OpenAI’s not-so-secret weapon in winning business customers? ChatGPT

Those individuals are paying for ChatGPT Enterprise or ChatGPT Team, which make up what OpenAI calls its “business products," according to OpenAI Chief Operating Officer Brad Lightcap. ChatGPT Enterprise is aimed at large companies with over 100 employees, and ChatGPT Team is geared at smaller firms with less than 100 employees, the company said.

OpenAI said it doesn’t disclose the number of businesses that pay for its services. It said 92% of Fortune 500 companies are using ChatGPT in some form, and 100 million people actively use ChatGPT on a weekly basis.

“We want to go maximize the number of companies that have a ChatGPT license, but if people are not actually using the technology underneath, then who cares? It’s shelfware," Lightcap said in an interview. “As we build better products, we just want to see that number [of individual users] go up."

The San Francisco-based AI company, which has been at the center of the current boom in generative AI, is up against many well-funded competitors in pursuing business customers, including tech giant Microsoft, its primary backer and partner. Unlike innovations in previous tech booms, generative AI is ideally suited for transforming the way businesses operate—and a growing number of firms are looking to cash in on it.

Owing much to its relationship with Microsoft, OpenAI has become one of the de facto vendors companies initially use to test generative AI, many chief information officers say. But increasingly, open-source models from companies such as Meta Platforms—which can be less expensive to use—are also putting pressure on OpenAI’s lead.

While OpenAI said it doesn’t have a market share figure, it referenced venture-capital firm Andreessen Horowitz’s survey, which found that 100% of the 70 enterprises it surveyed are using OpenAI’s models in some capacity, with 66% using its models in their operations, and 34% in testing.

With such a broad base of users—whether corporate IT-sanctioned or not—bringing ChatGPT into their workplaces, Lightcap said the chatbot is an “efficient" way for businesses to quickly understand the benefits of applying generative AI at work.

“Once you’ve introduced this appreciation for AI in your enterprise, through a product like ChatGPT, you’re starting to give your workforce an appreciation for how the model works," he said.

Software companies have similarly aimed to sell their products through a “bottoms-up" or “freemium" approach, where individual employees or developers bring free or paid tools into their workplaces—thereby opening the door for vendors to upsell corporate executives on a packaged solution.

OpenAI now has a team of about 200 developers, researchers, and sales and support staff to help directly sell its AI technology to companies, according to Lightcap, making its discussions with CIOs, chief technology officers and chief executives, “a very direct sales motion."

That figure is dwarfed by the size of enterprise sales groups at large, legacy software firms, which can number a thousand or more.

The most common business-use cases, so far, are “workforce enablement" or employee productivity, and companies that use OpenAI’s models to power their own products, such as customer support platforms, Lightcap said.

Home-improvement retailer Lowe’s is using a custom version of OpenAI’s GPT-4 model to improve search results on its website, and insurer Oscar Health used OpenAI models to build an assistant for managing claims, OpenAI said.

ChatGPT Enterprise is priced based on the number of users at a company, and ChatGPT Team costs $30 a month for each user when billed monthly. ChatGPT Plus, the company’s paid product for consumers, costs $20 a month. The free version of ChatGPT includes access to GPT-3.5, a less powerful version of its language model.

Launched last August, ChatGPT Enterprise is designed to address businesses’ concerns about securing their proprietary data, Lightcap previously said. It is built on GPT-4, the company’s advanced language model, and designed to address businesses’ concerns about securing their proprietary data. Lightcap said OpenAI doesn’t train its models on businesses’ data.

ChatGPT Team, launched in January, also includes access to GPT-4, but with some restrictions on usage and fewer management and security controls than the enterprise version.

In addition to ChatGPT Enterprise, OpenAI helps businesses develop their own apps and products, Lightcap said. The cost of building custom apps and products for businesses varies greatly, he said, depending on underlying model training costs and service costs, but typically ranges in the millions of dollars.

Still, some CIOs say OpenAI needs more enterprise credibility before they’re willing to go all in on its services. Startups are considered more risky than established enterprise sellers, they say, and many don’t want to take on additional vendor risk.

In the wake of OpenAI’s leadership turmoil last November, some enterprises decided to spread out their AI bets across multiple vendors, and are using a mix of proprietary and open-source models.

Lightcap said the company is seeing the opposite. “Trying to manage this across a lot of vendors is complex," he said. “We’re increasingly hearing companies tell us they really want to go with one partner."

Businesses still have the choice, though, between working directly with OpenAI to access its technologies and using Microsoft’s Azure cloud service, which also provides access to OpenAI’s most powerful models.

Some CIOs say it’s easy to flip a switch and add AI services to their existing Microsoft software or Azure setup. Cloud providers also offer data privacy and content safety assurances, regulatory compliances and software that makes it easier for corporate developers to put the AI models to use.

“We have our own business now that’s very independent," Lightcap said, referring to the firm’s relationship with Microsoft. “We look at ChatGPT really as the access point that any workforce in the enterprise should have to AI."

Write to Belle Lin at belle.lin@wsj.com