I attended the Evident AI Symposium yesterday. Several roundtable discussions highlighted the current state of AI in the financial sector. While there’s huge enthusiasm around AI, banks remain hesitant to integrate generative AI into their processes. I went along to work out what’s holding the sector back – is it the regulators or a lack of vision, or something else?
According to Tony Kim, head of tech sector fundamental equities at BlackRock, technology has recently gone through an epoch moment: “LLMs are like an AD/BC moment for tech investment. It’s the biggest driver of value for tech companies today.”
The public markets agree - NVIDIA recently became the world’s most valuable company, and the five biggest companies in the world by market cap today are all technology companies – NVIDIA, Microsoft, Apple, Google Amazon. There’s never been a stronger signal of the disruptive potential of technology.
So why aren’t banks in production with AI? And more specifically GenAI? Listening to yesterday’s Evident AI Symposium, it was obvious why there’s reluctance – fear of the regulators, fear of customer backlash, a lack of confidence in their abilities to overcome challenge. But it also felt like there was a real lack of vision. Just as fintechs lead the way with open banking, digital & embedded finance, and payments, the scene is set for them to lead with AI too.
“What I really want is AI to answer emails.”
Head of AI Research at JP Morgan, Manuela Veloso, said that financial services only works with humans. She said “when we’re thinking about the stack of AI, humans are the final layer. There’s a component humans bring. Past knowledge, correlation. Humans are always needed in financial services.” The challenge, she said, is that humans are “victims of Microsoft Office… What I really want is AI to answer emails.”
We can all relate. Most of us can feel like we’re stuck in a permanent cycle of spreadsheets, documents, PowerPoints and emails. However, this line of thinking is what’s preventing a serious rollout of AI in banking. We’re thinking about tasks where AI could help with rather than reimagining processes. The world probably already has enough chatbots.
“The challenge is it’s not deterministic”
If there was one thing holding back generative AI, it’s the fear of hallucinations. The word of the day at the conference was “accuracy.”
Ian Glasner, Group Head of Innovation, Ventures, and Digital Partnerships at HSBC, gave a summation of banks’ reluctance to change: “Unlike traditional AI, which is deterministic and uses maths, generative AI uses transformers and sometimes gets it wrong, as we've seen in the press many times.”
Ian gave three challenges to adopting GenAI:
Ian’s conservative approach was typical of the discussions. “We’re still in the early stages” was the refrain of the day.
But I’d argue do humans always meet the requirements of these three challenges? If we stay hung up needing to prove 1+1=2, we’ll never move forwards. It’s about creating a management structure with the checks and balances for AI to flourish – just as we do with human employees.
It’s time to lay the foundations and break the silos.
Banks are of course conservative for a reason. Regulators push them to be cautious, and as a result innovation doesn’t come easily. But there’s certainly a strong argument that they simply aren’t thinking big enough yet.
It needs an outsider’s perspective to set the record straight.
SambaNova’s CEO Rodrigo Liang argued that banks need to lay the foundations of AI because data, which is their most valuable asset, is under leveraged and siloed between teams.
Rodrigo said: “Accuracy is the most important but the question is it worth it. But it’s not a linear curve. You’ve got to spend $10 to get $1 of value, but at some point you’ll spend $100 and get $10,000 of value. We’re not there yet but that’s the journey.”
Conclusion
The Evident AI Symposium underscored the significant challenges and opportunities for AI in the banking sector – and despite the enthusiasm for AI's potential, banks remain hesitant due to regulatory fears, potential customer backlash, and, frankly, a lack of vision.
There’s a need for a foundational shift in thinking, moving beyond task-specific AI applications to reimagining entire processes.
The journey ahead involves leveraging data effectively and addressing accuracy concerns, but the time to start is now. By laying the groundwork today, banks can unlock immense value from AI in the future.