Open AI’s GPT-4 enters financial services with Stripe partnership

AI-driven services are already in demand and considered an integral part of the lending industry's future. With the advent and increasing popularity of OpenAI’s ChatGPT, 2023 might be the year for AI banking solutions. 

The basic version of ChatGPT is free and driven by GPT-3, which is a language model that leverages deep learning to generate human-like text (output). But its newest edition ChatGPT Plus based on GPT-4 requires a monthly subscription for premium access to the service.

OpenAI has partnered with Stripe to enable payments for ChatGPT Plus and DALL·E (deep-learning model to generate digital images from natural-language descriptions) in a move to commercialize its consumer applications. In addition, Stripe has integrated OpenAI’s new GPT-4 into its products and services to improve the user experience for its customers.

Last year, Stripe began working with OpenAI to implement GPT-3 throughout its user support journey. GPT-3 has been helping support Stripe’s agents with problem diagnosis, routing, and solving users’ issues more quickly.

What does Stripe bring to the table for OpenAI?

Stripe supports and facilitates more than 25 payment methods for OpenAI.

OpenAI uses Stripe Billing to set up subscriptions to ChatGPT Plus and usage-based pricing for DALL·E. It is also deploying Stripe Checkout to accept payments via a prebuilt payments page. 

By using Stripe Link, which saves and auto-fills payment and shipping information for Stripe customers, so they don’t need to enter their payment details manually, OpenAI customers check out 40% faster, driving increased conversion and revenue.

“OpenAI isn’t just capitalizing on Stripe to bolster its sales. In fact, the tech firm is building an entire engine for commercializing these products with the payments firm — in a move to manage its revenue with Stripe as much as earning it,” said Emily Glassberg Sands, Stripe’s head of information.

As it grows its international footprint, OpenAI is also using Stripe Tax to facilitate its sales tax calculations.

Furthermore, closing the month-end books as fast as possible is one of the most common ― and time-consuming ― challenges companies have to face. To better manage its finances, OpenAI is tapping Stripe’s Revenue Recognition product to keep its accounting data organized, ensuring all monthly transactions are accounted for.

What does it mean for Stripe itself?

For Stripe, pooling its resources for the purpose of accomplishing specific tasks means adding value for its existing customers.

"GPT-4 will help us build better products and experiences for our millions of users,” said Sands.

The firm is starting by integrating GPT-4 into its developer documentation, Stripe Docs, which is a big part of how Stripe’s clients interact and grow. 

The documentation has a three-column layout for organizing navigation, content discovery, and live code snippet execution. While the top navigation breaks down the documentation into Stripe’s different product areas, such as payments, business operations, and developer tools, the left-hand side navigation offers a corresponding Quickstart guide for developers to get started.

According to Sands, GPT-4 will lend a hand to find, extract, and summarize the desired information that isn’t yet in narrative form, as well as narratively summarize the right sections to provide correct answers to user questions in real-time. That’s opposed to sending users to search for information themselves, for example. This way, businesses can spend less time searching and more time building. 

“We’re carefully training our GPT-powered Docs to provide accurate responses. By nature, it’ll improve as it goes from fine-tuning on our side and from feedback provided by our users and communities of expert developers, which is incorporated automatically,” she added.

Is ChatGPT in financial services ‘scary’ after all?

ChatGPT gained 1 million users in just five days after launching. However, it is still in the early stage of verifying its industrial value and still has a long way to go before gaining ground in banking and financial services.

“We’re in the early days of using GPT-4 at Stripe, and at the moment we’re focused on using generative AI to improve user experience. To be clear, we’re not using generative AI in our payment solutions yet or to complete any transactions on customers’ behalf,” said Sands.

Stripe is currently exploring and experimenting with the GPT-powered model on a limited number of products to better assess its impact on its consumers and different areas of the business – operations, and services.

GPT-4-powered Stripe Docs is one of the first among those products.

Stripe has been using non-generative machine learning techniques in its processes and products, including its fraud prevention tool Stripe Radar, Stripe Billing to process failed payments, and Stripe Adaptive Acceptance which selectively retries payments declined by issuers in real-time before a response is returned to the customer.

Other fintechs have also initiated exploring generative AI, too. BNPL provider Klarna uses OpenAI’s protocol to build an integrated plugin for ChatGPT. The plugin will be used to roll out a personalized shopping experience by introducing consumers to curated product recommendations via ChatGPT.

However, in addition to the current limitations of ChatGPT, and the existing issue of ML and AI bias – some experts have raised concerns that ChatGPT may have negative implications in the financial services sector. 

ChatGPT may not have the necessary security and privacy measures to handle sensitive customer information, such as account numbers or personal identification numbers – or the secure framework to process, analyze, and interpret KYC or transaction data. Additionally, trust remains the most important determinant for consumers when seeking financial advice. Lack of personal connection could result in consumers having a hard time trusting an AI model enough for their financial needs.

Elon Musk recently tweeted about ChatGPT being a “dangerously strong AI”, which indicates that the more advanced the service, the bigger the scale of (cybersecurity) risks involved.

Other experts are of the opinion that ChatGPT won't likely replace the need for all-human financial advice, as users will always want personalized counsel. But it has the potential to push humans up the chain.

“Anyone who wants to work in ML responsibly has to think about these kinds of challenges, and Stripe is no exception. Humans will always be in the loop as we build and test new applications of GPT-4 to ensure that our customers have all the tools and trusted support they need to run their businesses online,” noted Sands.