Back to Insights
Automation

Automating Logistics in Abu Dhabi: How AI Voice Agents Cut Reporting Time by 90%

Jan 1, 2026
4 min read
Verge Sphere Team
Automating Logistics in Abu Dhabi: How AI Voice Agents Cut Reporting Time by 90%

In the logistics sector of the UAE, time isn't just money—it's reputation. Recently, a prominent transportation client in Abu Dhabi approached Verge sphere with a critical operational bottleneck: Incident Reporting.

Their fleet of 200+ trucks was managed by drivers who spoke a mix of languages—Arabic, Urdu, Hindi, and English. When a breakdown occurred on a remote stretch of highway, the driver would call a dispatcher. The dispatcher, often overwhelmed, would manually transcribe the driver's verbal report into an Excel sheet. This game of "telephone" resulted in:

  • Inaccurate GPS coordinates: Leading to recovery trucks going to the wrong exit.
  • Data Loss: Maintenance details were often lost in translation.
  • High Latency: It took an average of 45 minutes just to log the ticket and dispatch help.

The Solution: Polyglot AI Dispatcher

We didn't just build a chatbot; we built a voice-first interface using React and OpenAI's Realtime API. We realized that forcing drivers to type was the problem. They needed to speak.

The Workflow

  1. One-Tap Recording: The driver opens the PWA and holds a big red button.
  2. Natural Language Processing: They speak in their native tongue (e.g., Urdu mixed with English terms).
  3. Entity Extraction: The AI identifies Vehicle ID, Issue Type (e.g., Tyre Burst), and Location.
  4. Structured Logging: The system automatically pushes a clean JSON object into the company's Notion Database and alerts the fleet manager via WhatsApp.

The Technical Challenge

The biggest hurdle was latency and accents. Standard speech-to-text models struggled with the specific "Gulf English" or mixed-language dialects. We fine-tuned the system instructions to recognize local landmarks and fleet-specific terminology (like "LPO number" or specific tyre brands).

The Result? Reporting time dropped from ~2 hours of admin work per day to less than 60 seconds per incident. The monthly maintenance reports in Notion are now auto-generated, giving management clear visibility into fleet health without hiring extra data entry staff.