A new version of the ChatGPT generative artificial intelligence program has entered the market, and is getting a quick test in financial services.
Klarna has added ChatGPT Enterprise, implementing OpenAI's application programming interface to provide access for 2,500 employees, or about half of Klarna's staff, with the goal of getting all of the company's employees to use the technology.
ChatGPT Enterprise is similar to the standard ChatGPT in that it uses AI and language recognition to produce original content. What's new is a focus on companywide usage, bringing the tools to customer-facing sales and service staff. Klarna's experience will provide clues as to how a large payment company can broadly use generative AI when most firms are still experimenting with early, limited and
Klarna is using ChatGPT Enterprise primarily to expand access to generative AI while also improving the sophistication of original content for customer service and product recommendations.
These uses are good implementations for generative AI, according to Aaron Press, research director of worldwide payment strategies at IDC.
"These are language intensive concepts where inaccuracies have minimal impact, and where the value of generative AI has been well demonstrated," Press said. "They are not really fintech applications, but can add value to fintech platforms."
When it comes to other activities, like risk management or the actual payment, generative AI is not ready for prime time — and may not be a great fit anyway, Press said.
"Risk models require a full understanding of the reasons behind a decision, and current gen AI systems lack transparency and explainability," Press said. "Data-intensive payment messages require much more precision than large language models can currently provide. Non-generative machine learning and AI models, however, have proven to be very effective at reducing risk and optimizing payments."
OpenAI, the Microsoft-backed developer which launched the original ChatGPT in late 2022, at the end of August introduced
The original ChatGPT took the technology world by storm, making it one of the most downloaded programs in history in late 2022 and early 2023, and
OpenAI, which did not provide comment for this article, says in promotional materials that one of its clients, social media graphic design firm Canva, is using ChatGPT Enterprise to enable engineers to troubleshoot bugs, help data analysts organize broader data sources and aid financial analysts who are writing spreadsheet formulas.
Klarna's ChatGPT strategy
Klarna, which has a banking license in Sweden and operates as a payments company in multiple countries, is best known in the U.S. as a digital payment processor and buy now/pay later lender.
Like most companies that offer BNPL as a core product, Klarna is hoping to recover from a
Klarna did not make an executive available for an attributed interview, but said the company's integrated plug-in with ChatGPT offers curated product recommendations to users who ask the platform for advice, as well as links to shop for products via Klarna's search and compare tool.
The difference between Klarna's ChatGPT-powered shopping tool and other recommendation engines is that Klarna is aiming for shoppers who don't immediately know what they want.
While most people shop online for something specific, such as "running shoes," Klarna is addressing queries such as "I'm looking for a gift for my 5-year-old niece." ChatGPT uses a language model to parse language and vague search terms and match that with Klarna's product database for specific recommendations.
Consumers install the Klarna ChatGPT plug-in, and like all AI programs, the recommendations are designed to improve as the consumer uses the product more often. That's an incentive for repeat business, and a way to cross-sell Klarna's other products such as its BNPL installments.
For merchants, Klarna is offering access to an AI-powered shopping feed that digs into the retailer's product-level data and matches that with shopping behavior such as purchase history, browsing habits and preferences. This data feed updates in real time. Most shopping analytics systems perform these tasks already, but Klarna's goal is to provide recommendations faster, using larger amounts of data and producing original content for users. And it wants to place these tools into the hands of a larger group of employees.
Advanced large language models, which recognize conversational word patterns and uses data to produce original content, can improve the shopping experience by understanding complex user queries, deciphering context and delivering highly relevant search results, according to Daniel Fallmann, CEO of Mindbreeze, a company that sells cloud services for enterprise search and applied AI.
For example, a consumer searching for "comfortable tennis shoes for men" can receive recommendations based on the keywords of their query, Fallmann said. That can reduce operational costs and improve customer interactions.
Hurdles for advanced generative AI
Business opportunities from LLMs also come with risks, particularly in data security and privacy, according to Fallmann. "Enterprises must navigate the fine line between personalization and intrusion."
Also, the potential for algorithmic bias in LLMs should be addressed to ensure fair treatment of all users, according to Fallmann, adding that another challenge is data quality and reliability.
"Overreliance on public models like ChatGPT Enterprise without proper validation could lead to misinformation or errors in customer interactions," Fallmann said. "Businesses must balance technological advancement and responsible, ethical use to successfully innovate and positively transform company processes with enterprise-tiered LLM integrations."
Klarna may face challenges in its strategy to enable broad use of ChatGPT, and it's still not clear how much demand there is among consumers.
Rather than a blanket deployment of ChatGPT for all employees, the utility of the technology is based on a specific person's role, according to Aaron MacPherson, principal at AFM Consulting.
"Leaders should be doing research to find where it would be most useful, and then do targeted training and integration to improve employee productivity or satisfaction," MacPherson said. "Giving this to somebody in the wire room is probably going to lower productivity, as they are diverted from their main responsibilities."
The research on consumer demand for generative AI for financial services is mixed, even among financial technology companies that are deploying the technology.
Only 10% of U.S. consumers ranked AI-generated advice as a feature that would be "most interesting" in financial services, according to
ChatGPT Enterprise represents a significant step for OpenAI, said Wendy Li, senior vice president of emerging technologies at Marqeta. The new feature may not be the most optimal for Marqeta's needs, but the company is actively exploring and evaluating all available AI options, Li said.
"ChatGPT Enterprise is a compelling option for businesses that might not have their own in-house generative AI expertise," Li said, adding that it's crucial for every organization to evaluate the solution in the context of its specific needs.
"Enterprise ChatGPT, as a robust cloud version, is designed for companywide use. It presents immense value for many organizations, particularly those without a dedicated AI team," Li said.
