. 6. Use the map as booking context for a workflow pilot. The map has enough owner, assistive work, system behavior, and decision context for a first run.

Workflow map Editable workflow
Run Wait Stop
Press enter or space to select a node. You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.
Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.
Receipts
named ownergoverned handoffstop conditionresult record
03 / 04 Practice before authority

Work through ten steps and review each result.

Use the workbench to plan a task, set permissions, and review tests and results before considering live access.

Operator journey 01 / 10

Ten stages. Ten inspectable artifacts.

Progress follows the work produced, not time on page or content consumed.

Stage 01 Delegation Card

Enter

What work am I responsible for?

Name one workflow, owner, stakes, and current pain.

Missing artifact: Operator name or role, Workflow, Accountable owner, Stakes

Internal preview Not certification

Practice Receipt

Generate only after every artifact and governance binding is inspectable.

04 / 04 State and limits stay attached

Claims stay attached to state and limits.

Verified, review, and draft records stay deliberately different. Governed Agent Delivery remains a proposed flagship defense until its artifacts, counterexamples, rollback, and independent verdict are examined.

Source evidence04 records

Verified, review, and draft are different states.

Read the source record and its limits before relying on a result.

  1. 01 Operating claimSignal → Decision → ProofThe category method stays attached to workflow ownership, policy boundaries, and inspectable evidence. verifiedTHESIS-01Current public method
  2. 02 Workflow modelMarketplace review queueA read-only Map story shows the signal, automated preparation, human review, stop condition, and result record. verifiedMAP-01Representative local map
  3. 03 Prototype proofMarketplace workflow compilerThe proof route keeps representative fixtures, active-development status, and the absence of production writes visible. reviewCASE-01Public limits intact
  4. 04 Flagship defenseGoverned Agent DeliveryThe repo-owned case follows a Linear signal through bounded implementation to production proof. It remains a proposed defense until the full evidence packet is examined. draftDEFENSE-01Not yet publicly defended
Flagship defense · unresolved boundary

Governed Agent Delivery: from Linear signal to production proof.

This case follows work from a Linear issue to production. At every step, it must keep the owner, allowed actions, required checks, approval points, and rollback evidence visible.

A proposed case is not a passed defense.

Passing still requires artifact inspection, counterexample review, a rollback check, and an independent verdict with real consequences.

Read the test results
Review your practice

Leave with a plan and a record of your checks.

The practice result includes a workflow plan, tested permissions, and review notes. The review may call for changes or show that the approach does not work. When the task and person responsible are clear, bring the plan to a mapping session.

Request a mapping session Read the test results For service providers Use the Practice with clients
Owner
Accountable practitioner
Authority
One named workflow only
Proof
Artifact bundle + review verdict
State
review