Case Studies / German Congress of Surgery (DCK 2024)
HEALTHCARE · CONFERENCE OPERATIONS
Automating session guidance and visitor support at Germany’s largest surgical congress
The 141st Congress of the German Society of Surgery runs four days with hundreds of parallel sessions, thousands of attendees, and a packed exhibition floor. Finding the right session, checking schedule changes, and navigating the venue was a constant challenge – handled by information desks that couldn’t keep up during peak hours. We deployed an AI-powered assistant that handled these tasks through real-time voice conversation, available continuously across both a large screen kiosk and a VR experience.
Client: 141st Congress of the German Society of Surgery (DCK 2024) – Leipzig Congress Centre, April 23–26, 2024.
KEY RESULTS
4 days
Continuous visitor support across the full congress
100s
Parallel sessions navigated — personalized recommendations
Voice
Natural speech interaction — no typing, no app download
INDUSTRY
Healthcare / Congress Operations
USE CASE
Automated session guidance & visitor support
AI APPROACH
LLM + RAG + speech processing
INTERFACES
Screen kiosk + VR (Meta Quest 3)
EVENT
DCK 2024, Leipzig (Apr 23–26)
SCALE
Design thinking → Agile delivery

The challenge
Large medical congresses are operationally complex. The German Congress of Surgery runs four days with hundreds of sessions in parallel — lectures, workshops, symposia, and live demonstrations spread across multiple halls. Thousands of attendees need to find the right session at the right time, check for schedule changes, locate exhibitor booths, and get information about the organizing institution.
Traditionally, this falls on information desks staffed by event personnel who need to know the full schedule, venue layout, and exhibitor details. During peak hours – especially at session changeovers – these desks get overwhelmed. Printed programs are outdated by the time the first schedule change happens. Event apps help but require downloads and typing. The result: attendees waste time, miss relevant sessions, and organizers spend money on staffing that still can’t cover the demand.
The core problem: a four-day congress with hundreds of parallel sessions and thousands of attendees – and no scalable way to provide personalized session guidance and event information without dedicated staff at every turn.
What we built
We deployed an AI-powered assistant that attendees could simply walk up to and ask questions – by speaking naturally, the way they’d ask a colleague. No app to download, no interface to learn, no typing on a phone.
Personalized session recommendations. With hundreds of sessions running in parallel, the most valuable feature was helping attendees find the right ones. The system knew the full congress schedule and could answer questions like “What’s happening right now on colorectal surgery?” or “Which workshops still have availability this afternoon?” – something an information desk staffer would need to flip through a binder to answer.
Full event knowledge. Beyond session scheduling, the assistant could answer questions about exhibitors, the organizing institution, venue navigation, and general congress information. All of this was powered by a retrieval system connected to the complete event database, ensuring answers were always current and accurate.
Voice-first interaction in a noisy environment. Congress floors are acoustically hostile – hundreds of simultaneous conversations, PA announcements, foot traffic. We combined professional microphone hardware with software-level noise reduction to maintain reliable speech recognition even in the loudest areas. Response times were optimized through parallelized processing to keep conversations feeling natural rather than stilted.
Two access modes. The assistant was available on a large screen kiosk for quick walk-up interactions, and also as an immersive VR experience through Meta Quest 3 headsets – giving attendees a choice between fast utility and a more engaging, exploratory interaction.

The results
BEFORE
Printed programs, static signage, and staffed information desks. Limited capacity during peak hours. No personalized session recommendations. Schedule changes communicated slowly.
AFTER
AI assistant handling visitor questions on demand. Personalized session recommendations from the full schedule. Continuous availability over four congress days – no waiting, no queuing.
The assistant operated across all four congress days, giving booth visitors immediate, personalized answers about which sessions to attend, where to find exhibitors, and how to navigate the venue – the kind of guidance that would otherwise require someone manually checking the printed program.
For event organizers and exhibitors, this demonstrates a practical model for enhancing visitor experience at large conferences. The same system can be reconfigured for any event with structured schedule data – trade fairs, corporate conferences, medical congresses – making it a repeatable solution, not a one-off installation.
Technology used
More Case Studies
See how we’ve helped other companies

AUTOMOTIVE (UNDER NDA)
Replacing reactive maintenance with a predictive maintenance roadmap
Unplanned equipment failures were causing costly production stops. We assessed the manufacturer’s infrastructure and delivered a concrete architecture for predictive maintenance, a roadmap from reactive repairs to data-driven prevention.
PdM
Architecture
Roadmap
Delivered

MANUFACTURING · RADAWAY
Making email-based order processing reliable with LLMs
Staff were manually reading customer emails, identifying products, and entering orders by hand. We turned a promising AI prototype into a production system that handles it end to end, across languages, formats, and attachments.
-90%
Manual intervention
95%+
Match accuracy

LOGISTICS · FR. MEYER’S SOHN
Eliminating manual data extraction from thousands of daily shipping emails
Operations staff were manually reading German and English logistics emails to pull out routing and scheduling data, every single day. We built an AI pipeline that extracts, structures, and delivers the data automatically.
–80%
Manual effort
2 langs
DE & EN
On-prem
Deployed

