The world of financial services is in a state of rapid evolution, with new trends and technologies emerging at an astonishing pace. As the demand for faster, cheaper and hyper-personalized products continues to grow, fintechs need to balance customer service with managing risk and compliance in a heightened risk environment.
AI and machine learning are powerful tools to meet this growing priority: 91% of financial services companies are already driving critical business outcomes with AI investments, according to the 2023 State of AI in Financial Services report, and 80% of fintechs are leveraging machine learning in particular. Fintechs are deploying AI-powered applications, in combination with new data sources such as open banking, to drive innovation in use cases across the customer journey, from seamless customer onboarding, workflow automation and more effective risk underwriting, to the identification and prevention of fraud, deeper and personalized user engagement and world class customer service.
But no single organization can do it alone, which is why partnerships have become increasingly attractive to fintechs. By teaming up with established innovators like AWS and NVIDIA to build applications for these use cases, fintech startups can find a quicker and more effective way to achieve their digital transformation and deliver deeper, more personalized AI-powered solutions at scale throughout the customer experience. These partnerships unlock a key opportunity to grow revenue and customers in new markets and successfully navigate the digital trends with the complex regulatory landscape.
Enhancing customer service
Traditionally, self-service options like a chatbot or virtual assistant have been governed by predefined scripts. When a customer asks one question, you give one answer, but if they ask enough questions, ultimately the script breaks. There are only so many branches.
Generative AI and large language models will add flexibility to virtual assistants and chatbots, which will no longer be governed by these predefined branches of questions and answers. Fintechs can use generative AI with appropriate conversational guardrails, while giving them the ability to speak to company policies, product requirements and questions about a customer’s personal situation, and they will answer far more questions than we’re accustomed to today.
With the accelerated computing platform that sits within AWS powered by NVIDIA, chatbot assistants like Cleo can do this in near real time, leveraging Amazon SageMaker, a fully managed machine learning service. Cleo uses a casual language style to answer finance and wealth management questions, reducing awkwardness and making the conversation feel natural. Cleo can give users an overview from multiple accounts and show the bigger picture on their spending habits, while using a machine learning classifier to interpret users’ intents and give relevant, personalized responses.
“Fintechs differentiated themselves from traditional financial institutions precisely because they’re able to provide game-changing experiences for their customers,” says Sam Edge, Global Head of FinTech Business Development for Startups and Venture Capital at AWS. “AI and machine learning are key tools for fintechs to continue to delight customers while also managing risk.”
Besides generative chatbots, AI plays an important role there in call centers as well. A large part of an agent’s responsibility has traditionally been note-taking, so they can properly document all that’s happened. Now they can leverage speech-to-text AI to transcribe the call. That transcription can then power other AI features, like recommendation systems, for both customers and the agents on how to get satisfactory resolutions faster in the future.
Fraud detection and identity verification
Companies that are leveraging deep learning today, some of whom have been fighting fraud for decades, are seeing meaningful improvements. With AI solutions, banks and fintechs can look at prior transaction data and what they know about the customer to detect the potential of fraudulent purchases and reduce false positives.
“If you bought a flight ticket to go to San Francisco, would it make sense that you’re suddenly in South Florida paying for dinner?” explains Kevin Levitt, Head of Global Industry Business Development for the Financial Services industry at NVIDIA. “This approach to fraud detection can help more transactions flow more seamlessly, the customer satisfaction goes up, and the banks and fintechs reduce the work on the backend for the investigators who have to pursue the fraudulent actions.”
This also applies to identity verification. The average customer maintains a slew of usernames and passwords, while companies have placed various protections to make sure that people aren’t applying for financial products under aliases or stolen identities. Deep learning tools, such as graph neural networks, are able to map and identify relationships across entities that weren’t previously identifiable or known.
Veriff, an identity verification solution that partners with AWS and NVIDIA, analyzes more than 10,000 variations of government-issued IDs from over 190 countries in 40 different languages via machine learning and through its intelligent decision engine. This empowers fintechs who use Veriff to expand across borders, further building on its decision engine and making it even smarter and more effective over time.
By reducing false positives and more accurately identifying fraud, fintechs can improve their compliance performance and reduce the burden on their auditors. Using deep learning powered by accelerated computing, NVIDIA and AWS customers have seen a 50% reduction in false positives, which translates to a 50% reduction in workload, leading to cost savings as well as better customer experience.
Personalized recommendation
Across each stage of the customer journey, AI can also provide recommendations on optimal next best actions — beginning with recommendations that drive higher conversions. An AI-powered application can analyze data to automatically generate advertisements that are hyper-personalized for a specific audience, leading with the most relevant content for potential customers.
For example, Capital One leveraged state-of-the-art recommendation architecture to improve conversion attribution methods for its personalized online ads. Accurate attribution is essential for this use case: If an algorithm can’t determine which impressions were most likely to motivate a conversion, then it will struggle to serve the most relevant ads to similar audiences, limiting overall conversions. By partnering with AWS and NVIDIA — specifically, by building with the NVIDIA Merlin framework for recommender systems available on AWS — Capital One unlocked a superior attribution method that outperformed alternative models by a wide margin.
Thanks to frameworks like NVIDIA Merlin on AWS, this kind of hyper-personalization capability is now readily available through AI recommendation systems because it’s able to analyze all the data that’s relevant within a bank or fintech about a customer’s situation, and then use that data to give more salient recommendations. An AI-powered application can analyze data to generate not just ad recommendations, but also the right piece of content for customers throughout their journey — whether that’s saving for retirement, or budgeting for a big purchase, or creating a more effective investing strategy. What started with industry buzz around robo advising has now turned into a full suite of recommendations systems that address any sort of need of the customer along their financial journey.
Risk management
Fintechs that operate on AI analysis are able to use more than just a traditional credit report when making an underwriting decision. By looking at alternative data and using deep learning models in their underwriting to assess it, they’re actually able to underwrite or approve creditworthy customers who otherwise would’ve been declined by traditional banks and credit scoring.
Aella Credit is one of the startups on a mission to provide easy access to credit to the underbanked users. Using Amazon Rekognition, the financial technology company is dedicated to providing instant loans to customers in Nigeria through a mobile loan application platform. Rekognition is a deep learning technology solution that enables the addition of image and video analysis to applications, and Aella Credit uses it to enable facial detection and recognition for new customers.
This technology enables more consumers to make financial progress by getting their identity easily verified and receiving access to personal loans. Traditional financial institutions have seen the power of this because they started to leverage deep learning underwriting models in their own lending decisions now.
“It is really a financial empowerment story,” Levitt says. “The deep learning technology enables fintechs to write lower into the credit spectrum, have higher approval rates, and loan more money than a traditional bank could, while maintaining an appropriate risk profile.”
Download NVIDIA’s State of AI in Financial Services report to learn more about key AI trends, challenges and opportunities.