Sally and the University of Mannheim: ZIM Project Shaping the Future of Meeting Transcription

Sally is already one of the leading AI-powered meeting assistants in the German-speaking market. But the demand for precise, reliable, and context-aware meeting transcription continues to grow - especially because real-life meetings are far more complex than traditional online calls. To meet these technological challenges, Sally is being further developed through a joint ZIM project in collaboration with the University of Mannheim.

The University of Mannheim is one of the most renowned institutions in the fields of data science, AI research, and empirical analysis. Together, the project pursues one central goal: to create a new generation of AI-driven meeting transcription and analysis that clearly outperforms the standards of modern business communication.

Why This ZIM Project Matters

Meetings are an essential part of everyday work - yet the quality of their documentation often falls short of what companies truly need. Standard transcription systems quickly reach their limits: background noise, multiple speakers, overlapping discussions, informal language, or industry-specific terminology all significantly reduce accuracy.

This is where Sally, together with the University of Mannheim, steps in. The project researches and develops advances in:

• transcription models that actively suppress background noise
• speaker diarization for dynamic, real-world meeting situations
• automatic detection of action items directly from the conversation
• recognition of company-specific terms, abbreviations, and product names
• contextual understanding to capture meaning more accurately

These elements are not new - but achieving them in combination at a consistently high quality remains an open technological challenge. The ZIM project aims to close exactly this gap.

Transcription at the Highest Level: Sally Sets New Standards

Sally already achieves up to 98.8% transcription accuracy, outperforming most systems in the German-speaking market - especially in real meeting environments rather than controlled video-call conditions.

But the collaboration with the University of Mannheim goes far beyond this.

The goal is to deliver transcription that remains reliable even in demanding communication scenarios:

• multiple speakers talking simultaneously
• frequent speaker changes
• varying distances to the microphone
• natural room noises like chairs, paper, or keyboards
• industry-specific jargon and technical terminology

To achieve this, specially engineered test datasets are used, designed to realistically mirror real-world meeting conditions - fully anonymized and GDPR-compliant. This creates a scientifically robust foundation that goes well beyond traditional training methods.

Where Research Meets Application: Sally Evolved Through Science

What makes this ZIM project unique is the combination of academic depth and practical real-world applicability. The University of Mannheim contributes advanced expertise in AI modeling, speech processing, and machine learning. Sally provides the technical platform to translate this knowledge into tangible product innovation.

The result is a meeting assistant that doesn’t just transcribe - it understands.

With each step forward in the joint research project, Sally becomes more precise, more context-aware, and more reliable - setting new standards for the future of meeting transcription.