Main Conference Day 2, Wednesday, 5 December 2025

• How agentic workflows and no-/low-code platforms enable

business users to design AI-powered automations without full

developer dependency.

• Highlights the organizational challenge of balancing creative

freedom with compliance, quality control, and long-term

maintainability.

• Early lessons from deploying visual orchestration tools

for internal operations, with a view toward secure external

applications.

Eric Joachim Liese

Strategic Advisor and Lead Architect for AI & Data
BSH Home Appliances Group

See Bio

• Define which roles across business and tech functions need

• AI-related upskilling, and to what depth.

• Compare centralized vs. decentralized models for both GenAI

• implementation and capability building.

• Clarify how headquarters and local entities split responsibility

• for enablement, innovation, and compliance.

Slot reserved for one of our business partners

• Exploration of AI-driven change management techniques to

enhance operational excellence.

• Use of AI for strategic decision-making, including predicting

and managing change impacts.

• Automation of routine processes to increase productivity and

streamline organizational transformation.

Wioletta Strączek

Senior Director, GBS Finance Transformation and Lean Partner
Jacobs

See Bio

• How GenAI transformed our R&D Risk and knowledge

management by introducing our R&D chatbot “nexi”

• A solution built for 120 R&D specialists iand its integration into

a global IT AI strategy for a >10.000 employee company.

• Lessons learned and how we are approaching challenges

like organizational boundaries, resource and competence

constraints, accuracy, relevance, hallucination, speed, AI

adoption in the teams and balancing the AI hype within the

company

Ulrike Petrusch

Project Manager
Nexperia Germany Gmbh

See Bio

• Application of a signal-native generative model that

transforms magnetic and ultrasonic waveforms into 3D

corrosion and wall-thickness profiles for infrastructure like

pipelines and pressurized vessels.

• Integration of synthetic signal generation and simulation-

trained embeddings to overcome data scarcity, reduce expert

dependency, and deliver explainable, regulation-aligned output.

• Future potential for adapting the same architecture to other

signal-rich domains such as medical diagnostics, where real

data is scarce, and model bias must be minimized from the start

Benjamin Wolters

Head of AI, Data & Architecture
ROSEN Group

See Bio

• Why multi-agent systems are critical for complex, creative

tasks beyond the reach of single LLMs

• The architectural trade-offs: centralized vs. decentralized,

freedom vs. structure

• Examples from real-world cooperative agent + human

systems and upcoming frameworks

• Use of labeling platforms, entity recognition, and relationship

extraction tools to convert PDFs, emails, and legacy wikis into

structured knowledge graphs.

• Integration of curated knowledge into chatbots, copilots, and

decision support tools for employees and customers.

• Scaling annotation with human-in-the-loop QA and

alignment with business taxonomies and security controls

• Use of vector databases and semantic chunking to build RAG

pipelines that deliver accurate, explainable outputs in legal,

tech, and service use cases.

• Architecture patterns for combining public LLMs with private

corpora using on-prem, hybrid, or sovereign infrastructure.

• Governance of document updates, access control, and

hallucination risk mitigation in production RAG deployments.

• Addressing knowledge loss from workforce turnover by using

LLM-powered systems to support onboarding and reduce

training dependency on senior experts

• Deployment of retrieval-augmented generation using vector

databases to enable natural language queries across internal

documentation

• Managing inconsistencies in source material through

AI-supported synthesis and contextual reasoning across

fragmented content sets

Werner Rohrer

Leitung Data Science and Analytics
Öbb

See Bio

• Preparing and formatting internal content for AI systems

• Challenges with hallucinations and prompt tuning

• Demonstrating ROI and ongoing optimization of GenAI tools

Mike Hales

Product Owner
EE

See Bio

• How generative AI can address foundational challenges

in healthcare AI, including limited labeled data, and privacy

constraints

• The potential of synthetic medical images to support fairer,

more generalizable models and improve diagnostic workflows

in radiology and oncology

• Current limitations and open challenges in evaluating

generative models, including clinical relevance, realism, and

unintended artifacts

• Quality of life is a complex, multidimensional concept that is

difficult to continuously measure using conventional methods.

• Vocal biomarkers offer a promising alternative because

voice production is highly complex and sensitive to changes in

physical and mental health.

• A machine learning model was developed to analyze vocal

characteristics and objectively assess quality of life.

• Due to the complex data structure, a quantum-based

machine learning model was also investigated to capture

deeper patterns.

• Use of generative models to create brand-consistent

imagery across marketing, ecommerce, and documentation

workflows.

• Deployment of controlled image generation pipelines

using structured prompts, visual references, and predefined

templates to reduce reliance on stock photography.

• Integration with human review, metadata tagging, and digital

asset management systems to maintain aesthetic coherence,

IP safety, and regulatory alignment.

• Use of agentic AI systems to generate explainer videos,

training content, and product walkthroughs by combining

narration, imagery, and animation in real time.

• Chaining of LLMs, text-to-video models, and editing tools to

produce scene-structured outputs based on user intent and

enterprise knowledge.

• Incorporation of brand rules, compliance filters, and

multilingual adaptation to support global scalability and

internal reuse.