Concerns about the disruptive effects of artificial intelligence in the workplace often dominate discussions about how the emerging technology will affect the labor market.
Many comments on the topic range from gloomy predictions about job destruction and the obsolescence of traditional skills to celebrations of the fortunes on offer for those who can unleash AI to improve performance.
However, for some employers and educators, AI is already helping to facilitate skill acquisition and improve existing jobs. They say technology can help organizations assess workers’ skills, plan for emerging needs and train their staff, boosting corporate productivity and staff’s career prospects.
“What we’ve found is that one of the best ways to learn about AI is to use it,” said Jim Swanson, executive vice president and chief information officer at Johnson & Johnson.
The pharmaceutical company uses an artificial intelligence-powered process called “skills inference” to evaluate and plan its entire workforce, in ways that would not be possible manually. “It is proving to be an important asset in helping us understand and improve the capabilities of our workforce,” Swanson says.
DHL, the international delivery company, uses AI to compare the skills that staff have and those needed in open positions. Through its “career marketplace,” staff can be guided toward the right training to advance their careers more effectively, and managers can be supported to fill vacant positions.
This use of AI encourages internal hiring, which is less expensive and faster than external hiring, explains Ralph Wiechers, executive vice president of human resources at DHL. It also means that candidates are more likely to be a good fit.
AI has other applications for quickly identifying and creating training materials for new skills, ideal when business needs evolve rapidly. “For an organization to be adaptable. . . To acquire the right skills, you need to automate it, compared to what happened in the past where you could prescribe a training pattern that remained stable,” says Wiechers.
Many companies that use AI in their workforce management infer skills using data generated throughout the organization, for example, existing job titles, the work that staff perform, technology activity, and supervisor reports.
At J&J, a dedicated team developed a company-specific skills taxonomy with 41 “future-ready” skills, such as data management or process automation. It then trained the AI to identify where these skills existed in the organization, based on workers’ previous experience, roles, and current positions. Workforce management systems, updated by employers and managers, create a data set to train AI models to assess skills and rate them on a competency level from zero (no skills detected) to five (thought leadership) .
In addition, AI also adapts recommendations for learning and development, suggesting to users the courses they should take to advance their careers in the company. Mapping the organization’s skills in this way “helps our leaders make informed decisions about hiring, retention and movement of talent,” Swanson says.
Other organizations are using AI to improve training itself, through simulations or giving more people access to personalized feedback.
At Bank of America, employees can use AI to practice difficult conversations, such as discussing sensitive topics with customers. By testing approaches with a simulation, staff can “practice real-world interactions in a completely safe environment,” says Michael Wynn, senior vice president of learning innovation and technology.
“It gives them a chance to develop some confidence and test their skills. . . that traditional methods do not allow them,” says Wynn. Managers can see where staff are improving faster by responding to the feedback AI gives them, and also where staff are struggling, suggesting areas for educators to focus on.
“Something that really helped us navigate the maze of technology was understanding that our students don’t want to learn the same way,” adds Wynn. “They don’t just want to read or watch training materials; “They want to be an active participant.”
Nick van der Meulen, a scientist at MIT who focuses on supporting organizations with technological changes, says AI automation allows employers to assess more skills, potentially with greater accuracy than existing approaches.
“You can give people an idea of how their skills stack up. . . you can say this is the level you need for a specific role, and this is how you can get there,” says van der Meulen. “You can’t do that with 80+ skills through active testing, it would be too expensive.”
But while the technology is “tremendously promising,” van der Meulen is also aware of its limits and the fact that developing the infrastructure requires work.
Similar warnings from others in the field underscore the idea that, despite the hype, moving assessments and decisions to artificial intelligence can still be complicated. Skills assessments are only as good as the data they are trained on, and human input is crucial for a system to work.
“It is necessary to have a definition [of skills] that is easy to understand and useful for an algorithm,” says van der Meulen. He admits that AI may not be “100 percent accurate” and that problems may arise, for example, if employees “do not make the effort to ensure that their digital footprint is complete.”
That means that, in most cases, it should be recognized as a rough assessment of skills that staff and managers can correct and expand on, rather than something definitive.
To overcome this issue, J&J allows staff to edit their skills history and add information (goals, interests, certifications) that may not automatically be in the data sets, to ensure the AI has as much information as possible.
These limitations mean caution is still advised when using the technology, says Nimmi Patel, director of skills, talent and diversity at Tech UK, the British trade body. “AI can process large amounts of data very quickly. But algorithmic evaluation as it exists today could struggle to understand the nuances of individual growth and development trajectories.”
She believes that “high-risk growth and assessment decisions are best suited to remain under human supervision” through a hybrid approach.
At J&J, Swanson emphasizes that AI skills assessments are not used in day-to-day performance management. At both J&J and DHL, participation is optional. But early numbers show that AI platforms have been popular in both organizations. “It’s about understanding the big picture of our organization’s skills and helping people know exactly where they should focus their learning,” Swanson says.