It’s no secret that A.I. has become a workplace mainstay.

Plenty of companies have for years incorporated A.I. into their business practice. The Associated Press has used automation technology to write quarterly earnings reports since 2014, and Amazon first piloted its cashierless Amazon Go convenience stores in 2016. But the recent emergence of A.I.-based tools like OpenAI’s ChatGPT has opened the door even further. Netflix recently used A.I. to draw backgrounds for a new anime short, and several airlines, including Swiss International Airlines and Lufthansa, are using A.I. for scheduling, weather forecasting, and estimating fuel usage.

Though new A.I.-driven possibilities are extensive, they usher renewed and very human anxiety about job replacement. Nearly 69% of individuals with graduate degrees fear losing their jobs to artificial intelligence (55% of non-graduate degree respondents report the same fear), according to a survey of over 1,200 respondents from chatbot developer Tidio. In 2020, the World Economic Forum (WEF) estimated that the labor shift from humans to machines would displace 85 million jobs by 2025.

While A.I. certainly stands to disrupt the workforce, an occurrence employers worldwide have been bracing for the last decade, some experts disagree that automation will lead to job obsolescence. “What we consistently see at the moment is some task replacement and jobs that substantially evolve,” says Julia Dhar, director and managing partner at Boston Consulting Group, who leads BeSmart, the firm’s behavioral economics and insights initiative. “But instead of a complete replacement, you’re seeing a shift in the way labor supply and demand is fulfilled, and so you actually can end up with shortfalls in really key skills and in-demand occupations.”

The same WEF report also found automation could create 97 million jobs by 2025, meaning employers must rethink how they train employees for the future A.I.-driven workplace.

A.I. has created the need for employers to rethink the skills and development workers now need to thrive professionally. “I don’t want us to think about humans in the workplace as just all about their efficiency,” Christie Smith, Accenture’s global lead of Talent & Organization/Human Potential, tells Fortune. “In the new employer-employee relationship, the organization needs to build the right skills and capabilities and invest in their employees in a whole different way.”

Dhar advises employers who want to fully reap A.I.’s benefits to consider the 10-20-70 formula, in which 10% of the organization’s effort lies in designing an adequate algorithm, 20% involves building the right technology and data to support the A.I. system, and 70% comes from how business processes are transformed, including appropriately supporting human workers to use A.I. tools effectively.

But squeezing all the juice out of that 70% requires employers to understand the skills workers lack and create customized learnings to upskill them.

However, employers should expect some aversion to upskilling. “Change doesn’t come super comfortably to human beings. And if executives and leaders are consistently out there only saying that this change is exciting and energizing, you’re unlikely to bring people with you,” says Dhar. Instead, leaders should design an upskilling plan that considers potential antipathy rather than anticipating only positive responses.

One way to quiet aversion is automating time-consuming tasks, particularly those burdensome to employees.

“Think of A.I. being used in a health care scenario by an X-ray machine operator. You make the machine so smart that it can do X-rays faster, and you free up time for the X-ray machine operator to do more creative work,” says Beena Ammanath, executive director of the Global Deloitte A.I. Institute. While more efficient, it’s not likely that more bone fractures will fill up the technician’s queue, so it’s important that employers proactively think about how to fill the free time that A.I. creates.

It’s also worth remembering that most Americans lack robust A.I. literacy. “What we see is a lot of what’s known as algorithmic aversion, where workers don’t want to trust the A.I., or they’re not exactly sure how to use the A.I.,” says Kate Kellogg, a professor of business administration at MIT Sloan. Over 84% of respondents to a 2021 Allen Institute for A.I. survey failed to exceed a passing score of 60% when quizzed on A.I. capabilities.

At Accenture, clients undergo an assessment of skills they have, skills they need, and the breadth of their skills gap. Then, client organizations develop specific learning paths to upskill and reskill employees. The most common skills needed fall into two buckets: data and analytics capabilities and human skills around transparency, empathy, and connection.

“I don’t think you can separate those [skills], and we’ve seen that our clients won’t separate them. They need to go hand-in-hand,” says Accenture’s Smith. For example, training on in-demand technical skills and effective management if an employee transitions from an individual contributor to a supervisory role.

Also, consider the non-automatable skills that workers must learn. Dhar identifies three essential skills to invest in: critical inquiry, assessing the quality of what A.I. tools produce, and ethics.

“That’s exciting and the scary and ambiguous part of the moment we’re in,” Dhar says. “We’re between the fully human world and automation, and giving people the skills to say: What does a good-quality question look like? Can I validate whether I have received a good-quality answer? And is that answer safe, appropriate, and ethical for me to be able to take action on? Those are the skills of the future.”

On-the-job learning is among the best ways to upskill employees for A.I., and managers can even use the technology to train workers. For instance, introducing chatbots or creating A.I. content and learning modules based on employees’ specific roles and functions. Such customized training drives engagement “because if you are a finance person using an example for marketing… [that] may not be as relevant, or it might not be easy to grasp,” says Ammanath.

But addressing A.I.’s rapidly expanding presence in the workplace also requires a critical look at the corporate culture. In a 2021 Accenture survey of 1,100 C-level executives and 5,000 workers, just one in six employees reported feeling highly connected to their organization and colleagues, and one in four felt that their employers were meeting their needs.

“In this new world of work, it’s not just about the technological skills,” says Smith. “To win in the labor market and attract, retain, and access the right talent, you’ve got to make sure that you are building those very human skills that you can’t replicate in A.I. or any kind of technology.”

As A.I. becomes all the more pervasive it will fall on leaders to future-proof their workforce or, consequently, miss out on the benefits of technologically empowering their employees.