Duty Graph · Northstar Parts — Advisor Training 2026-09-05 21:13

Northstar Parts — executive training example

Show the shape of a reviewed advisor-to-client deliverable using synthetic data.

Reviewed for manual delivery · 2026-09-05 · Company revision 70

Audience: Alex Morgan, fictional sponsor; Jamie Park, fictional fulfillment lead

Synthetic training example. This company and its people, claims and measurements are illustrative. This is not an actual client finding.

4people in the engagement roster

4 / 4people with a returned response

3 / 3tasks with current human confirmations

Executive summary

TRAINING EXAMPLE — All people, accounts and measurements are fictional. Three task cards have completed the scripted local participant-confirmation flow. A reviewed manual workflow contains one completed case and one exception. The practice measurement is inconclusive; no automated release or production change occurred.

Decisions for the client

Keep release decisions with people. Review the missing customer reference in TRAIN-002. Continue testing incomplete intake against the competing credit-capacity explanation.

Next steps

Jamie: resolve the exception and gather a matched cohort before the next review. Alex: review scope and the outcome limitations. Advisor: refresh evidence and regenerate the report when source records change.

Scope and limitations

Entirely synthetic training data, not a client case study. Four people are in the engagement roster, not all 24 fictional employees. Local invitation assurance only. No independent authority verification, AI provider, email delivery, signed execution ledger or customer runtime is connected. A task confirmation describes work; it grants no permission.

Scope: Synthetic training: standard order intake through release. Coverage concerns this engagement roster. Raw sources and private response content are excluded.

Selected work and findings

engagement · v1 · reviewed

Northstar order-flow training engagement

Outcome: Understand delays in standard order release and agree a bounded measurement plan.
Systems: Training order register; Training credit checklist
In scope: Four fictional participants; standard orders from intake to release.
Out of scope: Payments, real customer data, production access and automated release.
Review cadence: Weekly practice review
Success criteria: Three reviewed task cards, an explicit handoff contract, a manual case, and a reviewed client packet with limitations.

Record 6f675d4b-19ae-499f-9a33-aec70b146a78 · content hash bf3f07679c587011edcb121e316883e0388460ec17528e7b473aca22e0e1aec7

duty · v1 · reviewed

Prepare standard orders for fulfillment

Scope: Standard orders in the fictional depot.
Purpose: Maintain an accurate, complete release packet.
Linked tasks: Check the incoming order; Record the credit-check result; Release the checked order
Proposed duty owner: Jamie Park

Record 3fb8c7fc-73c0-4305-904c-afb0a2ef8804 · content hash ff37ac93962a37ba571d17304965fcd4afa6b6bf5fea1b5cceb9ec52bc0e7edf

task · v2 · confirmed

Check the incoming order

Duty: Prepare standard orders for fulfillment
Inputs: Customer reference, item and quantity
Output: Complete intake checklist
Purpose: Move complete, checked orders to fulfillment without bypassing credit review.
Trigger: A standard order arrives
Performer: Sam Rivera
Not authorized: Change customer credit limits; Release orders automatically; Use production credentials
Human checkpoint: Jamie reviews exceptions before any human release.
Accountable owner: Jamie Park
Described actions: Read the fictional order; Record a human-reviewed checklist result

Record aa6c9162-e00f-4da5-99e0-4cfba672f672 · content hash 5d34a79907415f407f0795d6437410a2a12fb4ad4b566841777b76c700f1e172

task · v2 · confirmed

Record the credit-check result

Duty: Prepare standard orders for fulfillment
Inputs: Complete intake checklist and customer reference
Output: Recorded pass or exception result
Purpose: Move complete, checked orders to fulfillment without bypassing credit review.
Trigger: The intake checklist is complete
Performer: Robin Ellis
Not authorized: Change customer credit limits; Release orders automatically; Use production credentials
Human checkpoint: Jamie reviews exceptions before any human release.
Accountable owner: Jamie Park
Described actions: Read the fictional order; Record a human-reviewed checklist result

Record 5563fd51-12d2-44d2-9c14-b9cb3f78f6c7 · content hash 29d4fb1e5b9cd788f61fd228eabcd5710475a412fa17875d0f64cccb5421e1d2

task · v2 · confirmed

Release the checked order

Duty: Prepare standard orders for fulfillment
Inputs: Complete order packet and passing result
Output: Order released by a human in the training register
Purpose: Move complete, checked orders to fulfillment without bypassing credit review.
Trigger: A passing credit result is recorded
Performer: Sam Rivera
Not authorized: Change customer credit limits; Release orders automatically; Use production credentials
Human checkpoint: Jamie reviews exceptions before any human release.
Accountable owner: Jamie Park
Described actions: Read the fictional order; Record a human-reviewed checklist result

Record ecf8b304-e60a-4e07-981a-1a3c8183ed31 · content hash c464938d2064d43884c48fe3c410b268013360a81333734bdd91d5b265fa0c20

handoff · v1 · reviewed

Intake to credit review

To: Record the credit-check result
From: Check the incoming order
Output: Complete intake checklist
Condition: Required order fields are complete
Failure action: Stop this case and ask Jamie to resolve the missing information.
Required input: Complete intake checklist and customer reference
Exception owner: Jamie Park
Maximum retries: 1
Receiving check: Receiving person checks order reference and completed checklist.
Escalation hours: 24

Record 24702022-5f08-4615-9891-1bc4153938fa · content hash 082de1a188bdbf8b33ac09949887c334ab512b171212a7cdd326f73c0d3a9f35

handoff · v1 · reviewed

Credit result to release

To: Release the checked order
From: Record the credit-check result
Output: Recorded pass or exception result
Condition: Credit result passes and no exception remains
Failure action: Stop this case and ask Jamie to resolve the missing information.
Required input: Complete order packet and passing result
Exception owner: Jamie Park
Maximum retries: 1
Receiving check: Receiving person checks order reference and completed checklist.
Escalation hours: 24

Record b6a69656-f0cf-44fd-9816-ba585b20f12e · content hash e66cba51938eff812e4d7ba4b3a0cfc93e40d320b7f0787cb90a31b51de37af5

candidate · v1 · constraint hypothesis

Incomplete intake may delay order release

Hypothesis: Incomplete order packets may create waiting and repeated checks.
Throughput unit: Standard orders released per week
Discriminating test: Compare intake waiting with credit review duration in matched cohorts.
Competing explanation: Credit reviewer capacity may instead constrain release.
Global counterfactual: If complete intake removes the longest wait, end-to-end release time should fall while credit quality remains unchanged.

Record a4c0be58-b6e0-48b4-b1dc-87bc77e3479c · content hash 3e0be18babe68160ff0df94849fd2bf0e25f2beba44e2c8253c563261eb2279c

metric · v2 · defined

Median order release duration

Unit: hours
Cohort: 20 standard orders per fictional cohort
Target: 7
Formula: Median of release timestamp minus intake timestamp
Baseline: 10
Question: Does complete intake reduce overall release time?
Guardrail: No increase in credit exceptions or incorrect releases.
Observation count: 1
Missing-data reason

Record 17cf3014-9723-4440-a7c9-d4418dd487c9 · content hash c3f5cb1579f972da5407f717446088b33707827473f28c8596670d2225992d93

intervention · v1 · proposed

Try a complete-intake checklist

Change: Use the checklist on the next matched practice cohort.
Prediction: Median release duration falls from 10 hours to at most 7 hours without more credit exceptions.
Review date: 2026-09-12
Stop conditions: Stop if incorrect releases or credit exceptions increase.
Accountable owner: Jamie Park

Record bacbf17e-af66-4d34-958f-c82791f7831c · content hash a1921c0a80f4e95e94ec8fc9dfbcc803a484ae0b8d9ada7ffd46e8a9dae27ba9

outcome · v1 · reviewed

Practice week outcome: inconclusive

Result: inconclusive
Coverage: 20 invented baseline orders and 20 invented practice orders; this is training data.
Confounders: Small sample, different product mix and unmeasured staffing differences.
Next action: Collect a larger matched cohort and preserve the original prediction.
Interpretation: The invented median improved to 8 hours but missed the 7-hour target. The exercise does not establish causation.
Original prediction: Median release duration falls from 10 hours to at most 7 hours without more credit exceptions.

Record af7973ea-460e-4165-960d-7980e7391cba · content hash 22bec5f93049d693008cfdd120d5a561569e2af5592a47e4f95b19e15d1b282b

review · v1 · open

Practice review: resolve TRAIN-002 and improve measurement

Due: 2026-09-12
Owner: Jamie Park
Decision: Keep all release decisions human-controlled. The current hypothesis remains unproven.
Next action: Resolve the missing customer reference and collect matched observations.

Record 60d3762c-357a-4e01-8a7d-5f3f26a8d896 · content hash 789dbd93b725c52fb7471e6ae30661cc8daa98d764ff926bb9590fc9c4c81491

agent · v1 · draft

Order checklist drafting proposal

Tasks: Check the incoming order
Purpose: Explore drafting a completeness checklist for human review. No permission or deployment is requested by this training artifact.
Deployment: not_deployed
Authorization: None — proposal only
Accountable human: Jamie Park

Record 284ce006-984b-40c9-a36f-7c292d776b60 · content hash 3fc1a91f8e78fae187a7300564379a838d8534b108eb958b186d59b29cd76eef