Customer Ignite Talk: Maurizio Poletto (Chief Platform Officer & COO, Erste Group) & OpenAI

### TL;DR Erste Group’s Chief Platform Officer and COO Maurizio Poletto shares lessons from scaling AI in a regulated banking environment: connect customer data from day one (rather than starting with risk-free data), embrace trial and error, and aim to serve the 80% of customers who have never received financial advice. ## Why AI Became a Platform-Level Priority at the Board Level Maurizio notes that the board was excited about AI from the start, but initially that excitement lived only in private reading and PowerPoint decks. Now the excitement remains, but it comes with plenty of real-world questions: how to roll it out to employees, how to keep it legally compliant, and how to convince the security team to open up rather than lock down. He believes these challenges are exactly where value is created—because if you don’t solve them, nothing gets delivered. Unlike blockchain (which was also exciting but lacked challenges to solve), AI is real and here to stay. ## Deciding When to Move Fast and When to Move Slow in a Regulated Environment Banks have to slow down when handling customer data. Maurizio stresses that trust is the foundation of banking—once broken, it is very difficult and expensive to regain. That’s why, when they started exploring AI about 2.5 to 3 years ago, they chose the hard path: instead of starting with customer-free content like help centres or generic information (as many banks do), they connected customer data right away. This led to many mistakes along the way, but also to invaluable knowledge that now lets them accelerate with more confidence. ## Lessons and Strategies for Rolling Out ChatGPT Internally Erste Group took a very simple approach to productivity tools like ChatGPT: each country could mostly choose freely, so they could move fast and learn. For customer-facing parts, however, they set a clear boundary—those were off-limits and handled centrally by the group, because customer conversations on the shared digital platform (George) involve sensitive data across all countries. They found that many retail customers aren’t ready for conversational banking. Their use of the digital banking app is more tool-oriented (transfers, checking balances) than chat-based. So they adopted two AI modes: - **Reactive AI**: the customer initiates the conversation. - **Proactive AI**: the bank pushes conversation prompts to guide customers into trying AI features. Proactive AI delivered huge results, because customers often like AI but don’t know what to ask. They are now balancing both modes to help customers become familiar with AI. ## The Vision for the George App: From Self-Service to Financial Advisor Maurizio thinks it’s uncertain whether banking will become fully conversational. He points to predictions that the Alexa era would make people naturally talk to devices, but in reality he sees people pulling out their phones to check balances on public transport rather than speaking to them. He argues that if the interface is well designed, tapping a few buttons to complete a transfer can be more efficient than a conversation. The bigger challenge today is that only about 20% of the customer base (typically wealthier and older) walk into branches for financial advice. The remaining 80% use the tool functionally, yet they are often the ones who need financial guidance the most. Erste’s vision is to care for customers’ financial health the way doctors care for their physical health—it’s the second most important thing in life (after health). AI can help democratise access to tools and capabilities, serving all customers, not just the wealthy. ## OpenAI’s Role as a Partner Maurizio half-jokes that OpenAI puts him in an awkward position because they set an extremely high experience bar (e.g., ChatGPT being a $23/month consumer product). The AI conversations banks deliver must meet that same standard—customers won’t lower their expectations just because the institution is a bank. So OpenAI needs to help banks reach that bar, acting not just as a technology provider but as a standard setter and co-builder. ## Advice for Regulated Companies Starting Their AI Journey Maurizio emphasises that you have to accept making mistakes. Erste’s AI platform is already on version two, and the first version had to be completely rebuilt; he expects the third version will also need a total overhaul in 18 months. In some organisations this might be seen as waste, but they see it as the only way to get things done. The key is to shorten the cycle (e.g., with coding assistants) and to accept you’ll be wrong a few times before you get it right.