ChatAI with ChatGPT – what already works?

Put simply, at its core ChatGPT is a chatbot – an application that uses artificial intelligence (AI) to communicate with people in natural language. Users can ask questions, which the chatbot answers in natural language. It doesn’t sound very spectacular at first, but it takes on greater significance when you realize how rapidly AI technology has developed in recent years – and how many areas it is already being used in: language processing, data analysis, and the generation of images.

ChatAI – also known as ConversationalAI – is already being used in the finance industry, particularly to optimize customer service. With the help of intelligent chatbots, financial service providers are automating their communication. For prospective customers or customers in self-service portals or digital application lines, it is usually not obvious whether a human or upstream artificial intelligence is responding.

ChatGPT is a NLP (natural language processing) and AI-based solution that has been developed on the basis of GPT-3 (Generative Pre-trained Transformer 3). The generative language model is capable of generating human-like text and responding to questions. Since ChatGPT is an AI model, the application can function within the scope of its programming and the available data and may need to be supported by humans.

The range of services is broad: Some users use the software to generate emails, blog posts, or texts on LinkedIn and Twitter, while others use it for spell-checking, translating, or rewording texts. IT professionals use ChatGPT to check and improve their software source code – or have entire software components developed by ChatGPT. Recently ChatGPT proved its capabilities by successfully passing tests at an elite university. But subject-specific applications are likewise possible, such as market or customer group research with the goal of more precisely recording credit risks, sorting and ranking keywords according to search intentions, or generating typical questions on a certain topic.

 

Use of ChatGPT in the finance industry

The financial sector in particular is facing ever greater challenges with regard to the digitization and automation of processes. The technology can be used in various fields of action to automate processes, enrich data in processes, analyze risks in more detail, and better understand customer needs. One thing is clear: Whoever succeeds in processing customer requests, applications and processes faster and more conveniently than ever before, and in meeting the requirements of customers, will gain a strategic competitive advantage.

With ChatGPT, ChatAI offers financial institutions the ability to analyze large amounts of data quickly and accurately. This can help, for example, to minimize the risk of investment portfolios, to improve the efficiency of trading strategies or to accelerate the review of credit risks and obtain more reliable results for the granting of loans.

 

Credit risk review with ChatGPT – the revolution?

Note: The use of ChatGPT is currently (as of January 2023) not in productive use in any fintus customer environment. A library for integration into processes is available for initial evaluations. The evaluation of the requirements for data protection, the type of commissioned data processing, and also the evaluation of the risk due to incorrect assessments and result types, for example in credit process processing, are still pending.

Given the dynamic development of the fields of application of artificial intelligence in the financial sector, it is surprising that the use of the technology in credit and receivables management has been rather sluggish to date. The appetite for risk and the willingness to engage in discussions with the regulator represent the supposedly highest hurdle for most financial institutions. Yet such intelligent applications as ChatGPT have the potential to permanently change and revolutionize the banking industry.

A snapshot: Overly optimistic risk assessments combined with payment defaults and misreporting/fraud endanger the profitability and thus the business model of banks. It is therefore essential to assess the creditworthiness and risk of the exposure. Particularly in more complex credit situations, for example evaluation of a borrower along with assessment of the industry and regional risk, this increases the workload and prevents a high level of automation – on the one hand. On the other hand, it is becoming increasingly difficult for companies and private households to extend their credit lines in the current climate. As a recent survey by the ifo Institute shows, banks are currently very cautious in granting new loans. According to the survey, 30 percent of all companies negotiating loans with banks now report a strong reluctance to lend. Back in September 2022, the figure was only 24 percent. Small businesses and solo self-employed individuals are particularly affected by this caution in lending.

With the help of technologies such as ChatGPT, various non-structured/public data sources can be included for the assessment of creditworthiness or risk, probabilities of default can be calculated more accurately, conditions can be adapted to changing framework conditions – and ultimately information can be contributed that supports automation or at least content enrichment of the decision in an automated manner.

Possible use scenarios of ChatGPT:

• Evaluation of a risk due to a concentration of similar companies in a region.
• Query about environmental risks in the applicant’s region based on historical data and media sources.
• Explanation of the business model based on third-party sources rather than research by the credit specialists.
• Detailed strengths and weaknesses analysis based on intelligent combination of relevant search queries.
• Background check on an individual’s business suitability based on the company/individual’s public visibility and possible ratings.
• A competitive analysis of the top 5-10 competitors in a segment or region, including a listing of recent media coverage with high reach.

Automotive Industry Risk Evaluation using ChatGPT

 

Low-code – the easiest way to process integration

The benefits of integrating artificial intelligence for automating decision-making or enriching content to prepare decisions via low-code are obvious:

1. Greater agility

One of the biggest benefits is the agility that low-code platforms bring. They enable banks to quickly adapt to changing requirements. Both questioning and interpretation as well as basic integration take place in the shortest possible time, creating an edge over the (analog) competitor institution.

2. Lower development costs

Low-code platforms require little or no programming skills, which increases the speed of implementation while reducing their cost by freeing up internal or external developers. Research has shown that low-code platforms can reduce development time by up to 90 percent.

3. Flexible integration

Low-code platforms integrate third-party systems into processes using standardized interfaces (REST) or as small Java delegates. Both the input to the interface and the result of the request can be processed immediately as part of this – and flow into the interfaces of the software application or into credit protocols.

Using ChatGPT in conjunction with low-code platforms provides competitive advantages for banks, factoring and leasing providers, development banks, lending institutions and other businesses. In particular, the high workload involved in initiating and closing more complex commercial contracts can be taken to the next level by combining innovative technologies such as ChatGPT with low-code. Financial institutions benefit from optimized and accelerated credit decisions with low effort and development costs thanks to low-code.

The advantages of low-code development over full-code development – explained by ChatGPT

 

Automated credit decisions using low-code solution from fintus

The innovative low-code banking platform from fintus is already successfully using AI in automated lending in the area of document processing to make lending decisions faster and more efficiently and to simplify the lending process for customers and employees. The fintus suite always uses state-of-the-art technologies to continuously optimize processes and adapt them to current developments, such as the possible use of ChatGPT.

 

Conclusion: The combination of ChatGPT and low-code represents a promising approach to credit risk assessment. The technology has the potential to fundamentally change the way information is collected and decisions are made in the financial industry. Challenges will include the ever-changing data environment, learning interpretation of artificial intelligence information, and thus limited reproducibility of decisions.