Case Study

SwiftFreight
AI Workflow Transformation

How we eliminated 22 hours of weekly manual data entry at a mid-sized logistics company by deploying a custom AI workflow engine connected to their Salesforce CRM.

22 hrs

Saved Per Week

91%

Data Accuracy Increase

4.2x

ROI in 6 Months

Start Your Project AI Workflow Services

AI Processing

400+ docs/day automated

ROI Achieved

4.2x in 6 months

Project Overview

SwiftFreight is a mid-sized logistics and freight-forwarding company managing over 400 inbound shipment documents per day—Bills of Lading, customs declarations, and carrier invoices. Each document required a staff member to manually read, extract key data fields, and type them into Salesforce CRM records, creating chronic bottlenecks, transcription errors, and delayed invoicing.

Intellectual Clouds was engaged to design and deploy a custom AI workflow engine that could read unstructured PDFs and emails, extract the relevant data fields, and automatically create and update Salesforce records—completely eliminating the manual entry process.

Client

SwiftFreight Ltd.

Industry

Logistics & Supply Chain

Timeline

8 Weeks

Services

AI Workflow Automation, Salesforce Integration, Custom AI Development

The Challenges

Massive Manual Data Entry Volume

Operations staff spent a combined 22 hours per week manually typing data from inbound shipping documents into Salesforce—time that could not scale with business growth.

High Error Rate

Manual transcription resulted in a 12% data error rate on Salesforce records, causing downstream billing errors and compliance flags with customs authorities.

Delayed Invoice Processing

Because data entry was a bottleneck, invoices were being generated 2–3 days after shipment arrival, causing cash flow issues and late payment penalties.

Unstructured Document Formats

Carrier documents arrived in over 30 different PDF formats and email templates, making traditional rule-based automation completely infeasible.

Before Automation

22 hrs/week

manual data entry

Our Solution

We designed a 3-layer AI pipeline that intercepts incoming documents, intelligently extracts structured data using a fine-tuned extraction model, and syncs the output directly to Salesforce via its REST API.

Intelligent Document Parser

We deployed a custom LLM-based extraction model (built on GPT-4o) trained to recognize SwiftFreight's specific document vocabulary and field structures across all 30+ carrier formats.

Email & Webhook Ingestion Layer

Documents arriving via email were automatically parsed via a Make.com webhook pipeline, while carrier portal PDFs were uploaded via a custom drag-and-drop interface we built.

Salesforce Auto-Record Creation

Extracted JSON data was mapped directly to Salesforce Shipment records, automatically populating 18 key fields including weight, origin, destination, carrier, and customs reference numbers.

Human-in-the-Loop Validation

For low-confidence extractions (below 85% confidence score), the system flagged records for a 30-second human review rather than auto-creating, ensuring accuracy without sacrificing speed.

Implementation Timeline

Week 1–2

Discovery

Document format audit
Salesforce field mapping
API integration design

Week 3–4

Model Training

Sample data labelling
Extraction model fine-tuning
Confidence threshold calibration

Week 5–6

Integration Build

Make.com pipeline build
Salesforce REST API connector
Human-review UI development

Week 7–8

UAT & Go-Live

Parallel run testing
Staff training (2 hrs)
Full production handover

The Results

Within the first month of full deployment, SwiftFreight's operations team recovered more than 22 hours per week of staff time, achieved near-perfect data accuracy, and slashed invoice generation cycles from 3 days to under 4 hours.

22 hrs

Saved weekly across the ops team

91%

Reduction in data entry errors

4 hrs

Average invoice cycle (was 3 days)