Fintechs – pioneering use of generative AI in banking and finance

By Urban Roth, IBM Innovation Studio leader, Stockholm

The race is on for leaders to make meaningful use of generative AI and its transformational power for their organisations.

Current state of the banking industry – approaches, perceived opportunities and risks

In a recent IBM survey of 600 banking executives, 2024 Global Outlook for Financial Services, the tactical approach of exploring singular use-cases currently dominated, with 8 out of 10 adhering to that approach. The business areas perceived to hold the greatest potential were risk and compliance reporting, shortly followed by client engagement. However, over 50% of executives hesitate to implement solutions in risk and compliance, likely due to their concerns around governance and AI-associated risks. Top concerns include cybersecurity vulnerabilities, legal liabilities, and accuracy and bias assessment. These hesitations and perceived risks can help explain why digital assistants are the most common starting point.

How can we inspire wider applications of AI, while dealing with the risks that come along?

Insights from IBM’s 2024 Global Outlook for Financial Services

Industry pioneers showing the way!

In an effort to do just that, inspire the industry by learning while doing, we decided to engage with pioneers of the industry, fintechs. We designed a program IBM fintechx where IBM specialists would validate use-case ideas and co-create prototypes together with a group of fintechs. Then sharing the outcomes to the industry at a Demo Day. Thus contributing to their ventures and existing services, while learning and inspiring the industry. To help us select the most interesting ideas and help spread good results, we gathered a group of industry experts from banking, investment and industry organisations (Handelsbanken, Wellstreet, Ålandsbanken, Resurs, Crosskey and Findec). Seven fintech companies where ultimately invited to join IBM fintechx: Edger Finance, Esgaia, Flowpay, Oxide.AI, TrustAnchorGroup, Asteria and Finterai.

So, what were the applications developed with the fintechs?

IBM fintechx kick-off days, 14-15 November 2023

Applications from fintechx collaborators

Five of the fintechx project’s outcomes are published and briefly summarized below. A full report on our aggregated learnings and identified design patterns is available here:

Report: Learnings from IBM fintechx

Fintechx applications of generative AI:

1. Edger Finance is on a mission to create EU financial market information equivalent to US’s EDGAR by gathering and aggregating all stock market information. Their prototype focused on automating company quarterly report data extracts and summarizing the main points from 30 page reports. Here the team had to work with accuracy of data when summarizing. Solutions rely on IBM watsonx.ai and watsonx Assistant. Read more, Case Study.
   
   
2. Flowpay began its journey with a lot of data and one idea: provide simple, online, operational financing for small and medium-sized enterprises (SMEs). To enter a new market, their cashflow based risk scoring model needed an initial dataset to interpret and categorize PSD2 transactions from that country. A prototype was developed using a large language model to generate that dataset automatically from publicly available data. Flowpay’s underwriting process was also targeted with a prototype generating summaries of an applicant’s financial health and associated risk. The summaries were in non-technical format and based directly from PSD2 transaction data from the applicant’s bank accounts. Read more, Case Study
   
   
3. Oxide AI want to empower all of us to find exciting investment opportunites by creating a social media like investment feed to users. The feed is created by AI agents, trained in quantative and qualitative analysis, crawling and analyzing stock markets to identify non-obvious opportunities. In their new solution prototype, they moved from GPT-4 to Llama 2 model on watsonx.ai. A significant step towards using open models, fit for enterprise grade production. They achieved 37% faster average sequential response times while maintaining quality levels. Read more, Case Study.
   
   
4. Trust Anchor Group helps financial institutions and businesses monitor or invest in real-world assets by providing asset digitalization technology. They built a dynamic asset valuation service using watsonx.ai that can calculate asset values in real time. It’s done by querying unstructured and structured data sources. It also enables business users to ask questions about their assets and deep dive into the valuation findings. They too worked on ensuring data summarization accuracy. Read more, Case Study.
   
   
5. Asteria has an ambition to provide SMEs with access to powerful cash management and a higher degree of resilience in their business operations. They provide their services through collaborations with established banks. Asteria created an online advisor capable of summarizing SME’s financial health, answering follow-up questions and recommend financial products from the bank relevant to the SME’s current needs. The advisor has potential of further strengthening SME resilience while saving time for the bank. Read more, Case Study.

So, what were the common traits across projects and what can we learn?

Common patterns provide disruptive opportunities beyond client engagements

From the projects we conclude that the technology brings disruptive opportunities within and beyond the area of client engagements, validating the potential that surveyed industry executives saw. It comes with repeatable design patterns applicable across financial business areas and brings an outcome accuracy risk that needs to be managed with competence and care.
The range of applications all derive from the gen AI capabilities of:
– getting useful information from unstructured data
– transforming structured data and information into knowledge and insights
– creating natural language user experiences that deliver value to users

We think the time is here for organisations to start looking at which applications could contribute to their goals and include fintech collaborations in the mix.
If you are interested in having a dialogue about how to create this approach, or in any of the fintechs we collaborated with, you are welcome to reach out to us.

IBM Innovation Studio’s mission is to inspire meaningful use of technology and define a path forward for organisations considering new initiatives. Our work is done prior to commercial agreements are put in place.

Urban Roth
Studio leader, IBM Innovation Studio