ChatGPT and other generative AI applications are becoming increasingly important

For some time now, terms such as chatbots, voicebots, conversational AI, generative AI, ChatGPT and large language models (LLMs) have been gaining in popularity and are increasingly being taken up by the media and politicians. Companies are also following developments closely and discussing them. However, a closer look at the market reveals that many Swiss financial companies are currently still hesitant to use technologies such as chatbots or voicebots, generative artificial intelligence (AI) or large language models (LLMs) in their customer and internal communication. As part of the latest IFZ Conversational and Generative AI in Finance study, together with my colleagues Nils Hafner and Florian Schreiber, as well as the partners Adnovum and Spitch, I researched more about the possible applications and obstacles of Conversational AI and Generative AI solutions in banking and insurance.

What is the IFZ study Conversational and Generative AI in Finance about?

For the latest IFZ Conversational and Generative AI in Finance study, 47 employees in senior positions at Swiss financial companies were surveyed on the potential uses of and obstacles to chatbots, voicebots, generative AI and ChatGPT.

Specifically, the following research questions were formulated together with the participating partners and answered as part of the study:

• What hurdles do you encounter when introducing Conversational AI solutions?
• What does the ideal combination of man and machine look like?
• How should chat and voicebots be combined with communication channels operated by humans?
• What (new) opportunities arise from the use of Generative AI (e.g. ChatGPT, GPT-4 and similar technologies)?
• How is the acceptance of Conversational AI solutions based on Generative AI?
• What concerns does your company have today and in the future with regard to the use of Generative AI? How can these concerns be reduced?
• What should be considered with regard to the design of Conversational AI solutions (e.g. tonality, customer approach, naming, etc.)? Do best practices exist in this regard?
• What best practices exist in terms of integrating digital assistants into the overall service, sales and marketing culture?

How do the results differ between banks and insurance companies?

The results generally show only minor differences between the banking and insurance sectors. Most of the results are equally valid for banks and insurance companies. There are only a few differences in the use cases. This can be explained by the fact that banks and insurance companies have different customer concerns. While reporting a claim is a frequent reason for contact for insurance customers, bank customers are more likely to inquire about their account balance.

How widespread are chatbots and voicebots at Swiss financial companies?

The results of the survey clearly show that only just over half of the companies surveyed already use chat and voicebots. In general, chatbots are much more widespread than voicebots. However, there are also companies that use both chatbots and voicebots. Some companies already have several chatbots in use at the same time.

Where are chatbots and voicebots used?

The study shows that the majority of companies that already use conversational AI use this technology in customer service, marketing or the IT helpdesk. The following three main objectives are being pursued: Increasing customer satisfaction, increasing efficiency and reducing costs. Chatbots are primarily used for product advice, answering frequent queries or forwarding to a suitable employee. Voicebots, on the other hand, are mainly used for forwarding to a suitable employee.

How are chatbots and voicebots connected to employees?

If the chatbot or voicebot does not know a suitable answer, the dialog is either forwarded directly to a responsible employee or, in the case of a chatbot, a link to the contact form is displayed. According to the employees surveyed, there are also companies that have not yet considered a direct link between bot and human employee. In these cases, the conversation between customer and bot ends as soon as the bot is unable to answer a question. Customers are therefore “forced” to seek contact independently via another channel. While most chatbots have a name and often even an avatar, voicebots usually lack a name. Since there is no visual communication with voicebots, an avatar is not required. Customers typically reach the chatbots via the websites of the banks and insurance companies surveyed, usually via the chat window on the right-hand side. If the chatbot project is still in the test phase, the bot may only be accessible via a link on the company’s contact page.

What are the most important requirements for AI projects?

The study also provides clear answers to stumbling blocks and challenges when implementing a Conversational AI project. Sufficient human and financial resources are among the most important prerequisites for the successful implementation of such a project. The issue of data protection should not be underestimated either. All open questions must first be answered and all data protection risks must have been eliminated.

How widespread is Generative AI and ChatGPT at Swiss banks and insurance companies?

Generative AI, including ChatGPT and similar technologies, is becoming increasingly important for Swiss financial companies. Almost all companies state that the topic of generative AI is being discussed internally and initial considerations are being made regarding use cases and possible applications. In addition to improving the customer experience, the focus is primarily on increasing internal productivity. This increased productivity should be particularly noticeable in the form of cost savings. One of the most common applications for Generative AI appears to be the automated answering of emails. The improvement of existing chat services and the automated creation of content for blogs and social media also play an important role. However, as with the use of conversational AI, there are also concerns about data security and data protection when using generative AI.

More insights on Conversational and Generative AI in Finance?

In addition to these results, we have also derived successful use cases for the application of Conversational and Generative AI. You can download these and other findings from the IFZ study Conversational and Generative AI in Finance.

And as always, please feel free to contact me if you have any further questions about the study. Just send me a WhatsApp message or an e-mail.