Artificial intelligence
AI that is in production.
We use AI models in running systems. This page lists the applications we have built and how we integrate them into existing workflows.
Voice bots for appointment booking and customer service
Our voice assistants answer calls, book appointments and recognize the caller's request. When needed, they hand over to a member of staff. The bots are built on Parloa and connected to the customers' booking systems. All dialogs are in German.
- Appointment booking for tire changes through the booking interface of a tire and car service network
- Further voice bot projects for a nationwide car service chain and an association of tire dealers
- The same dialogs run as a chatbot on WhatsApp.
Receipt and image analysis
Multimodal models read receipts, forms and product photos, check rules and return structured data. In our sweepstakes platform this analysis runs continuously, with duplicate detection and rule checking.
Movement and posture analysis
Pose detection captures 33 body landmarks and runs locally in the browser. No image reaches a server. The method also applies to sports, ergonomics and rehabilitation.
Speech to text and text to image
Spoken instructions are transcribed with Whisper, for example for notes in exercise templates. We generate images with ComfyUI workflows on our own GPU hardware. No data is sent to an external image service.
Integrating AI into your systems
The most common case in practice: a model needs to become part of an existing workflow. We connect speech and image models to CRM, point-of-sale, booking systems or databases. This includes logging, cost control and defined behavior when the model does not respond. We use retrieval from your own documents (RAG) and multi-step agents where the benefit is measurable.