Fintech
GenAI’s “lack of field-tested scenarios” hampers its use in financial products, says German fintech Solaris
Generative AI is playing a significant role in many industries, in some cases driving productivity through the roof.
The financial sector is also attracted by its potential.
Examples included Capital One and JPMorgan Chase leveraging GenAI to strengthen its fraud and suspicious activity detection systems, Goldman Sachs using GenAI to develop internal software and Morgan Stanley introducing a AI tool that helps its financial advisors find data.
But here is Carina Grühser, VP engineering of the German fintech Banking-as-a-Service Solaris, which he recently raised 96 million euros in a financing roundoutlines some of the challenges financial institutions face when implementing GenAI in a financial environment.
Why doesn’t Solaris use ChatGPT?
“Implementing large language models like ChatGPT internally in a regulated organization is a significant undertaking. Care must be taken when using AI, as security and data protection are always top priorities.
“For example, some AI apps may use third-party AI model providers, often data sharing is part of the terms of use. When using any software as part of the Solaris technology stack, we must ensure that they are have clear processes and controls in place to safeguard customer data from unauthorized access, breaches or misuse.
“Additionally, continuous monitoring and validation processes need to be established.
“Instead of implementing tools like ChatGPT, our approach is to first build internal expertise, as AI expertise is scarce, especially in a regulated environment with banking know-how.
“A deep understanding of the business context is critical to identifying application areas that truly add value, especially in generative AI.
As part of this process, one use case identified by our AI task force is SolarisGPT, a chatbot that draws on all the information and knowledge from our Solaris systems to enable our employees to have all the information at their fingertips and be more efficient.
“They have designed the architecture and implemented a proof of concept already successfully. Developing such systems in-house means they have been designed with operation in a regulated environment at their core and gives us greater control over data usage and privacy.”
Why is implementing GenAI in a financial environment a significant undertaking?
“We constantly evaluate the improvements new technologies can bring to increase partner experience and our operational excellence. At the same time, our customers’ and partners’ data must be safe above all else.
“With proven third-party software, including those using regular artificial intelligence or ML, there are many policies, know-how and measures available to ensure we can use them in a robust and regulatory-compliant way.
“However, when it comes to GenAI applications, because the technology is in its infancy, there are much less proven use cases and proof points when working in regulated financial environments.
“This means that each potential opportunity must be thoroughly analyzed from scratch, which results in a significant and often time-consuming process.”
Is GenAI still very far from becoming a Solaris product?
“While we see great potential in using AI and machine learning in certain scenarios, such as risk assessment, fraud detection, personalized services and automated customer services, the broader implementation of Gen AI is still in the exploratory phase.
“The lack of proven, field-tested scenarios for Gen AI in regulated environments means we are not yet at the stage where we expect Gen AI to be used in Solaris products.”