Agentic AI in Banking: How bunq Automates 97% of Support via Amazon Bedrock
Coverage of aws-ml-blog
In a newly released case study, the AWS Machine Learning Blog outlines how bunq, a leading European neobank, utilized Amazon Bedrock to transform its AI assistant, Finn, into a fully agentic system capable of resolving the vast majority of user inquiries without human intervention.
Scaling customer support in the financial sector presents a unique set of challenges. Banks must navigate strict regulatory environments while meeting consumer expectations for instant, 24/7 assistance across multiple languages and time zones. Traditional support models often struggle to balance cost-efficiency with quality, leading to long wait times or impersonal, scripted interactions. In a recent post, the aws-ml-blog explores how bunq has addressed these friction points by deploying Agentic AI.
The analysis focuses on the evolution of bunq’s in-house AI assistant, Finn. By migrating to Amazon Bedrock, bunq reportedly upgraded Finn from a standard conversational interface to an "agentic" system. Unlike traditional chatbots that simply retrieve information (RAG), agentic systems are designed to autonomously plan and execute multi-step workflows. This capability allows the AI to not only answer questions but also perform actions on behalf of the user, such as processing transactions, analyzing spending habits, or resolving account issues directly.
According to the publication, this architectural shift has allowed bunq to automate approximately 97% of its user support interactions. This figure is significant as it demonstrates that Generative AI has moved beyond experimental pilots into critical, high-volume production workflows within regulated industries. The system handles the complexities of multilingual support autonomously, ensuring consistent service levels regardless of the user's location or native language.
While the post focuses on the high-level success metrics rather than the granular engineering specifics-such as the specific foundation models selected or the precise guardrails implemented for financial compliance-the signal is clear: Agentic AI is proving to be a viable solution for operational scaling in fintech. For engineering leaders and CTOs, this case study serves as a proof point for the ROI of integrating managed AI services like Amazon Bedrock into core business logic.
We recommend reading the full article to understand the scope of bunq’s implementation and the potential for agentic workflows to redefine customer service standards.
Read the full post at the AWS Machine Learning Blog
Key Takeaways
- High-Volume Automation: bunq reports that its AI assistant, Finn, now handles 97% of all customer support inquiries autonomously.
- Shift to Agentic AI: The implementation moves beyond simple text generation to agentic workflows, enabling the AI to make decisions and execute banking tasks.
- Enterprise Scalability: The use of Amazon Bedrock allowed the neobank to deploy these capabilities globally, addressing multilingual and cross-time-zone challenges.
- Operational Efficiency: The case study illustrates how AI agents can resolve the traditional bottleneck of scaling human support teams in the banking sector.