Industries such as retail and tourism have long relied on AI programs such as ChatGPT to support them in their daily work. The new "IFZ Generative AI in Finance Study" by Lucerne University of Applied Sciences and Arts shows for the first time what banks and insurance companies can learn from other industries when using such generative AI software. The researchers have developed two useful tools to support this: the "Gen AI Scale" and the "AI Process Compass".
Artificial intelligence is becoming increasingly important for many companies. Companies in the healthcare, tourism and retail sectors, for example, are using AI programs such as ChatGPT. These generate completely new content with the help of machine learning and language models and are therefore referred to as ’generative AI’. In the new ’IFZ Generative AI in Finance Study’, researchers at Lucerne University of Applied Sciences and Arts (HSLU) identified successful applications of generative AI from other industries and examined their applicability in the financial sector. The aim was to provide banks and insurance companies with recommendations for useful generative AI applications and to provide them with two helpful tools.
As a first step, the researchers conducted interviews with senior employees from the areas of customer service, information services, compliance, marketing and communication. This enabled them to find out in which areas AI is already being used. The results of the surveys were then mirrored with experts from the financial sector in order to assess the transferability of generative AI applications to the financial industry.
’Gen AI scale’ helps with orientation
The researchers introduced the ’Gen AI Scale’ to classify Generative AI applications for banks and insurance companies. ’It aims to identify opportunities, risks and upcoming fundamental decisions in the company,’ says study author Nils Hafner.
Figure 1: The ’Gen AI Scale’ helps to categorize AI applications. (Click to enlarge)
On the ’Gen AI scale’, AI applications can be categorized between two poles: One pole comprises applications that are operated directly by customers. The other pole includes processes that run entirely within the company and are sometimes not even visible to employees. Due to the lack of visibility, this is referred to as dark processing in the insurance industry, for example. In between are applications that support all or individual employees.
The use of AI is seen as particularly useful in the ’creation of marketing texts using generative AI’ and in the ’creation of product descriptions’. ’This is easy to understand against the backdrop of increasingly complex financial products and the associated global increase in regulatory compliance,’ says Nils Hafner. This is where Generative AI shows its strengths in terms of the complex adaptation of texts in different languages and legislative areas.
The researchers at HSLU have developed another useful tool in addition to the ’Gen AI Scale’: The ’AI Process Compass’. This guide shows the up to 15 steps that companies should go through for the long-term and successful introduction of Generative AI applications: from the analysis of existing customer dialogs to a feasibility check and technical implementation through to rollout.
"The AI Process Compass serves as a checklist for the structured introduction of applications," says co-author of the study Sophie Hundertmark. "We found that purely in-house applications of generative AI require far fewer steps before they can be introduced." For companies, this creates clarity about the effort involved in the introduction.
For the ’Generative AI in Finance Study 2024’, the Institute of Financial Services Zug IFZ at Lucerne University of Applied Sciences and Arts, together with partners, investigated how other sectors such as retail, healthcare, tourism and many others are already using generative AI and evaluated the transferability of these use cases to Swiss financial companies.
