iWMS O2O AI Agent Workshop Guide v3

Purpose

Workshop Goal

Goal: business users learn how to ask the Agent the right operational questions, verify the outputs, understand the logic behind each finding, and continue running structured test cases before the next follow-up session.

Pre-workshop

During workshop

Participants follow step-by-step instructions. Their job is to verify results and discuss business logic, not discover the process from scratch.

Post-workshop

Participants rerun the same scenarios for additional dates, stores, and issue types, then bring feedback to the next session.

Schedule

July 15 Agenda: 13:00-17:00

13:00-13:30Setup buffer: Agent login, store scope, LINE readiness

Everyone can follow the same session without technical interruption.

13:30-13:45Agent overview and Skill map

Understand what the Agent can answer today and what is Phase 2.

13:45-14:20Module A: OTIFS diagnosis and outbound briefing

Read North Star status and split On-time vs In-full.

14:20-14:50Module C: POD and order closure monitor

Find ready-for-pick, ready-for-send, overdue and near-cutoff risks.

14:50-15:20Module D: Shortage, replenishment, picking, manpower

Know which Skill to use when In-full or picking capability is the issue.

15:20-15:30Break

Collect quick questions.

15:30-16:00Module E: LINE push, template, snapshot, scheduler

Use the pre-bound LINE account to review test cards and scheduled reports.

16:00-16:25Module F: Knowledge Q&A, troubleshooting, operations guide

Ask process/config/metric/operation questions and verify KB answers.

16:25-16:55Module G: Build a Cloud App dashboard

Let the Agent build, publish, and open a shareable data dashboard.

16:55-17:00Post-work cases and feedback plan

Leave with repeatable practice cases and a feedback template.

Before 13:00

Pre-workshop Setup

00Readiness Checklist

Complete this before business users start the guided session.

30 MIN BUFFER

1Confirm Agent login and store authorization

Each participant logs in using their normal account.

Expected outcome: the Agent can show the current authorized warehouse/store context.

Verify: store name and scope match the participant's role.

2Confirm system connection

Participants ask the Agent a business-friendly readiness question. The Agent performs the technical checks in the background.

Am I successfully connected to the system now?

Expected outcome: the Agent confirms that the system connection is ready and shows the current authorized store or warehouse scope.

Success example: You are successfully connected. Your account can access the current store data, and the data service is ready for the Workshop.
If not ready: Agent should explain the business impact, such as login expired or data service temporarily unavailable, without exposing technical command details.

3Bind LINE account for notifications

Participants bind LINE before the hands-on push workflow. If binding cannot be completed for everyone, the facilitator completes one demo account and the rest observe.

Please help me bind my LINE account for Agent notifications.

Expected outcome: the Agent uses line-bindcode to generate a bind code or QR page, and the participant's LINE account is ready to receive test notifications.

Verify: participant can confirm the bound account or understands that the hands-on LINE push will be demonstrated by the facilitator.

4Generate feedback sheet with the Agent

Participants ask the Agent to create a feedback sheet template. During and after testing, they can ask the Agent to append findings into the same structure.

Create a Workshop feedback sheet template for me. Include these columns: scenario, prompt used, expected result, actual result, business comment, data gap category, priority, and follow-up owner.

Expected outcome: the Agent creates a reusable feedback sheet template for post-workshop test execution.

Follow-up prompt: Add this test result to my feedback sheet: [describe the scenario, actual result, and business comment].

Capabilities

Agent Skill Map

The Agent is not a single report. It is a routed skill system: business questions are mapped to the right data, knowledge, push, or builder skill.

North Star and outbound

makro-otifs-trackermakro-outbound-dashboardmakro-store-outbound-rankingwms-order-querywms-trip-querywms-pack-query

POD and order closure

makro-pod-monitororder-info-querypick-info-querypack-info-query

Shortage and replenishment

makro-shortage-monitormakro-replenish-trackermakro-duplicate-cn-trackerwms-shortage-query

Picking and workforce

makro-first-pick-ratemakro-picking-efficiencymakro-shift-pick-ratemakro-employee-performancemakro-manpower-shiftwms-worker-efficiency-query

Knowledge and troubleshooting

process-explainconfig-explainmetric-explainoperation-guidetroubleshootwms-knowledge-fmcli

LINE, report, and automation

line-bindcodeline-template-buildersnapshot-builderscheduler-manager

Builder and extension

fmcli-runfmcli-adhoc-queryfmcli-create-metriccloud-app-builderapp-builder

Sales and operational add-ons

makro-sales-performancecloud-app-buildersnapshot-builder

Migration Map

From Excel Macro to Agent Skill

Original workbook	Business purpose	Agent path	Status
Top 50 store focus	Store performance, OTIFS split, shortage and follow-up focus.	makro-store-outbound-ranking	MVP ready Basic store outbound ranking. Area/OD mapping is a later enhancement.
Check Order / POD	Late-stage order follow-up, ready-for-pick, ready-for-send, overdue risk, POD/payment closure.	makro-pod-monitor	MVP partial WMS-side order closure ready. POD/payment and Locus are Phase 2.

Module A

OTIFS Diagnosis and Outbound Briefing

AStart From the North Star

Use OTIFS as the entry point, then drill down into On-time and In-full.

35 MIN

1Ask for today's overall status

How is today's outbound performance? Start with OTIFS, then show On-time, In-full, and the main risk.

Expected outcome: Agent opens with OTIFS and splits the issue into On-time and In-full.

2Ask for the likely cause

OTIFS is low. Please explain whether the issue is mainly On-time or In-full, and tell me what to check next.

Expected outcome: Agent routes to shortage/picking/trip based on the weaker side.

3Verify the logic

Participants check whether the recommendation matches their operational experience for the store.

Discuss: Is the issue actually timing, shortage, picking capacity, or another process bottleneck?

Module C

POD and Order Closure Monitor

CFind Orders Needing Follow-up

WMS-side MVP behind the Check Order / POD macro workbook.

30 MIN

1Ask for closure risk

Please check today's POD and order closure risk. Show ready-for-pick, ready-for-send, and overdue or near-cutoff orders.

Expected outcome: Agent separates Ready for pick, Ready for send, overdue/near-cutoff, and POD/payment availability.

2Confirm Phase 2 gaps

Important: POD/payment pending count, POD amount, POD aging, Locus actual delivery, and customer/sales owner mapping are Phase 2.

Expected outcome: participants know that "not connected yet" is not the same as "no issue".

3Open the output file

pod_monitor_daily_YYYYMMDD.xlsxpod_monitor_summary_YYYYMMDD.json

Expected outcome: participant can review the highest-risk bucket and sample orders.

Module D

Shortage, Replenishment, Picking, and Workforce

DFollow the Root Cause Chain

Practice which Skill to use after OTIFS is split.

30 MIN

Question	Prompt	Expected Skill
Shortage issue	Which departments or items are driving shortage today?	makro-shortage-monitor
Replenishment follow-up	Which replenishment tasks still need manager attention?	makro-replenish-tracker
First pick issue	How is first-pick success today? Which area is weak?	makro-first-pick-rate
Picking speed	Is picking efficiency below target? What should I check next?	makro-picking-efficiency
Shift capacity	How many orders are picked per picker this shift?	makro-shift-pick-rate
Manpower	Do we have enough manpower by shift and task type?	makro-manpower-shift

Expected outcome: participants can choose the right Skill based on the business symptom instead of asking one broad question.

Module E

Use LINE Push, Message Template, Snapshot, and Scheduler

ETurn Results Into Push Workflow

Use an already-bound LINE account to preview, test, and schedule manager-review notifications.

30 MIN

1Preview a LINE template

Show me a LINE card template for OTIFS alert with OTIFS, On-time, In-full, and top action options.

Expected outcome: Agent uses line-template-builder and shows a manager-review card, not raw JSON.

2Send a test notification to the bound LINE account

Send a test LINE notification to my bound account using the OTIFS alert template. Mark it as test only.

Expected outcome: participant receives a test LINE message and verifies the content, severity, metrics, and action options.

3Create detail page snapshot

Create a detail page for this store outbound ranking result, with KPI summary and table.

Expected outcome: Agent uses snapshot-builder to create a shareable detail page concept.

4Discuss scheduled push

Set up a daily 07:30 outbound briefing push for manager review.

Expected outcome: Agent routes to scheduler-manager. Participants understand schedule, template, recipient, and manual review requirements.

Module F

Knowledge Q&A and Troubleshooting

FAsk Process and Operation Questions

Use the knowledge skills when the question is "what does this mean" rather than "show today's number".

25 MIN

Question type	Sample prompt	Skill
Process	Explain the outbound order flow from order creation to handover.	process-explain
Metric	What does OTIFS mean and how should I interpret it?	metric-explain
Operation guide	How should a picker handle an OOS item step by step?	operation-guide
Configuration	Why is this feature not visible for my store?	config-explain
Troubleshooting	This order is stuck before handover. What should I check?	troubleshoot

Verify: knowledge answers must come from the knowledge module. If the knowledge base has no answer, the Agent should say so and not invent one.

Module G

Build a Cloud App Dashboard

GTurn Insights Into a Shareable Dashboard

Use the cloud-app-builder skill to let the Agent build, publish, and open a data dashboard — no coding from the business user.

30 MIN

1Describe the dashboard you want

Tell the Agent the topic, data source, and layout. The Agent confirms the scope and proposes the modules it will call.

Help me build a cloud app dashboard for outbound overview. Show OTIFS, on-time, and in-full as cards, add a date picker, and use yesterday's data by default.

Expected outcome: the Agent uses cloud-app-builder and confirms the dashboard topic, data modules, and a slug name before generating code.

2Let the Agent prepare the backend modules

The Agent checks whether the needed Groovy cloud modules exist and creates any missing ones. Each module must be published to Center before the dashboard can call it.

Verify: the Agent lists the modules it will use (for example POST /api/pick/counts) and confirms they are published. If a module is missing, the Agent creates and tests it first.

3Create and publish the app

Ask the Agent to create the app and publish the frontend files. Each publish creates a new version (v1, v2, ...) that you can roll back.

Create and publish the app with slug outbound-overview. Name it "Outbound Overview".

Expected outcome: the Agent runs fmcli app create and fmcli app publish, then confirms the app slug and version.

4Open and verify in the browser

Ask the Agent to open the app. It auto-generates an access token and opens the dashboard in your browser.

Open the outbound-overview app so I can verify it loads correctly.

Expected outcome: the dashboard opens without a 401 error, cards render real data, and switching the date reloads the numbers.

Verify: each data region loads independently — if one module fails, the others still show. The Agent should not expose SQL or internal API details in the page.

5Share the dashboard (optional)

Share the app with a colleague by login ID. Each viewer gets their own session and sees data scoped to their authorized store.

Share the outbound-overview app with [colleague login ID].

Expected outcome: the Agent runs fmcli app share and the colleague can open the app with their own store scope.

6List and manage your existing apps

Ask the Agent to list every cloud app you already have. Useful before creating a new one to avoid duplicates, or later to review, roll back, or clean up apps.

What apps do I have now? List the names and versions.

Expected outcome: the Agent runs fmcli app list and returns every app you own with its slug, display name, and current version.

Verify: the list matches what you actually built. Use fmcli app versions <slug> for history, fmcli app rollback <slug> --version N to roll back, or fmcli app delete <slug> to remove one.

Ready-to-use

Prompt Library by Scenario

Morning briefing

Please give me today's outbound briefing. Start with OTIFS, then split On-time and In-full, and show the top risk.

Store ranking

Generate yesterday's store outbound ranking. Explain the weakest metric and the next follow-up action.

POD closure

Check today's order closure risk. Show ready-for-pick, ready-for-send, overdue or near-cutoff orders, and any unavailable POD/payment data.

Shortage

Show today's shortage situation by department and identify the items or departments that need follow-up first.

Picking

Check today's picking efficiency and first-pick success. Tell me whether the issue is capacity, speed, or shortage-related.

LINE push

Create a LINE-ready OTIFS alert card for manager review. Include severity, key metrics, root cause, and 1-3 action options.

Schedule

Create a daily 07:30 scheduled outbound briefing push. Use the manager-review template and include a detail page link.

Knowledge

Explain how outbound order handover works, and list the normal checkpoints where orders may get stuck.

Post-workshop

Practice Cases Before Follow-up

Case	Task	What to verify
1	Build a cloud app dashboard for outbound overview and open it in the browser.	App opens without 401, cards render data, date picker reloads numbers.
2	Ask for order closure risk during a busy shift.	Ready-for-pick, ready-for-send, overdue or near-cutoff buckets.
3	Ask why OTIFS is low.	Agent splits On-time vs In-full and recommends the next Skill.
4	Ask a process or metric explanation question.	Answer comes from knowledge module or says knowledge is missing.
5	Try a LINE template preview or test push flow.	Template content, test notification, detail page link, manager review step.
6	Record one missing business rule.	Classify as Area/OD mapping, POD/payment, Locus delivery, metric definition, or permission issue.

Feedback workflow: after each practice case, ask the Agent to append the result to the feedback sheet with scenario, prompt used, expected result, actual result, business comment, data gap category, priority, and follow-up owner.

Reference

Metric Appendix and Current Limitations

Store outbound ranking formulas

Field	Current formula
totalOrders	Count of outbound orders in the date range.
nonCancelledOrders	Orders where cancel time is empty.
shippedOrders	Orders where outbound time exists.
onTimeOrders	Shipped orders where timeout flag is 0.
inFullOrders	Non-cancelled orders where shortage flag is 0.
otifsOrders	Orders that are shipped, on-time, and in-full.
otifsRate	otifsOrders / nonCancelledOrders * 100.
onTimeRate	onTimeOrders / nonCancelledOrders * 100.
inFullRate	inFullOrders / nonCancelledOrders * 100.
cnShortageRate	Shortage orders / nonCancelledOrders * 100.

Current limitations to explain clearly

Area/OD

Current command is not Area/OD ranking.
mappingStatus = not-applicable is expected.
Final Area/OD ranking needs store-to-Area/OD mapping.

POD and delivery

POD/payment pending count and amount are Phase 2.
POD aging and payment method split are Phase 2.
Locus delivery status and delivery on-time are Phase 2.

Final decisions

Agent provides diagnosis and options.
Manager reviews and decides action.
Agent does not make staffing or operational final decisions.

Data permissions

Users only see authorized store data.
Cross-store comparison must stay within authorized scope.
Permission mismatch should be captured as setup feedback.