Generative AI Helps Firms Take Aim at Legacy Cost Centers

CFO Voices: Modernizing Back-Office Operations

As large language models (LLMs) like ChatGPT keep growing, enterprise headcounts might start shrinking.

This, as a new paper from OpenAI researchers estimated that the potential implications of generative pre-trained transformer (GPT) models, and next-generation software tools built atop them, could impact up to half of the tasks necessary for 19% of the jobs in the United States.

“Our analysis indicates that the impacts of LLMs like GPT-4, are likely to be pervasive,” the study stated, adding that four in every five U.S. workers could see at least 10% of their tasks affected in “some way” by ChatGPT.

Jobs that already involve a variety of software-driven tasks could potentially face more disruption from artificial intelligence (AI) tools, whereas industries reliant on manual labor, including food service and hospitality, forestry and others, are likely to see the least impact.

“While LLMs have consistently improved in capabilities over time, their growing economic effect is expected to persist and increase even if we halt the development of new capabilities today,” the paper stated.

Read more: Automation Does the Work Accountants No Longer Want To

What Would You Trust AI to Do?

Technical innovations powered by generative AI have taken on a rapidly increased prominence across the marketplace.

Already, OpenAI’s latest AI chatbot, ChatGPT-4, is capable of passing the bar exam, extracting insights from financial reports, program computer code, and draft marketing copy.

But will it one day replace lawyers, accountants, software engineers and ad agency creatives?

Microsoft, an early investor in OpenAI, is integrating the new GPT-4 AI model into its revamped suite of Office applications, including Excel, PowerPoint, Outlook and Word.

“ChatGPT is going to be in everything,” General Motors Vice President Scott Miller said earlier this month in reference to his company’s plans to use the chatbot in its vehicles.

Still, enterprise reception to the use of AI-powered chatbots by their employees has been mixed.

Nearly half of human resource leaders are currently in the process of formulating guidance on employee use, while leading financial institutions including Bank of America, Goldman Sachs, Citigroup, Deutsche Bank and Wells Fargo have banned the use of ChatGPT outright. J.P. Morgan has limited its use by employees.

Meanwhile, Bain & Company and OpenAI have together formed a global services alliance to bridge the synergies of Bain’s digital implementation capabilities and strategic expertise with OpenAI’s AI tools and platforms in order to help enterprise clients realize the future-fit value of the intelligent technology.

While activating AI to create efficiencies and make operations more seamless, there remain areas where it is best suited as a supplement rather than a replacement, at least for now.

For example, healthcare patients are likely to feel more comfortable having a physician review any AI-powered assessments before prescribing a course of care. AI for healthcare is better suited for automating the back office and surfacing medical histories in real time than it is at performing back surgery.

At least for now.

How AI Is Helping Payments Executives

Companies have long been aware that non-strategic finance is a cost center.

There’s been an explosion of opportunity for the automation of remittances and payments, which in turn improves the accuracy and productivity of those workflows, and the accounting and finance office has already proven to be one area where applications of AI tools can have their greatest impact.

“We’ll see fewer people in these departments in three years, and I’m a firm believer that humans are not doing this job at all in five to 10 years.” OpenEnvoy CEO Matt Tillman told PYMNTS’ Karen Webster in an earlier conversation.

Payment platform Stripe has been serving as an enterprise beta tester for OpenAI’s Chat GPT-4 interface, trialing the efficacy of its next-generation capabilities across operational areas that include streamlining user experience as well as combating fraud.

Stripe’s head of information and data science, Emily Glassberg Sands, told PYMNTS CEO Karen Webster earlier this month that Stripe currently has 14 GPT-4 prototypes in the works.

The prototypes include a way for Stripe’s software developers to type out a question and receive summarized answers, as well as a customer-facing solution that allows Stripe customers to make queries and receive answers about their own business analytics.

“There is a lot of opportunity to build new user-facing products, or those that better delight users in an existing experience, using AI,” Sands said.

And for each of those 14 prototypes, Stripe will be able to restructure the teams previously responsible for interacting with customers and surfacing data to respond to queries.

For all PYMNTS B2B coverage, subscribe to the daily B2B Newsletter.