How Danube Group Cut Response Time 70% With a Bilingual AI Assistant

The situation

Reaching a salesperson was slow, and answers were inconsistent.

Reaching a Danube salesperson meant a convoluted, time-consuming path, so customers faced long waits and inconsistent answers. Manual handling introduced errors, and every lead depended on a person being available to capture and qualify it. For a business selling to large enterprises and local buyers alike, slow and error-prone service was a direct drag on conversion and satisfaction.

What we built

Aisha, a bilingual assistant that captures leads in chat.

We built Aisha on Azure OpenAI and the Microsoft AI Bot Framework, with natural-language support and English and Arabic voice. A simple typology matrix lets prospects explore projects by location, and the assistant captures contact details, records project interest, integrates an IVRS call-back and books site visits directly in chat. Fabric KQL logs every conversation for conversation intelligence. A responsible-AI guardrail keeps answers scoped strictly to Danube content; it was tested for bias, and we also used AI to test the assistant and improve its answers.

How it works

The architecture, stage by stage.

1. Engage

   • Aisha assistant with English and Arabic (EN / AR) voice
   • Azure OpenAI with the Microsoft AI Bot Framework

2. Capture and learn

   • Lead capture, IVRS call-back and site-visit booking
   • Fabric KQL logging every conversation for conversation intelligence

What changed

The outcomes, as the client measures them.

• Response time fell 70 percent.
• The error rate in interactions fell 75 percent.
• Reliance on manual customer-service intervention dropped 60 percent.
• Customer satisfaction and lead capture both rose.