AI In Banking: Proven Best Practices From Industry Experts
Big tech will not like bankings' tech agnostic approach to GenAI.
This is my daily post. I write daily but send my newsletter to your email only on Sundays. Go HERE to see my past newsletters.
HAND-CURATED FOR YOU
Euromoney interviewed dozens of AI experts and created this insightful and fun-to-read “best practices playbook” for AI in banking. It should be on everyone’s reading list!
This report is full of “pearls of wisdom” from industry practitioners at the major banks, conveniently dispersed throughout the report.
But of all the pearls of wisdom about LLMs, the biggest one will stick in the throat of Big Tech.
The report's interviews repeatedly state that banks should remain tech-agnostic. The LLMs are developing so fast that committing to a model or company makes no sense.
Try telling that to OpenAI or Google!
While the journalists at Euromoney have done a great job assembling the thoughts of the practitioners they interviewed, they also made a sobering observation about banks’ claims:
“Inevitably, there is a gap between claims of radical innovation and reality. This is particularly true in agentic AI, given its still hazy definition.”
So when you read bold typeface figures in the report NatWest claiming a 150% increase in customer satisfaction or that 50% of Morgan Stanley’s staff is using AI, take them with a grain of salt.
👉 Nine steps to implement AI in Banks:
01 | Adopt a decisive but flexible approach to AI development
Adopt an LLM-agnostic approach
Balance agility with decisiveness
Use traditional AI where appropriate
Focus on end goals
02 | Strengthen a central AI function plugged into businesses
Build a plugged-in central AI function
Formulate internal AI standards
Keep a lid on costs
Invest more in broader use cases
03 | Upskill across the organisation, not just in AI research
Upskill across the firm
Have different levels of AI training
Launch AI mentors and evangelists
Attract scientific talent
04 | Develop processes for selecting third-party AI models
Develop a framework for model selection
Keep your specific needs in mind
Assess performance, security, pricing
Ignore advantages likely to be temporary
05 | Build platforms to boost model flexibility and security
Favour strategic relationships
Maintain pricing pressure
Stay agile and switch when needed
Invest in a flexible LLM platform layer
06 | Consider smaller models which meet your needs
Look beyond the big LLM players
Customisation is key
Be thoughtful of your advantages
Adapt to your jurisdiction
07 | Start with low risk but impactful use cases
Don’t overlook simple use cases
Start with internal deployment
Develop skills to improve AI output
Allow users to check at the source
08 | Ease into customer-facing opportunities
Expect the unexpected
Test and learn, repeat
Customer-facing deployment takes its time
Humans stay behind the scenes for now
09 | Build trust to succeed in agentic banking
Recognise the threat of agentic AI
Treat agentic AI as a growth opportunity
Experiment with less material agent actions
Build a framework around checking requests
If you know someone who would like this newsletter, please share it with them and help grow our Asia, CBDC, and AI aficionados community!
