How has ChatGPT changed the way we search

By Gautam Bandyopadhyay

In today’s digital era, artificial intelligence (AI) has emerged as a game-changer, revolutionising various industries by automating tasks, improving efficiency, and enhancing user experiences. One remarkable AI breakthrough is Chat GPT (Generative Pre-trained Transformer), a chat based interface to a large language model that uses artificial intelligence algorithms to generate coherent and contextually relevant responses to user queries. Since the introduction of the chat feature in Bing, Chat GPT has transformed the way we search for information, offering users a more personalised and user-friendly experience and offering real-time information. Chat GPT’s improved natural language processing makes it easier for users to find the information they need, without having to go through multiple search results and doing nested searches to deep-dive the topic of choice, keeping the context of earlier searches in the session. Google or Bing, it seems certain that GPT will transform the search engine landscape by providing more accurate and relevant search results. But can this powerful AI model also change the landscape of lending?

I believe generative AI has the potential to revolutionise the lending industry as well. The banking and financial sector has been at the forefront of adopting cutting-edge technologies and I am sure it will be the same when it comes to the adoption of generative AI-based tools. This list can be endless, but let me list a few key ones which I think can benefit the customers as well as the lenders alike.

Product discovery

When a customer has a need, the time taking and inefficient process to find the best loan product to suit the requirement can be a frustrating experience. Going through the FAQs, not finding the relevant details, and painfully waiting to connect up and speak to the right expert from your bank leaves us exhausted. Add to this, the plight of the not-so-savvy customers who can only hope to walk to the branch and manage to speak to someone empathetic to their requirements. This results in unmet customer needs, lower financial inclusion, and loss of good business for the bank.

According to a study conducted by Forrester Research, 55% of consumers prefer to interact with a company using natural language instead of a traditional user interface. GPT technology can be used to help the customer discover the right product, and engage the customer in their vernacular language of choice to generate personalised loan offers in real-time by processing vast amounts of data in the context of the customer’s needs.

This capability can also be used as a sales assistant to the relationship managers of the bank, ensuring that skill gaps are filled by the AI co-pilot. End result is a happy customer, an efficient bank staff, and a bank that can grow exponentially.

Credit underwriting

The inability to underwrite a customer, especially New to Bank (NTB) or New to Credit (NTC) leads to low credit penetration and loss of business. AI models have been used by lenders, but generative AI has multiplexed the potential. While both traditional machine learning and generative AI involve learning from data, their goals and methods differ. Traditional machine learning algorithms focus on understanding data and making accurate predictions. Generative AI, however, seeks to create new data that resembles the training data. The ability to generate synthetic data helps ensure the privacy of the original source of the data that was used to train the model. GPT based models can also be used in a more dextrous manner to learn from fraud data and predict newer cases of fraud that otherwise may take longer to learn and detect for traditional ML based models. This real-time analysis can result in faster loan processing times, reduced risks, and improved decision-making for lenders. Moreover, the personalised loan offers generated by Chat GPT can be tailored to individual borrowers, ensuring that they receive the most suitable financing options. Better underwriting models means better portfolio health hence, lowering cost of credit to the end customers. My take is that in the next one year we will see early adopters of GPT based underwriting models and such models may become mainstream in the next five years.

Customer servicing

Additionally, GPT in lending can also be used to provide personalised customer service. Lenders can use LLMs to answer customer questions, resolve issues, and provide support. Natural language processing (NLP) features allow it to answer customer inquiries, complaints, and information requests swiftly and efficiently. AI models trained on customer past interactions are skilled enough to handle edge scenarios, are available at all hours and provide contextual and personalised customer service. This can help improve customer satisfaction and loyalty. It can also be used to automate loan processing by bringing efficiency to tasks such as data entry, risk assessment, and loan approval and improve efficiency.

Improved collections

Synthetic data generation can help simulate customer behaviour and payment patterns. By training the models on the synthetic data, financial service providers can predict payment patterns and optimise collection strategies. By analysing customer data, transaction history and other relevant data, generative AI can help to optimise the customer communication strategy, including creation of personalised reminders, preferences of channels and timing and the rescheduling of reminders when needed. Basis on sentiment analysis, it can also help predict the probability of collection and alternate debt collection strategies. GPT based models can be trained to even assist in the negotiation process and result in more optimal settlement benefiting both parties.

Challenges

However, there are some obstacles that need to be overcome before Chat GPT can be fully leveraged by the lending industry. These include the accuracy of identity verification, biases and fairness of the model, interpretability and explainability of a model, data residency, privacy and security, regulatory compliance and the data quality and quantity. Historical data often does not capture emerging trends, market dynamics and changing economic environments. These require very careful handling of the data and algorithmic approaches to create ethically fair lending models.