Kick-Off Presentation of IBM Challenge Zürich
- 1. IBM watsonx Gen AI Challenge Welcome Event & Lecture Information Zurich, 27.2.2024
- 2. Agenda 18:15 Welcome speech by Christian Keller CEO IBM Switzerland Course introduction by Dean Heizmann 18:50 Use case presentations: PostFinance Twerenbold Freitag Schweizer Armee Victorinox Komax Group Zweifel Hirslanden SIX Group 19:50 Questions Closing & Aperitif
- 3. Course Description • Real use cases – real partners • All use cases in the field of generative AI • Your task is to form groups and think of a solution • Develop a PoC • Present it to the use case provider • We provide you with: o Technical tools and necessary education o Guidance on how to tackle your use case
- 4. General Information • Cross university Master's course - 6 ECTS • 6 Assignments (Not graded but indispensable) • Lectures online – Tuesday's 6:15PM to 8 PM • Lectures are recorded • Midterm presentation (ungraded): 23.4. • Final presentation: 29. 5. & 30. 5. • Best teams receive an invitation to visit the IBM Research Lab in Rüschlikon with an aperitif and a certificate • Unregister deadline: 1.3. 12:00
- 5. Goals and Requirements • Learning Goals: o Generative AI in business o Management of an AI project o Development of a Gen AI PoC • Requirements: o No coding needed but helpful o Technical affinity required o Communication and taking the initiative is key! • Target group: Wide range of master students from several economics and technical backgrounds.
- 6. Evaluation Grading happens in semi steps 1. Are the main requirements satisfied? 2. How are the technical possibilities explored? 3. How was the solution packaged and presented?
- 7. Schedule Welcome Event 27.02.2024 | 6:15 pm - 8:15 pm General introduction, goals and expectations, Apero (aperitive) and networking Introduction to generative AI in business 05.03.2024 | 6:15 pm - 8:15 pm Technical overview, use cases, and challenges of generative AI Large Language Models and their use cases 12.03.2024 | 6:15 pm - 8:15 pm Watsonx.ai overview & technical lab deep dive Chatbots and user interaction 19.03.2024 | 6:15 pm - 8:15 pm Watson Assistant overview & technical lab deep dive Foundations of a successful AI project 26.03.2024 | 6:15 pm - 8:15 pm Requirement engineering and Design Thinking for AI projects RAG 09.04.2024 | 6:15 pm - 8:15 pm Watson Discovery and Neuralseek overview Technical Deep Dive 16.04.2024 | 6:15 pm - 8:15 pm Generative AI: Transformers, GPT, training, deployment. embeddings, vectors Midterm Presentations 23.04.2024 | 6:15 pm - 8:15 pm Presentation of current state of project Integration Possibilities 30th April 2024 | 6:15 pm - 8:15 pm Overview of possibilities and specific integration examples Q&A 7th May 2024 | 6:15 pm - 7:15 pm We’re here to support you with your questions; Questions to be handed in beforehand Q&A 14th May 2024 | 6:15 pm - 7:15 pm We’re here to support you with your questions; Questions to be handed in beforehand Q&A 21th May 2024 | 6:15 pm - 7:15 pm We’re here to support you with your questions; Questions to be handed in beforehand Final presentation 1 28th May 2024 | 6:15 pm - 8:15 pm Final Group Presentation of Solution Final presentation 2 29th May 2024 | 6:15 pm - 8:15 pm Final Group Presentation of Solution Final Event TBA | TBA Rüschlikon LAB Tour for challenge winner groups
- 8. Team
- 9. Next Steps • You will receive an invitation to Slack and Box • Use Slack to look for team members or to find a group (4 to 5 per group) • Use case descriptions are in the box • Write an email to dean.heizmann@ibm.com with: • Group members (email, university, spoken languages and course of study) • Your top 5 use case preferences • We will have to distribute use cases evenly considering required languages and technical difficulty • Important: To unregister from the course write an email to dean.heizmann@ibm.com until 1.3., 12:00 AM! • Please do it asap so we can free slots for students on waiting list!
- 10. Questions?
- 11. USE CASES
- 12. About PostFinance 2.5 million Customers Total Assets 114 billion Swiss Francs (CHF) 1.3 billion Payment Transactions 34Branches and 57Consulting Offices ~ 3,600 Employees Every year 567,026 Coffees are consumed across locations. Average per day: 2,250 tell me more… let’s connect…
- 13. BUSINESS PROBLEM DIRECTION OF SOLUTION Scenario: Employees at the Customer Center PostFinance aim to resolve client issues, offer new products/services, or enhance existing client relationships through customer interactions. The goal is to gain a deeper understanding of the reasons customers reach out to the Customer Center. Users: Customer Center, Product Owner, Product Development Pain points: 1. Lack of understanding of customer needs and problems 2. High, manual effort for post-processing of calls 3. Large, previously unused amount of data Goal of the solution: Automatically convert and analyze Customer Calls to identify intents and suggest product or service enhancements. Scope to address: Focus on Swiss German Customer Calls. The process includes: 1. Transcribing spoken recordings to text 2. Using AI models to understand and anayze conversation content 3. Displaying call intents in a UI of choice Data: Synthetic audio transcripts of customer calls ClientClarity PostFinance Technical Difficulty: 1 Language Required: German Use Case Value in a Nutshell 2.7 million Customers Calls in 2023 at PostFinance CC Transcribing Customers asks about saving account interest Client Clarity Customers seeks better investment options
- 14. • Classical group tour operator • 1‘000 Trips • 3‘700 Departures • 50‘000 Bookings • + 3‘000 pages of print brochures • Own busses & own river cruise boats • Online booking ratio >70% • 130 years old family-owned business • Digital strategy • Improve personalization of all trave services • Increase automation / efficiency • No own coding resources About Twerenbold
- 15. BUSINESS PROBLEM DIRECTION OF SOLUTION Scenario: A customer has request and needs support from a Twerenbold sales agent Salas assitant: • Use Twerenbold’s online Documentation to serve clients' interests Customer requests • Change of personal information • Travel information Users: Twerenbold Customers with an active/valid booking, prospects Pain points: •Those simple requests consume valuable time from our agents •Customer can call/answer during office hours only (9-12; 13:17 ) •Caller identification with GenAI (elderly people) Goal of the solution: • Customer requests are done manually. With the support of GenAI, we aim to achieve a higher client self-service adoption rate. • Office hours are restricted to 09.00-17.00; GenAi should increase service levels toward a 24/7 client service hub. Scope to address: • Focus on clients with an active booking in place. Data needed: • Client booking data provided by Twerenbold AG • Customer Data / Traveler Data • Trip information • Website / Catalogs (API) Travel Advisor Technical Difficulty: 1 Language Required: German
- 16. ABOUT FREITAG <<INTELLIGENT DESIGN FOR A CIRCULAR FUTURE>>
- 17. BUSINESS PROBLEM DIRECTION OF SOLUTION Scenario: "Guuru" is a chat application that involves the community. Our own community is our first port of call when a customer has questions and wants to chat with someone. So far, the data generated has neither been analyzed nor used for our own chatbot solution. Users: • Employees • Community • Customer Pain points: Lack of insight into user behaviour and customer problems. Lack of quality assurance over our “Guuru” community. Goal of the solution: • Evaluate the data from the tool • Chatbot as an output against which we can analyze the data in the tool. Scope to address: • 2023 until today transcripts • focus on English requests • metadata scope will be defined • LLM is needed to generate the output of the questions to be analyzed Data needed: To solve this challenge, we will either provide you with a SQL database or an API access to pull the data from the system What the F-uck are they talking about? Technological Difficulty: 1 Language Required: English
- 18. Kommando Cyber Auftrag The commando Cyber ... - responsible for the provision of power in the cyber and electromagnetic space (CER). This in the areas of action management and mission-critical ICT; -ensures preparedness and assesses the feasibility of CERs in the operational spheres; - protects the mission-critical ICT infrastructure of the army in the CER; ... in favour of the Swiss Armed Forces and their partners in the Swiss security network.
- 19. Kommando Cyber Mission We give the armed forces the necessary knowledge and decision- making advantage in all situations. We combine innovation, technology, know-how and enthusiasm for mission fulfillment in a powerful Kdo Cy of the Swiss Armed Forces. We provide the expected and demanded services for the Swiss Army, the SVS and partner always precisely, to the point, coordinated, robust and secure.
- 20. BUSINESS PROBLEM DIRECTION OF SOLUTION Scenario: This PoC aims to demonstrate how Gen AI can be utilized to aggregate, analyse, and disseminate complex and rapidly changing information related to military technology advancements, use cases, capabilities and global defence trends. Users: Cyber command: Anticipation & Innovation Pain points: 1. Data Overload 2. Accuracy and Reliability 3. Predictive Analysis and Trend Identification 4. Resource constraints 5. Adapting to Technological Advancements 6. Cross-Departmental Coordination Goal of the solution: Developing an anticipation engine which allows Swiss Armed Forces Cyber Command to navigate future trends in accordance to predefined business rules. Create a user interface to communicate findings. Scope to address: Define Sources ,Scrapping, Matching, Recommendations, Visualization of output Data needed: Use algorithms are designed to collect data from diverse and reliable sources. Including intelligence reports, global surveillance data, research publications, and real-time news feeds. The AI system then processes this data, identifying key patterns, trends, and actionable insights. Swiss Army Business Rules. The sources will be distributed to the participating teams. Swiss Armed Forces Cyber Command Anticipation Engine Technical Difficulty: 3 Language Required: German, English
- 21. About Victorinox 1884 1945 1989 / 1999 2005
- 22. About Victorinox 2021 Launch of our digital learning platform C.A.R.L. We create a positive and vibrant learning culture that embraces growth and transformation. • 200+ trainings • 2.300 users worldwide (1.850 internal users) • 8 languages
- 23. How can generative artificial intelligence support us in customizing our user’s experience (during onboarding) on the digital learning platform? About the challenge Internal Users • Indirectly productive employees • Sales subsidiaries (worldwide) • Own retail • Directly productive employees External Users • Distributors and Retailers total = 2.300 users worldwide Scenario • 50+ new trainings every year • Mandatory / recommended / voluntary trainings • Automatic assignment to new user accounts (all at once) during onboarding Pain Points • High number of trainings • Missing overview of where to start and where to end • Depending on the function/position number of trainings to complete varies • No automatic recommendations on trainings matching the employee’s job profile / work topics
- 24. About the challenge Scope • Creation of a recommendation engine for general trainings (e.g. Code of Conduct) & role-specific trainings (e.g. S&OP) • Focus on indirectly productive employees (=office workers) and Sales Subsidiaries; own retail and directly productive employees as target group are out of scope • Ensure that the ideas/concept developed may also be applicable to onboarding of external users. Goal of the Solution • Provide orientation on their learning/onboarding journey • Ensure relevance (match & recommend trainings based on user’s position within company, job profile, preferences & interests) • Automatically adjust the degree of difficulty to ensure users stay engaged and challenged • New approach (Victorinox already uses Bing AI Enterprise which could be integrated/applied; recommend other options?) Data Needed • Training Descriptions and Target Group Information • User Feedback / Ratings • User Profiles Technical Difficulty Language English German is beneficial
- 25. Questions? Joël Maier Digital Learning Manager & UX Architect Katharina Neumann Learning Consultant & Project Manager Victorinox AG Schmiedgasse 57 6438 Ibach-Schwyz Switzerland T +41 41 818 12 11 www.victorinox.com
- 26. About Komax Komax Group: Global market and technology leader in solutions for automatic and semiautomatic wire and cable processing. • Cutting and stripping • Attaching connectors • Testing • Production planning – MES software • And much more In virtually every industry: • Cars, trains, airplanes • Electronic devices, communications equipment, appliances • Medical devices • Etc. Wire Processing Testing MES Software
- 27. BUSINESS PROBLEM DIRECTION OF SOLUTION Scenario: The aim is to convert solution proposals from existing service cases into know-how articles. The service cases are available as e-mail dialogs between customers and service technicians. Users: Internal technical service team Pain points: • Very large amount of data that is not manageable by hand. • Complexity is too high to be solved with a normal algorithm. Goal of the solution: Solution approaches (know-how articles) for previous service problems that are as meaningful as possible. Scope to address: The LLM should be able to recognize — independently of the dialog — how to fill the know-how article (Excel). Data needed: • Export of the service cases as an e-mail dialog • Example of a know-how article Service Cases Technical Difficulty: 2 Language Required: German, English
- 28. • founded in 1958 and family-owned ever since • 400 employees • more than 70 types of chips & snacks • 10’000 tons of chips and snacks • Swiss made with swiss ingredients • SBTi committed About Zweifel Pomy-Chips AG
- 29. BUSINESS PROBLEM DIRECTION OF SOLUTION Scenario: This PoC aims to demonstrate how GenAI can be utilized to: 1. segment complaints accordingly (ideally also taking provided images into account) 2. create responses for the customer: the answers generated by the GenAI engine should be provided with an accuracy rating such that selective and manual checking is possible for ambiguous complaints Users: Zweifel consumer complaint department Pain points: • time consuming process à Consumer complaints department is currently reading and manually answering 2’000 cases p.a. • limited knowledge of the complaints structure and the changes over time • Adapting to Technological Advancements Goal of the solution: Developing an AI engine which allows Zweifel to first classify the type of complaint and create a ready-to-send response to the customer. Ideally, the complaint department must only quickly look over the text and can send it out without the need of further modification. Scope to address: • mandatory: use of language mode to cluster client text and generate replies • ideally: reviewing the existing accumulation of complaints and questioning them based on the most recent complaint data • ideally: use of image models to use uploaded consumer photos for complaint segmentation Data needed: existing complaints: about 13’000 data records including images (from 2017 until now) Complaint Segmentation and (Semi-)Automation Technological Difficulty: 2 Language Required: German, English
- 30. • Hirslanden represents high-quality, integrated healthcare for individuals at all stages of life, from birth to old age and from prevention to cure, both outpatient and inpatient. • The company differentiates itself through superior medical and nursing quality, service excellence, and a warm atmosphere, supported by highly qualified specialists, excellent care, and specialized medical centers. • As Switzerland's largest medical network, Hirslanden plays a crucial role and commits to innovation to meet the evolving needs of society and to continue providing high-quality healthcare in the future. Through additional services and the use of technology, the differentiation in service offerings is emphasized. About Hirslanden Group
- 31. Klinik Hirslanden Zurich Klinik Hirslanden has been committed to the well-being of its patients with passion and dedication since 1932. We are proud that we have been contributing to the provision of healthcare in the Canton of Zurich for over 90 years. Klinik Im Park Zurich For over 30 years, Klinik Im Park has been a guarantor of medical excellence, professional care and a warm atmosphere. We see it as our task to provide our guests with personalised care to ensure their safety and well-being. Medical services at the highest level and a modern infrastructure are a matter of course for us. About Hirslanden Group
- 32. BUSINESS PROBLEM Scenario: The Klinik Hirslanden and the Hirslanden Klinik Im Park operate their own LinkedIn accounts and have an internal communication app "Beekeeper," which functions similarly to a social media platform. The Hirslanden Clinics in Zurich specifically encourage employees and partner physicians to act as ambassadors for the clinics by publishing their own posts on these platforms. Users: Employees and partner physicians Pain points: For authentic content, grammatically correct and adapted in form and language to the Hirslanden CI/CD, as well as properly prepared for the two channels, there is often a lack of time or the necessary experience. DIRECTION OF SOLUTION Goal of the solution: Automatic creation of channel-specific prepared posts, incorporating the linguistic specifications of publicly available information from both clinics (websites, LinkedIn, etc.) and including the CI/CD of the Hirslanden Group. Scope to address: The essence of the prepared posts is crucial: authentic, grammatically correct, according to the guidelines of the Hirslanden Group, including the use of emoticons and the creation of image suggestions. It should be quick and easy for all employees to create LinkedIn or Beekeeper posts without extensive prior experience. Data needed: Websites, LinkedIn, Hirslanden Language Guideline Technological Difficulty: 2 Language Required: German, English Hirslanden Voicehub
- 33. SIX Swiss Exchange, BME Exchange, BME Derivatives Exchange, SIX Digital Exchange Listing Trading Market Data Exchanges Clearing Settlement and Custody Securities Finance Tax Services Trade Repositories Cash Connectivity (Open Banking) Debit and Mobile Solutions Billing and Payments Banking Services Securities Services Third-largest stock exchange group in Europe Smooth payment transactions Unbeatable post-trade services from A to Z and more Four Areas of Activity. One Company. Reference, Corporate Actions and Market Data Tax and Regulatory Services Indices ESG Data & Solutons Display and Data Feed Financial Information Data You Trust IT Human Resources Marketing & Communications Legal & Regulatory Risk, Security & Compliance Finance & Services Corporate Functions
- 34. Sensitivity: C1 Public ESG Data Challenge Supporting Swiss market participants on their ESG journey 258 Swiss Listed companies 2 Pillars: Environmental and Social 2 Themes: Carbon Emissions and Social Risks 3 Data source types: Annual reports. ESG reports, Company websites
- 35. BUSINESS PROBLEM DIRECTION OF SOLUTION Scenario: As financial infrastructure provider in Switzerland, we see it as our responsibility to support the Swiss market participants in their ESG journey. Today, there is a lack of transparency in the way companies perform, when it relates to ESG. Two main areas need urgent actions: environmental footprint and social risks. We want to provide SIX listed companies with greater visibility on how they compare against others, allowing them to further improve. Users: • Companies listed / considering listing on SIX Exchange • Investors • SIX Product Development, Marketing, Communication Pain points: 1. Lack of transparency 2. Lack of data-harmonisation 3. Dependency of data providers 4. Green- & social washing prevention – is the reported data the same as what is communicated? Goal of the solution: To increase ESG performance and transparency, with specific focus on carbon emissions and social risks. Use Case 1: As a company, the user can understand how well the company is doing compared to its peers. Use Case 2: As investor, the user can improve their investment decisions (e.g. «green» investments based on more reliable and neutral data). The solution could be made available on a portal, powered by ChatGPT, with the ability to run queries on specific themes and topics. Scope to address: • Carbon emissions : Scope 1, 2 and 3 • Social risks: modern slavery, human rights, child labour policies and or controversies • Listed companies on the SIX Swiss Exchange Data needed: Data should be fetched from reliable public sources such as Annual reports, ESG reports and company websites. ESG Carbon Emissions & Social Risk Benchmark Language Required: English, German Technological Difficulty: 3
- 37. Thank you!