# Use Case: Operations Connector — Customer PDFs to TMS-Ready Excel

URL: https://cloud28.ai/use-case-operations-connector.html
Category: Operations connector
Reference outcome: 5-day path from kickoff and sample review to a production-like test
Owner/operator: Cloud Nine Software · Contact: sales@cloudninesoftware.vn

## Summary

How Cloud28.ai adds agents to an operations workflow so incoming customer PDFs are extracted, checked, enriched, turned into Excel rows, routed for exception review, and prepared for transport-management-system (TMS) import.

## Client Profile

- Client: logistics or transportation operation receiving customer order PDFs by email.
- Workflow: shared mailbox monitoring, PDF intake, order extraction, container rows, port-code enrichment, Excel generation, folder delivery, and exception routing.
- Scale example: 10+ orders per day, with customer PDFs ranging from single-container orders to large multi-container files.
- Cloud28.ai role: add extraction, validation, enrichment, and exception-review support around the existing operations team.

## Why This Operation Needed Cloud28.ai

Customer PDFs contain the operating instructions that move containers from ports to warehouses, rail yards, or other destinations. The information arrives as documents, while the downstream process needs structured rows in an Excel template. Manual reading, copying, and port lookup cost delay, rework, and inconsistent handoffs. Cloud28.ai adds agents so routine extraction and validation happen continuously while exceptions still go to humans.

## Current State (Manual Path)

| Step | Manual pattern | Operational risk |
| --- | --- | --- |
| Email intake | Orders arrive as PDF attachments in a shared mailbox. | Urgent files wait until someone sees and opens them. |
| PDF reading | A user reads the order and identifies customer, container, port, rail, and destination details. | Important fields can be missed or interpreted inconsistently. |
| Excel row creation | Each container is copied into the required Excel template. | Large PDFs create repetitive work and copy/paste errors. |
| Port lookup | Users look up or remember port codes and mappings. | Unconfirmed mappings create downstream corrections. |
| Exception handling | Unclear layouts are handled ad hoc. | New customer formats can silently produce bad output if not caught. |

## Cloud28.ai Connector Roles & Human Control

| Connector role | What it does | Human control |
| --- | --- | --- |
| Inbox | Watches the agreed mailbox and routes new PDFs into the workflow. | Operations defines mailbox, labels, and priority rules. |
| Extraction | Converts customer PDFs into order and container-level fields. | Low-confidence fields are flagged instead of silently accepted. |
| Mapping | Enriches rows with port codes and stored reference mappings. | Unconfirmed mappings route to review and become reusable after approval. |
| Excel | Creates the agreed Excel format and delivery file. | Operations validates sample outputs before production use. |
| Exception | Sends unknown layouts or missing required fields to the team. | Humans decide how to handle new formats and edge cases. |

## Five-Day Delivery Path

- Day 1 — Kickoff, mailbox rules, target Excel template, sample PDFs, and first extraction path. Signal: a sample PDF becomes structured rows.
- Day 2 — Container-level parsing, required-field checks, and port-code mapping table. Signal: common fields extracted and mapping gaps visible.
- Day 3 — Excel generation, folder delivery, and exception notifications. Signal: the workflow produces a file and routes exceptions.
- Day 4 — UAT against representative customer PDFs, including multi-container examples. Signal: supported PDFs meet the required output format.
- Day 5 — Production-like test, handover notes, runbook, and improvement backlog. Signal: team can operate, review exceptions, and add new formats.

## Measures to Track

| Measure | Baseline | Target signal |
| --- | --- | --- |
| Manual effort | Minutes spent per container copied into Excel. | Routine supported formats require little or no manual retyping. |
| Cycle time | Time from email receipt to usable Excel file. | Supported PDFs move quickly unless an exception requires review. |
| Error and rework | Wrong container, port, customer, rail, warehouse, or missing-field corrections. | Structured validation catches missing and low-confidence fields. |
| Mapping reuse | Port codes and customer-specific mappings depend on memory. | Approved mappings are stored and reused. |
| Unknown layouts | New customer PDFs are handled ad hoc. | Unknown layouts are caught and added to the improvement backlog. |

## Control Principle

Agentic processing should not hide uncertainty. If a PDF layout is unknown, a required field is missing, or a port mapping is not trusted, the connector stops that file for review and makes the exception visible.