Case Study - AI Agents

Overview

OKM GmbH is the worldwide leading manufacturer of high-end metal detectors, ground scanners, and treasure hunting equipment. Their customers range from hobbyist detectorists to professional archaeologists and security personnel, resulting in a high volume of inquiries across WhatsApp, email, and web forms.

We built a multi-channel AI agent system that automates customer interactions across all three channels while seamlessly integrating with their existing CRM.

What the agents handle:

• Lead qualification — Instantly filtering spam from genuine prospects, ensuring the sales team only engages with qualified leads
• Technical support — Answering common questions about product features, specifications, and operation
• Scan analysis assistance — Helping customers interpret detector scan results and troubleshoot issues
• Sales conversations — Collecting vital information from prospects (budget, use case, timeline) so the human sales team has everything they need before reaching out
• General Q&A — Providing instant responses about products, shipping, warranties, and processes

When conversations require human expertise, the agents intelligently hand over complete context to the support team.

What we did

We drew on our prior experience building accessible, easy-to-set-up AI agents for Rödd AI in this project.

We built the AI agents using Python with LiveKit for real-time conversational capabilities. The system uses an embedded FAISS vector database for semantic search across product documentation and support knowledge base, enabling accurate, context-aware responses.

The deployment runs on Railway for reliable, scalable infrastructure. All channels (WhatsApp, email, web forms) connect via webhooks and feed directly into OKM's CRM, ensuring a unified view of every customer interaction.