How banks are deploying AI: best practices playbook – Euromoney report.

The latest Euromoney report on AI in banking offer an unrivalled deep look into how individual banking leaders are thinking about AI and how they are acting on it. How are they leveraging the technology? And which are the most successful and forward-looking initiatives in gen and agentic AI?

Euromoney held more than 30 in-depth conversations with those in charge of implementing gen and agentic AI at top global banks, and in many tech-leading national banks. They also spoke to banking-focused AI professionals at LLM vendors, and smaller AI-focused fintech firms.

„AI strategy in banking isn’t about centralising or decentralising. It’s about knowing when to do both. Also, the hardest part isn’t building AI tools. It’s deciding which ones to scale.” – said Arjun Vir Singh, Partner and Global Head of Fintech @ Arthur D. Little.

Here are his key takeaways:

🔶 Banks need a central AI team while letting each business unit use AI independently. This balance enables consistent governance while ensuring solutions address real operational problems.

🔶 Early internal AI applications are delivering measurable wins. They drive 15-30% productivity gains for developers and 70% faster customer service response times.

🔶 The most forward-thinking banks avoid vendor lock-in. They have built proprietary platforms to swap different AI models as technology evolves.

🔶 Customer-facing AI requires rigorous testing but can soar satisfaction. Some banks report 150% improvements when implementing carefully controlled gen-AI chatbots.

🔶 Technical flexibility is essential. Banks must be able to quickly abandon approaches that aren’t working while maintaining their strategic vision.

🔶 Some AI agents can independently handle dispute resolution and service inquiries. Still, banks are wisely putting guardrails around these capabilities.

🔶 Broad AI literacy across the organisation matters more than specialised expertise. Banks creating communities of data professionals are seeing faster adoption and better use cases.