By: Joy Macknight

Artificial intelligence (AI) was the main hot topic at Money 20/20 in Amsterdam both in and outside the conference sessions, as well as being the subject of numerous company announcements released at the show. For example, Experian launched its Agent Operating System to help financial services organisations move beyond AI experimentation and safely scale agentic AI.

Experian’s announcement is timely as both banks and fintechs are grappling with adopting and scaling AI in a highly regulated environment. Despite investing heavily into AI pilots, the vast majority never reach production. According to a 2025 Boston Consulting Group report, only 25% of banks are successfully using AI to gain competitive advantage. “The other 75% remain stuck in siloed pilots and proofs of concept, risking irrelevance as digital-first competitors accelerate ahead,” the report stated.

A Money 20/20 panel entitled ‘What will AI-native financial services look like?’, moderated by TTP, discussed the challenges banks and fintechs face when shifting to meaningful implementation.

Banks and fintechs face different issues when adopting AI, according to Pinar Ozcan, Professor of Entrepreneurship and Innovation at Saïd Business School, University of Oxford, and Academic Director at the Oxford Future of Finance and Technology Initiative.

“Large traditional financial institutions typically have siloed data that isn’t AI ready. Plus their compliance mindset means that they look at data from a different cultural viewpoint. For example, one institution called data ‘biohazardous waste’, effectively something they put away and don’t look at,” said Ozcan. “Fintechs, on the other hand, are digital natives and understand AI, but they don’t have the data. They need to either work with a large institution or go straight to the customer to access data.”

Ryan O’Holleran, Head of Enterprise and Startup Sales at US-headquartered AI company Anthropic, which works with the likes of Goldman Sachs, J.P. Morgan and Revolut, argued that one of the biggest challenges for any organisation is creating a robust AI strategy.

“A common trap is companies or teams that deploy AI once and walk away. This isn’t a technology that you can deploy into a chatbot, without checking on and supporting it. Models are being released at such a pace and frequency that organisations should be constantly testing and implementing emerging models, or they will fall behind,” he said.

Both panellists agreed that regulation can act as an enabler but also hinder AI adoption in the financial services industry, which deals with sensitive data, such as Personally Identifiable Information and Know Your Customer (KYC) data.

“Organisations are looking for guidance from regulators, for example to understand if they can completely automate their KYC processes,” said O’Holleran. “While progress has been made with the EU AI Act and some other regulations, there needs to be more guardrails in place so organisations can be confident in deploying AI in high return on investment use cases.”

Ozcan highlighted the positive impact of open banking and open finance regulations, which are enabling new entrants to access customer data. “Consumers can choose to share their financial data with licensed third parties,” she explained. “However, implementation of these regulations is still lagging and consumers don’t understand what it means, so much more education is needed.”

In addition, AI governance and ethical usage guardrails need to be put in place to remove biases embedded in historical data, such as gender and racial discrimination. Ozcan warned that data cleansing is critical for successful deployment, adding that AI can actually help with this.

Importantly, AI models shouldn’t be trained on sensitive data, added O’Holleran. “It’s not just ensuring that organisations have access to the data, but ensuring that data is utilised in an appropriate way,” he said.

In terms of what AI-native financial services will look like in future, O’Holleran outlined three pillars to AI deployment: internal productivity, such as using Claude Code for work tools; internal APIs use cases, for example automated KYC; and external AI-focused tooling, such as virtual CFOs.

“We’re seeing a lot of adoption in the financial services industry within pillar one, which are table stakes today, while opportunities are beginning to emerge in pillar two and three,” he reported.