Project_Finn__presentation_Big_Data_Expo__aelhassouni___3_.pdf

Transcript

1. Project Finn bunq's journey to GenAI-driven operational excellence Big Data Expo, Utrecht | September 13, 2023 | Ali el Hassouni
2. 01 02 03 04 05 The need for change What is GenAI? Project Finn Key Applications Agenda Results 06 Ethical considerations
3. bunq: the bank of The Free Non-traditional banking No physical locations, progressive mindset and always reinventing Rapidly growing 10+ million users across the EU AI for fraud detection, customer support, and performance marketing Tech-savvy, user-focused
4. ❖ Limitations of traditional rule-based models in various domains ❖ The demand for scalable and intelligent solutions The need for Change
5. • Lack of adaptability Rule-based systems can't quickly adapt to new challenges or insights. High maintenance Constantly updating rules to keep up with changes False positives Rules introduce a trade-off in terms of effectiveness. Ineffectiveness Struggles to optimize campaigns in real-time or predict future consumer behavior. Inflexibility Limited to predefined scripts, lacking personalized engagement. Limitations of traditional rule-based models
6. • Operational efficiency Need for solutions that scale with growth without adding complexity. Dynamic customer interactions Desire for systems that can offer personalized, context-aware support Data-driven decision making Solutions need to analyze large data sets in real-time to inform strategic decisions Future-proofing Ensuring the system can evolve and adapt to future challenges and opportunities Intelligent marketing Requirement for real-time adaptability and predictive capabilities Demand for scalable & intelligent solutions
7. Emergent abilities: go beyond training data, enhancing utility and flexibility Contextual understanding: revolutionizes interactions with data and other intelligent systems Advanced pattern recognition: uncovers hidden patterns in vast data LLMs Search and retrieve info Converse with AI Generate content
8. Project Finn Finn V0.1: Zero-shot learning Input prompt Proprietary pipeline including embeddings, vector dbs, ranking etc bunq knowledge base: Bunq together, Google Suite etc…. Query Context Generative AI LLM: Palm, GPT4, LLama etc… Relevant results Response
10. Draft acceptance: cuts agents’ time and effort. Automated tickets resolution: speeds up support, quicker resolutions. Intelligent Support Ecosystem: boosts user satisfaction and operational efficiency. User support
11. Personalized ad campaigns: adjusts campaigns in real-time improving engagements. Quicker A/B testing: automates split testing for better short-term and long-term gains. ROI synchronization: targeted, personalized approach enhances short-term and long-term results. Automated marketing
12. User-Centric navigation: analyzes user behavior to dynamically rearrange or highlight in-app features. Context-aware recommendations: offers personalized product or feature recommendations. Personalized user experience: increases overall app usage and loyalty. In-app personalization
13. Data privacy: stringent data protection measures to ensure user data is secure. Bias mitigation: active efforts to eliminate biases in AI models, promoting fairness and inclusivity. Transparency: commitment to transparency when it comes to AI at bunq. User consent: explicit user consent before each use of confidential user data. Accountability: regular audits and quality checks to uphold ethical standards. Ethical AI
14. Questions?