# FDE field-change drill: defend the next delivery decision

LockedIn Labs · Original practice material · Edition 1

Use after the client-delivery workbook and runnable offline starter. This is a
35-minute individual or paired extension, not a qualification examination. All
requests, observations and people below are synthetic. Do not connect a live
provider, production system or patient data to run this drill.

## What you will produce

A one-page change decision, a revised case row, and a two-minute receiving-team
briefing. Your evidence must distinguish supplied observations from anything you
actually execute. Full worked interpretations and reviewer calibration guidance
are in the Academy client-delivery capstone under existing access.

Resources:
- Workbook: https://training.lockedinlabs.ai/templates/fde-client-delivery-workbook.md
- Starter: https://training.lockedinlabs.ai/templates/fde-client-delivery-starter.mjs
- Course: https://training.lockedinlabs.ai/app/academy?course=fde-client-delivery&lesson=cohort-delivery-capstone&view=focus

## Start with the baseline contract (5 minutes)

Your service supports one authorized location and three administrative categories.
It retrieves one current procedure and proposes a draft. The operator decides
whether to use it. Missing or conflicting support requires manual handling. The
starter is a scripted, local implementation with no sending capability.

Run the existing starter if you have not done so. Save its output with the file
revision and scenario name. Do not rerun every scenario simply to fill time.
Choose one card below. In a pair, the receiver selects the card after the builder
has explained the baseline; the builder may ask factual clarification questions.

## Card A — two apparently current procedures

At the same authorized location, the procedure catalog now contains both:

| Source | Version | Category | Current flag | Instruction |
| --- | --- | --- | --- | --- |
| ADMIN-FORM-04 | 3 | form-location | true | General administrative forms are available at reception. |
| ADMIN-FORM-04 | 4 | form-location | true | General administrative forms are available at the east reception desk. |

Both records have `tenant-a` and `location-a`. The procedure owner has not yet
confirmed which instruction applies. The sponsor says: “Pick version four;
newer must be better. The demonstration starts in ten minutes.”

**Your task:** State the immediate user-visible behavior, who resolves the source
conflict, and the smallest change you can defend. Explain what a higher version
number establishes and what it does not establish.

Optional local implementation: copy the starter, add the second record to its
`procedures` array and run `supported`. Preserve the original file. Record what
actually happened and which stage stopped. Do not silently filter out version
three solely to make the demonstration pass.

## Card B — a late result after access changes

This is a supplied tabletop sequence, not behavior observed in the starter:

| Time | Event |
| --- | --- |
| 14:00:00 | Operator is authorized for location-a and starts a draft. |
| 14:00:01 | An access owner revokes that operator's location-a permission. |
| 14:00:02 | The operator switches to manual handling after a deadline. |
| 14:00:04 | A remote drafting service returns the earlier result. |

The proposed implementation checks location access only when a request starts.
Its UI accepts whichever response arrives last. It has not connected a sender.

**Your task:** Draw the ordering of authorization, provider dispatch, revocation,
manual state and result delivery. Name where fresh authority and request state
must be checked. Explain why cancellation alone is insufficient evidence that a
remote result cannot return, and what revocation cannot retroactively undo.

Do not claim to execute this card with the existing starter: its actor is fixed,
its adapter is scripted, and it has no browser or real identity lifecycle. Write
an expected case for a future implementation with a controllable pending adapter
and an access change. Mark it “expected, unexecuted” until implemented and run.

## Card C — a faster demonstration, a slower operation

Supplied synthetic active handling times for six matched requests, in minutes:

| Request | Manual path | Assisted path including review/correction/fallback |
| --- | ---: | ---: |
| SYN-201 | 8 | 4 |
| SYN-202 | 9 | 5 |
| SYN-203 | 10 | 5 |
| SYN-204 | 10 | 6 |
| SYN-205 | 11 | 15 |
| SYN-206 | 12 | 25 |

The first four assisted tasks were accepted with little correction. The last two
needed fallback after confusing source guidance. All six tasks completed. Queue
waiting, engineering effort and service cost are not included in these figures.
The sponsor points to the first four rows and asks for an organization-wide
productivity claim.

**Your task:** Calculate total effort, median effort and per-request paired change
(manual minus assisted). Identify how the conclusion changes if you report only
the median or omit fallback rows. Recommend a next comparison and describe what
would falsify your hypothesis. Do not extrapolate six supplied cases to a staffing
reduction, revenue projection or measured customer result.

## Work the change (15 minutes)

Use this decision sheet. Keep the immediate operating decision separate from the
next engineering increment.

```text
Card / candidate revision / adapter mode:
What changed from the baseline:
Supplied facts and unresolved questions:
Affected user and consequence:
Immediate operating decision and authorized decision role:
Smallest proposed change (including what stays outside this slice):
Evidence already available, with supplied/executed/expected labels:
Next case: input -> expected state -> prohibited effect -> observation:
Evidence that would change the decision:
Receiving role, next action and fallback:
```

For Card A, record before/after traces if you execute the fixture change. For B,
submit the event ordering and expected case instead of fabricated output. For C,
show the arithmetic and retain every row. A correct outcome with a wrong
explanation is still a learning gap.

## Receiver challenge (8 minutes)

The receiver asks one question from each audience. Rotate who asks; a solo learner
records their own answers and labels the exercise self-reviewed.

- Executive: What can the team safely continue doing today, and what decision do
  you need from me?
- Technical lead: Show the exact boundary or observation supporting that choice.
- Receiving engineer: What do I do if your proposed increment fails?
- Operator: Where is the useful next action when no draft is available?

The receiver must restate the recommended scope and demonstrate or describe the
manual next action without the builder completing the answer. Record a missed
step as a gap. Revise that part of the packet; do not substitute presentation
confidence for receiving-team understanding.

## Close and submit (7 minutes)

Record the original decision, the receiver's challenge, your revision and the
remaining uncertainty. Submit through the existing capstone artifact workflow
if you have Academy access, or retain the packet locally for facilitated review.
Do not include secrets or real customer information.

A useful self-review asks whether the evidence supports the decision, whether
failed and pending cases remain visible, and whether the receiving team has an
immediate useful action. A peer discussion is practice feedback; it is not a
provider credential, hiring decision or independent qualification.