Skip to main content

Smart Services with AI and Own Data: Insights from the Service Lunch

By Jürg Meierhofer, ZHAW

The recent “Service Lunch: Eating, Learning, Networking” event featured Chris Bochsler from Cando, who shared practical examples of how AI can enable innovative smart services in the field of energy management. By leveraging customer documents and real-time data, Cando successfully linked these resources to an LLM (RAG) framework, resulting in more relevant and valuable outcomes.

Cando’s approach involved integrating the LLM (RAG) framework into various apps. This strategic move also earned them the “Best of Swiss App” Award in Innovation.. By seamlessly incorporating AI capabilities, they enhanced user experiences and delivered valuable services. The LLM (RAG) framework demonstrated versatility by catering to both smaller and larger companies, adapting to different contexts and providing customized services based on specific needs.

One example is Koster Home, a home energy manager, which integrated the LLM (RAG) framework into its home system, including components like heat pumps and photovoltaic (PV) panels. The system helps users determine optimal times for charging their electric cars or suggests solutions for low shower water pressure.

Cando’s energy management system integrates numerous data sources, such as production trends or weather forecasts. For instance, if a steep temperature increase is predicted, the system recommends adjusting heating settings. Or it issues warnings when anomalies occur, translating technical information into a user-specific language (e.g., layperson vs. technical expert).

Learnings included the need to monitor quality, balance data quantity, and stay informed about rapid AI development. The technical approach using the LLM (RAG) framework proved effective. The system combined structured and non-structured data, contextualized information, and utilized vector embeddings. Looking ahead, Cando aims to ensure operational stability through AIOps, maintain quality assurance, or explore ontologies and knowledge graphs. Their innovative use of AI and data integration exemplifies how smart services can evolve and adapt.

Successful Workshop on Technical Risk Mitigation & Insurance

By Jürg Meierhofer, ZHAW

In a half-day workshop organized by the Expert Group Smart Services on March 28, participants from seven different organizations (data science, tech, sensoring systems, insurance, service design) came together to explore effective strategies for managing risks of operating industrial assets. The workshop focused on various aspects of risk mitigation, including preventive measures, impact reduction, and decision-making processes. 

Preventing risks before they occur is crucial for maintaining a secure and efficient business environment of industrial assets. Our discussions highlighted the preventive measures primarily by technical solutions. Leveraging technology, such as sensor-based monitoring systems, enhances risk detection and early intervention. By proactively identifying potential threats, organizations can prevent adverse outcomes.

However, technologies for preventive and reactive measures open the door for prescriptive measures. This can result in optimizing the performance of an asset over its lifecycle. In this manner, risk mitigation expands into a broader context of asset value optimization, while insurance adopts the perspective of investment optimization. Departing from the conclusion of the predecessor workshop in August 2023 (https://databooster.ch/databooster-workshop-managing-financial-risks-of-smart-services/), the discussion moved to the question of the extent to which technical solutions are scalable or have to be developed from scratch for each application case. We concluded that certain verticals can be effectively standardized for scalability, while others remain within the realm of specialized solutions.

Expert Day – FHNW

By Milena Perraudin, data innovation alliance

The Expert Day on March 19, 2024 was held for the first time at the University of Applied Sciences and Arts Northwestern Switzerland (FHNW) in Brugg-Windisch. Welcoming over 40 participants, the event enfolded plenty of opportunities for engagement within the three Expert Groups: ‘Smart Maintenance’, ‘AI in Finance & Insurance’,  and ‘Spatial Data’ as well as the Expert Group in planning ‘Governance for Growth with Data & AI’. For those interested in joining the latter group, please contact us at info.office@data-innovation.org.

The event started with two thought-provoking keynotes. The first ‘Eyes on Human-Data Interaction’ by Prof. Dr. Arzu Çöltekin, FHNW, highlighted how eye movement tracking can enhance human-computer interactions, emphasizing the importance of designing data-based products with the end user in mind. The second keynote, ‘Lessons learned on scaling after 1 year of GenAI’, by Dr. Marcin Pietrzyk, co-founder and CEO of Unit8, shed light on the significance of GenAI and its mostly underused potential in realizing value in production at scale. A fact that the data innovation alliance wants to address together with the Innovation Boosters Artificial Intelligence and Databooster, powered by Innosuisse.

Following the keynotes, participants engaged in four breakout sessions, each focusing on a specific topic. The following paragraphs will give you a short insight into the different breakout sessions.

Workshop ‘The Future of Financial Data Analytics’ by Expert Group AI in Finance & Insurance

The Expert Group meeting jointly organized by Marc Vendramet, Andreas Blum from Unit8, and Branka Hadji Misheva from BFH, delved into two primary topics. Firstly, it offered a comprehensive exploration of AI & analytics use cases within the finance sector. Secondly, it involved brainstorming on the group’s focus areas for the upcoming year, including discussions on new event formats and collaborations.

Highlights were talks by Nicole Königstein of Wyden Capital, who shared insights on financial times series prediction in the age of transformers, and Guillaume Raille from Unit8, who presented various examples of LLMs’ use beyond chatbots, along with a detailed discussion on the challenges and opportunities associated with applying advanced LLM technologies to real-world cases.

Workshop Governance for Growth with Data & AI

Participants of this workshop gathered to delve into the full spectrum of Governance and Growth theme, as well as the opportunities with Data & AI. Frank Seifert, from adesso Schweiz AG, introduced a comprehensive management model for Governance, while Dr. Omran Ayob, from SUPSI, delved deep into the critical aspect of explainability in data governance. Participants engaged in lively discussions about the challenge of achieving higher data or model transparency (Explainability) while avoiding privacy issues.

The tension between the desire for a minimum level of governance and the need to work simply and leverage Data & AI opportunities became apparent. On one hand, holistic approaches are necessary, but on the other hand, there is a need for concrete, easily implementable measures. It’s precisely this tension that makes the topic so intriguing.

The group will continue working in this area by examining various layers of the topic, especially legal, technical, and ethical aspects, and developing Governance approaches and impulses for Growth opportunities. Our current vision includes guidelines and processes on one hand, and concrete tips and tools on the other.

Workshop ‘Hybrid approaches to intelligent maintenance’ by Expert Group Smart Maintenance

First, Dr. Kai Hencken from ABB corporate research started with an overview of predictive maintenance strategies at ABB, emphasizing the utility of traditional statistical reliability models in addressing data scarcity challenges.

After, an open discussion among participants followed, exploring potential use cases for AI integration with domain knowledge in their respective companies. Examples included condition monitoring of trains and digital twins for gas turbines. Technical solutions such as few-shot learning and the role of LLMs and generative AI in advancing condition-based and predictive maintenance were also examined.

The day concluded with an Apéro, fostering further discussions about innovative ideas among participants. We are already looking forward to the next event, as we are convinced that together we move faster.