Lewitt referenced an AI in banking survey currently being fielded by NVIDIA.
“The early results are saying that just shy of two-thirds, about 57 percent, are developing their own [AI capabilities], 22 percent are co-developing with partners and just nine percent are in market actually buying the solution outright, with the remainder not currently active in any of those three right now,” he said.
“That gives you the view of where the banks are. Certainly, a lot of our customers skew more towards the big banks than they do the long tail of community banks, credit unions … A lot of the bigger banks do have the resources to build and develop on their own.”
Despite large budgets for internal development, Levitt does see banks adopting hybrid models in areas that are no longer purely on premise.
“There’s a lot more appetite to move to the cloud, but also appetite to maintain a dual strategy when it comes to flexibility of where data is stored and managed, or where your AI operations are taking place,” he said.
According to panellist Shnay Chohan, Cora AI experience lead at Natwest Group, banks can benefit from a hybrid approach, but should be cautious of developing a third party dependency.
“It is a combination of both really. If you start straight away but you don’t have the skillset to deliver [AI] to a high standard, then you’re going to need help from third parties to be able to deliver that,” he said.
“If you start by having some help from a partner but with the idea of that being handed over fully from external to then internal, I think that makes more sense. Then you’re able to build that skillset and actually a lot of these partners will help you get to that position – it’s not necessarily in their best interest to be depended upon for so long. Doing a bit of a blend is quite a good approach and has worked for us in the past.”
Also on the panel was Kate Rosenshine, head of Azure Cloud Solution Architecture at Microsoft UK, who said internal governance plays a large role for banks when expanding their AI capabilities.
“A lot of financial institutions already have an early form of AI in place, just look at the quants. That’s a predecessor to all of this, so it really depends on the approach the organisation takes,” she said.
“I see so much potential and sometimes it’s just a matter of how do you rebrand a certain division and just tweak it a bit to do something new … and then when you do want to go for those bigger, stickier projects, then you have a whole host of other considerations. It’s often better to start small, have some wins and then scale out.”
Despite banks expressing an increased interest in AI – its business value in banking is projected to reach $300bn by 2030, according to an IHS Markit report – pure machine learning is not necessarily the goal in front house development, according to Chohan.
“We don’t have machine learning models that run on their own; they are assisted meaning humans are looking at them and I don’t think there’s necessarily an appetite to do pure machine learning customer facing at the moment, but assisted machine learning models absolutely.”