ABN AMRO to pilot Wonderful's AI agents on multi-step banking workflows

ABN AMRO is preparing a pilot with Wonderful to test how AI agents can handle multi-step workflows inside the bank, with clear boundaries, human oversight and measurable results. Yorick Naeff, the bank's Head of Innovation, announced the pilot on LinkedIn on 30 September.

"The exciting part of agentic AI? Getting work done. The hard part? Making it work reliably inside a bank," Naeff wrote.

What is the pilot meant to prove?

Naeff set out three questions the pilot is meant to answer:

• "Can agents reduce repetitive work and genuinely help our colleagues?"
• "Can we make their actions traceable and keep people in control?"
• "What would it take to scale responsibly if the results are strong?"

The split of work is straightforward. "Wonderful brings its platform and expertise in AI agents. We bring our banking expertise and the security, risk and compliance requirements that any solution must meet," he wrote.

Neither company has said which workflows will be tested, when the pilot starts or how long it will run. Naeff described the next step as "test, learn and earn the right to scale."

Who is Wonderful?

Amsterdam-based Wonderful describes itself as the AI OS for the enterprise. It announced a $550 million Series C at a $5 billion valuation on 2 September, and lists banking among the industries it targets.

ABN AMRO is no stranger to conversational AI. In 2024 it moved its chatbots onto Microsoft Copilot Studio. The Wonderful pilot moves the focus from answering customers to agents carrying out work across multiple steps.

This article was prepared with the help of EDDIE, the AI editor-in-chief at Conversational AI News, and MARVIN, our AI research assistant.