BPE / FLEET — TRAINING

Copy this sample data exactly to register a pickup truck for testing:

Click "Add Vehicle". You will be redirected to the Vehicles list. Find WKL 5522 in the list. Its status should show Available.

After entering Destination and Estimated Days, click the purple "AI Suggest" button. Gemini will:

• Score all available vehicles (0–100)
• Rank by service buffer, expiry dates, and suitability for the destination
• Show reasons for each recommendation
• Highlight vehicles with critical alerts in red
• Match vehicle type to your task — e.g. "repairing high building" will recommend a sky lift, not a van

Click "Submit Requisition". The form is assigned a reference number (e.g. VRF-2026-0042) and status changes to Pending Fleet PIC.

On the requisition page, scroll to "No BPE Vehicle Available?" panel. Click "Mark as Rental Vehicle", enter the vendor name and PO reference, then click "Confirm Rental".

The requisition shows a RENTAL VEHICLE badge in Section A. The pre-trip BPE inspection is not required — proceed straight to HOD approval and handover.

At the bottom of the requisition, click "Verify & Forward to HOD". Status changes to Pending HOD. HOD is notified automatically.

Log in as HOD or PM. Go to Requisitions, filter by Pending HOD. Open the form and click "Approve Requisition".

Or click "Reject", enter a reason, then click "Confirm Rejection" — the requestor will see the rejection reason.

Go to Inspections, click "New Inspection". Link it to requisition VRF-2026-0042, set type to Request (before handover). Click "Start Inspection". On the next screen, upload the 7 photos downloaded above — tag each with the correct angle.

When you upload the odometer photo (tagged odometer), the system automatically triggers OCR AI. Wait 3–5 seconds for the km value to auto-fill.

Click the purple "AI Damage Scan" button. Gemini analyses all 4 exterior photos (front, rear, left, right). After ~10 seconds you'll see a result panel like this:

Click "Auto-apply to checklist" to pre-fill the checklist rows with AI findings. Then click "Confirm findings" to lock the AI result.

Go through the physical vehicle and rate each item. Use this reference:

Once photos are uploaded and no critical items are rated 3, click "Submit Inspection". The system generates e.g. INS-2026-0118 linked to VRF-2026-0042.

Click "Hand Over Vehicle (Mark In Use)". Status changes to In Use and vehicle status becomes in_use. The vehicle is now locked — it will not appear in AI vehicle matching for other requests until returned.

Log in as Requestor. Open VRF-2026-0042. Once a return inspection exists, click "Complete Section B (Return Vehicle)".

Upload the odometer photo from the return trip — AI will OCR the reading AI and check for mileage anomalies.

Go to Inspections → New Inspection. Set type to Return (upon return), link to VRF-2026-0042, click "Start Inspection". Repeat the 7-photo process and run AI Damage Scan again.

After the return inspection is submitted and approved, click "Close Requisition". Status changes to Closed. Vehicle status returns to available.

Total km driven = 54,842 − 54,320 = 522 km. This is recorded and the AI mileage anomaly check runs automatically.

Review the AI classification. Accept the category or change it if needed. Click "Submit Request" (button changes to "Log Urgent Request" if urgency ≥ 4). Reference number e.g. RPR-2026-0031 is generated.

Click the Q&A tab on the Documents page. Try these sample questions:

Upload the signed PDF undertaking letter (Appendix 1 of LGT-P02). Once uploaded and marked Valid, that user can have requisitions approved.

Go to Reports. Click "Generate Monthly Summary". Select the month (e.g. April 2026). This triggers a BullMQ job that:

• Queries all requisitions, inspections, repairs for the month
• Calculates fleet utilisation rate, total km driven, repair costs
• Gemini writes executive summary, findings, and recommendations
• Generates a PDF — download link appears when done (~30 seconds)

The system runs three scheduled background jobs automatically — no action required from Fleet PIC:

On the Dashboard, each alert has an "Acknowledge" button. Click it after you have taken action (e.g. booked road tax renewal). This removes it from the critical list. Acknowledged alerts are still stored — they appear in the monthly compliance report.

Click the purple Refresh All button at the top right. The system queues a job that walks every non-decommissioned vehicle, recomputes deterministic stats, asks Gemini for narratives, and writes everything back to the predictions table. Typical run: 30–90 seconds. The page polls every 3 seconds until the new generated_at timestamp arrives.

Each vehicle row shows four prediction blocks:

Each vehicle has a Risk badge in green / amber / orange / red:

If you've just logged a new repair or a fresh inspection for one vehicle, click the small Refresh button on the right side of that vehicle's card. This runs a single-vehicle predict_vehicle job (a few seconds) without re-processing the whole fleet.

Use the three filter chips above the vehicle list:

• All Vehicles — every active vehicle, sorted by risk score then failure ETA.
• At Risk Only — risk score 4 or 5. Use this view in your daily standup.
• Service ≤ 30 days — vehicles where the predicted service date falls within the next month, so you can pre-book the workshop.

Predictions are decision support, not autopilot. The system never auto-orders parts, books workshops, or pulls vehicles from active requisitions. Fleet PIC always confirms the action — exactly like every other AI output in BPE-FMS. Numbers come from real database rows; Gemini only adds the narrative reasoning. If the worker is offline or Gemini fails, the page still shows accurate baseline statistics.