ING: From Chatbots to Action-Taking Agents

In this conversation, Marco Li Mandri of ING explains how the bank is deliberately evolving customer-facing AI from simple information tools into action-oriented digital agents. While chatbots have already become a familiar part of daily banking, Li Mandri is clear that answering questions is only the first step. The real transformation begins when AI can do things, not just explain them.

Until now, ING’s chatbots have focused on handling frequently asked questions — balance queries, transaction explanations, and basic support issues that reduce pressure on contact centres. These tools have delivered clear efficiency gains and improved customer experience by offering fast, always-available assistance. But Li Mandri describes this phase as foundational rather than transformative.

The next step is execution. ING is now working to enable agents — particularly those powered by generative AI — to carry out concrete actions on behalf of customers. Rather than redirecting users to forms, menus, or human agents, future chat interactions will resolve issues end-to-end. A customer won’t just ask about an invoice; the agent will retrieve it. They won’t just ask how to fix a problem; the agent will fix it.

This shift represents a fundamental change in how customers interact with their bank. Conversations move from informational to transactional, reducing friction and cognitive load. Instead of navigating complex digital journeys, customers can simply describe what they want in natural language. The agent then orchestrates the necessary steps behind the scenes — securely, compliantly, and within clearly defined boundaries.

Li Mandri is careful to emphasise that this evolution does not remove humans from the loop. Execution-capable agents must operate within strict controls, with clear escalation paths when tasks fall outside predefined authority. ING’s approach is incremental: start with low-risk, high-frequency use cases in daily banking, learn how customers interact with agents, and then expand capability over time.

Invoices are a prime example. They are common, time-consuming, and frustrating for customers — yet relatively well-structured from a data perspective. By starting here, ING can deliver immediate value while refining governance, security, and monitoring processes. Each successful execution builds confidence, both internally and externally.

This progression reflects ING’s broader AI philosophy: use technology to remove unnecessary effort from everyday life, not to impress with novelty. Chatbots answering questions were useful. Agents that resolve problems directly are transformative.

As Li Mandri frames it, the future of digital banking is conversational, contextual, and action-driven — where customers speak, and their bank responds not with instructions, but with results.