As report after report prophesies how generative artificial intelligence will upend millions of jobs, many white-collar workers are wondering what it will all mean.

The release of ChatGPT in November spurred a surge of excitement and fear about the potential of this kind of technology to transform work. But change, especially at large organizations, takes time. In many cases, the hype has far outstripped companies’ ability to adapt. In the meantime, many employees — keen to stay ahead of the curve and lighten their workload — are waiting in limbo for employers to give clear guidelines and training.

Over 85% of employees believe they will need training to address how AI will change their jobs, according to a survey of about 13,000 workers across 18 countries by Boston Consulting Group Inc. So far less than 15%, though, have recieved any.

That may be about to change. While some companies have banned tools like ChatGPT outright or restricted its use, citing information security concerns, others have gone all-in on generative AI, scrambling to build and launch company-wide training programs to get staff up to speed.

For consulting firm PricewaterhouseCoopers LLP, that means rolling out mandatory training to its entire US workforce over the course of five months, starting in August. Given the concern among workers about what AI means for their jobs, PwC’s US Chief People Officer Yolanda Seals-Coffield said the first step is demystifying the technology. “The sooner we can get out and start to teach people about this technology, the sooner we can dispel some of that,” she said.

The company is dividing its workforce into three layers based on how deeply each needs to understand the new technology. The first and the broadest is mandatory training to bring all employees, regardless of role, up to speed on generative AI basics: what it is, how it works, best practices and how to use it ethically and responsibly.

A more defined second and third tier consist of software engineers, who need more technical training in order to integrate AI into internal systems, and senior leaders, who need a thorough understanding so that they can help clients transform their own businesses. “We don't want and don't need to have 75,000 deep subject matter technologists. That's not the goal,” Seals-Coffield said.

Though the training roadmap is detailed, the firm explicitly chose not to extend it past December. “Quite frankly we didn't go beyond that because we think the technology will continue to evolve,” Seals-Coffield said. “We want to make sure that we're not stuck and committed to something that by January will need to be completely redone.”

Meanwhile, other firms, such as Booz Allen Hamilton Inc., are taking a slightly less structured approach. The consulting company also offers formal training, though it’s voluntary and typically done on employees’ own time. Staff are given the opportunity to attend two virtual sessions per week on best practices, according to Jim Hemgen, head of strategic talent development. The firm also taps new hires for a full-time version of the same training program that’s recently been retooled with a stronger focus on generative AI.

Digital consultancy Publicis Sapient is tackling the AI question in a more targeted way. Although instruction will vary by the nature of job, the company will require that all employees learn prompt engineering, or the process of crafting precise questions to get the best answers from the chatbot, according to Chief People Officer Kameshwari Rao. As a first step, all engineers are being asked to complete this training by September.

Other companies are opting for a more learn-by-doing approach. Earlier this year Jeff Maggioncalda, the chief executive officer of online learning platform Coursera Inc., said the company would reimburse any employee who wanted to upgrade to the enterprise version of ChatGPT. Staff were encouraged to experiment with it as much as possible in their work and share what they learned in a dedicated Slack channel and in regular all-hands meetings. For Maggioncalda, speed was the most important thing, and that meant getting the new tech into the hands of his employees immediately so they could start learning as soon as possible.

Ultimately, though, Maggioncalda says he’s realized this kind of organizational transformation can’t be entirely bottom-up, with employees figuring out how to use it on their own, or top-down, with senior leaders dictating how everything should be done. Maggioncalda says middle managers will be a big part of the process, and that they will need to be taught how to teach their direct reports how to do their jobs differently — training that Coursera is now working on developing.

“I can’t tell every single person how the job is going to be different,” he said. “But you can’t just say, ‘Oh go figure it out.’”