Deutsche Bank has doubled down on its use of artificial intelligence. The Frankfurt-based bank announced on Wednesday a multiyear partnership with a U.S. tech company, Nvidia, to accelerate and improve artificial intelligence and machine learning throughout the bank. The companies declined to share how much the bank is spending on the initiative.
The two companies have already been working together for 18 months on AI projects that include risk models, an HR avatar and a metaverse presence, they said.
Other large banks have also been stepping up their use of AI in recent days and even giving executives
"There is an acceleration of AI in banking because financial institutions are keen to cut costs and they also deal with things such as financial transactions that are easily automated," said Darrell West, vice president and director of governance studies at the Brookings Institution. "Algorithms can analyze money movements and spot fraud, inefficiencies, or ineffectiveness."
How Deutsche is deploying AI
Deutsche Bank has already been using AI in what Perez refers to as traditional use cases, like know-your- customer compliance, sanctions compliance, payment categorizations and dynamic pricing.
"There are lots of use cases that we're using right now today for AI," said Gil Perez, chief innovation officer and head of cloud and innovation network at Deutsche Bank. "Including even within security, within trade surveillance, etc. We believe that there's going to be significant improvements on all of these, plus many more use cases that we haven't yet explored, with more modern and powerful AI and machine learning capabilities and a lot more data."
One thing Deutsche aims to do through the Nvidia partnership is make its risk models run faster. The bank has begun using Nvidia's graphics processing units and high-performance computing stack for its risk models, and has already seen a tenfold improvement, according to Perez.
The second is an HR avatar that answers employee questions. "Our focus has been on HR and internal usage," Perez said. "We are also considering exposing it to external HR prospects with regards to potential hiring."
The third is an avatar for the metaverse.
"A lot of people have spent the last couple of years in all kinds of immersive games," Perez said. "We believe that direction is only going to continue, and that kind of an experience is going to be expected in the workplace as well."
Perez noted that it's not yet clear which platform, if any, will be successful in the metaverse.
"There's various platforms that are vying to be the right ones," he said. "When we did our analysis, we decided that the common element across all of those platforms is actually the avatars."
Deutsche Bank would like to start conversations with people through avatars in the metaverse and then, at the right time, do a handoff to a human subject matter expert.
"We do believe that avatars are a natural evolution of self-service," Perez said. "Just like today, you use your application to check your balance and to do all kinds of routine banking services, we believe that in the future, you will want to engage with an avatar to do some of those routine functions, and then at some point, move over to a person when it's needed."
Deutsche Bank is building consumer protections like data privacy into its AI frameworks, Perez said, "to ensure that not only are the models nonbiased at the beginning, but to prevent any kind of drift." The bank is working with about 46 different regulators, including in the U.S. and in Germany, so that it will be able to roll out AI-based features globally.
Nvidia's AI hardware and software
The bank is using Nvidia's AI Enterprise and Omniverse Enterprise technology, both on premise and in the cloud (the latter through Nvidia's long-standing relationship with Google).
"We also see the potential of AI to transform the financial services industry," said Manuvir Das, vice president of enterprise computing at Nvidia, of Santa Clara, California. "We are hoping as we go forward, not only to help Deutsche Bank succeed, but to learn a lot about the use cases in the financial services industry so that the industry as a whole can benefit from the advancements to the platforms that we make."
GPUs, also referred to as graphics cards, are often used for AI because they can run about 10 times faster than standard computer chips.
Nvidia has been developing software to accompany its GPUs, Das said.
"What we quickly learned at Nvidia was that in order to benefit from this accelerated computing, the workloads and use cases and the tools and frameworks had to be adapted and retargeted to GPUs, and this is what we undertook and did," he said. "We built the software that enables those use cases to run well, whether it would be in visual AI like video analytics or speech AI for virtual assistants, to convert audio to text or text to speech or recommend systems that show people what to buy next or what to read next."
Deutsche Bank has been using several of these software development toolkits for some time now, he said.
The latest version of AI Enterprise, 3.0, comes with software development tools for speech AI, vision AI, cybersecurity and document recognition, Das said. It also comes with some pre-trained models for things like comparing a person's current behavior against normal patterns for security and fraud purposes. It works with cloud technology from Oracle, HPE, Red Hat and VMware as well as Microsoft Azure and Google Cloud.
Deutsche Bank already has an AI center of excellence based in Dublin that will help implement AI in other departments.
"We are engaged with each one of the divisions with regards to their North Star, where they want to take their business in the next couple of years," Perez said. "This is a commitment on our side to actually make AI an integral part of the way we function across the bank."
