In the five days after OpenAI launched ChatGPT late last year, the artificial intelligence chatbot achieved what took Netflix over three years—acquiring 1 million users. 

And since its stable release last month, it’s caused even more frenzy with an “at capacity right now” error message frequently appearing on the site’s homepage.

Whether it’s CEOs getting excited about the prospect of increased efficiency or content creators fearing for the future of their jobs, too many people are checking out the viability of the chatbot than the site can handle. 

What’s more, many leaders have been so impressed by what they saw (or rather, are so keen to get ahead of the trend) that they have already begun implementing the chatbot in their workplace, with research revealing that it’s even already begun nabbing jobs from humans. 

Fortune spoke to around 70 CEOs to see how they’re actually using the productivity tool. 

While not a single leader admitted to replacing workers with the bot, it’s clear that marketing, research, and recruitment are the most popular jobs businesses are already enlisting the help of ChatGPT for. 

How CEOs are using ChatGPT for creativity and communications

Even before ChatGPT launched, professionals in the communications industry have long been using A.I. tools like Grammarly for spell-checking and paraphrasing. 

Sameer Ahmed Khan, CEO of Martech firm, Social Champ, is implementing ChatGPT similarly to elevate, instead of replace, people’s work.

“While no department views ChatGPT as a threat, team members often tease the content team that ChatGPT will soon replace their jobs,” he says. “In reality, however, ChatGPT only complements their work and streamlines their workflow.”

For example, Khan’s marketing team uses ChatGPT to automate tasks like generating appropriate HTML meta tags for blog posts, conducting keyword research, obtaining email outreach templates, and identifying link-building opportunities. 

Likewise, Chris Camacho, CEO of the creative agency Cheil UK, has seen a surge in the service being used by busy marketers in the advertising industry to quickly generate copy based on trending topics, to improve traction and SEO. 

“Although I have been impressed by the results, I have seen some outlandish answers—the quality of response is very much dependent on what a user inputs into the system,” he warns. “For that reason, our teams are treading carefully.”

“A.I. technology has its inevitable limits,” Gareth Davies, CEO of the creative agency Leagas Delaney, echoes. “Whilst ChatGPT can regurgitate an established argument, what it can’t do is imagine a compelling new territory, nor can it present that idea in a fresh way.”

That’s why Michael Alexis says his employees are using ChatGPT exclusively for short-form content generation and leaving the genuinely creative work to people.

The CEO of teambuilding.com, which runs team-building events for clients like Apple, Amazon and Google, says “these tools can be handy at producing snippets of text such as meta descriptions, outlines, social captions, automated emails, and one-off blurbs for link-building.”

Alexis’s company has even invested in apps to automatically generate prompts for his writers.

His view is that by alleviating some of the administration and tedious tasks involved with writing and marketing for online consumption, ChatGPT can free up his writers to get on with exactly that—writing. 

But for businesses that have large numbers of workers dedicated to alleviating the administration of their creative department, “a radical reimagining of their talent strategies will be required,” cautions Davies.

“By its very existence, it will force businesses to change, to focus on problems that machines can’t solve, placing greater emphasis on human-centered strategy and innovative creative solutions,” he adds.

“Companies that understand that fastest will not only win but arguably produce the most persuasive content in doing so.”

Most CEOs are using ChatGPT for general assistance

“I’ve shaved off an hour of my day thanks to ChatGPT,” boasts Vishen Lakhiani.

The author, CEO and founder of the e-learning provider Mindvalley says that within two weeks the A.I. tool had permeated every aspect of his business. 

Most of the CEOs that Fortune spoke to are similarly encouraging their workers to use ChatGPT as a day-to-day assistant like Iron Man’s virtual butler J.A.R.V.I.S.

It’s made such a difference to productivity at Lakhiani’s firm, that prompt engineering—the knowledge of how to prompt A.I. to deliver results—is now the “number one thing” Lakhiani looks out for on job applicants’ CVs. 

“Although ChatGPT and its rivals are far from perfect, their ability to enhance productivity in knowledge work is plainly evident,” echoes Nigel Vaz, CEO of the global digital consultancy Publicis Sapient.

Vaz says that his workers are already using A.I. to assist with conducting research, assessing computer code segments for errors and communicating with clients in non-native languages—and it’s a productivity win for the business. 

“In the near future, our knowledge workers will experience productivity improvements ranging from 10 to 30% depending on their roles, thanks to the implementation of Generative A.I. assistants,” Vaz adds.

Why CEOs are still cautious

Much has been written about the dangers of A.I. reinforcing biases, which is a reason a number of CEOs remain cautious about implementing it, especially in client-facing tasks like recruitment and customer services.

“Despite its sophistication, ChatGPT is capable of racist, sexist, and other problematic responses to prompts,” says Michaela Jeffery-Morrison, CEO and founder of Ascend Global Media, the platform for women in tech. 

Jeffery-Morrison points to racial bias found in the U.S. health care system and gender bias appearing in Amazon’s old recruiting engine as examples of why she won’t be implementing the likes of ChatGPT until it’s “reached a point of maturity.”.

“I’m studying the ChatGPT concept but will not be jumping on the bandwagon right away,” chimes Baruch Labunski, CEO of the cybersecurity firm Rank Secure.

As a rule of thumb, Labunski says he never buys into “the first model of anything” because, much like that first pancake you cook, it is almost always imperfect. 

Although ChatGPT could ease a lot of businesses’ customer service administration, like answering common questions, the possibility that a chatbot would throw out an inaccurate, or worse, racist response isn’t worth the risk for Labunski.

But even if those issues were ironed out, the customers and clients Labunski spoke to would rather talk to a human than a bot. “They are all suspicious of it so it could cost me customers, which is a reason why I’m cautious.”