*Five Things to Know About Generative AI in HR*
By Workday Staff Writers
In a far-reaching conversation hosted by RedThread Research, Josh Lannin, VP of product technologies at Workday, was among the panelists discussing the knowns, unknowns, and no-nos of this red-hot emerging technology.
With generative AI breakthroughs coming at a breakneck pace, it’s hard to separate hype from true innovation, or to get a handle on how this might reshape your organization, if you’re not trying out the tools yourself.
Lannin encourages HR professionals to “get together with your teams, try some experiments, and get hands‑on.” Lannin has been leading prompt-writing workshops on how to best use ChatGPT, with teammates trying out techniques and assessing responses. It’s not about trying to change how people work—it’s about creating more comfort and fluency around this emerging tool, which is well worth the effort, Lannin says.
Generative AI algorithms are trained on large language models that contain vast volumes of data, enabling the AI to generate automatic responses to prompts or queries. For HR professionals who generate a lot of content—such as job offers, job descriptions, and statements of work—replacing the usual blank page with an AI-generated draft is a game changer. In fact, a March 2023 MIT study showed that workers using generative AI were able to complete writing tasks in 40% less time and with higher job satisfaction—and the quality of the content didn’t decrease, regardless of the user’s writing ability.
“When we think about efficiency gains for some of the rote writing that people do, if we can accelerate and augment that work, then we can allow them to move on to the work they really want to do and elevate the employee experience,” says Lannin.
Training generative AI isn’t something that happens only once, before the tool is unveiled to the world. Every time a user engages with the tool, the information they input is fed back into the large language model to further train the algorithm—and it can’t be clawed back or excised later.
So far, organizations have two paths to sidestep that peril: build their own large language model to run entirely on their own servers (an incredibly expensive undertaking), or establish a framework to understand how third-party tools and services using generative AI store your data. “I fully expect that different providers are going to have enterprise capabilities allowing customers to exclude data from training the large language model, as part and parcel of the system,” says Lannin. “But that’s still being worked out right now, and your employees may already be using these systems ad hoc.”
During the webinar’s informal poll, bias in generative AI emerged as the top concern for HR professionals, followed by accuracy and reliability. However, insights are another powerful antidote to passively overlooking bias, says Lannin. “The ability to benchmark how you’re doing as a company and understand where you’re doing well and where you’re not is key to giving people insights to help them avoid bias.”
The pace of generative AI can feel both dizzyingly fast and, paradoxically, somewhat slow: as technologists geek out over imagined capabilities and future potential, most HR users are toying around its edges. That disconnect doesn’t mean the transformation talk is all hype—only that “it’ll probably take more time than we all think it will, to see the fruition of some of this tech,” says Lanin.
HR professionals who may be underwhelmed by generative AI’s utility now would be wise not to discount it entirely. Rather, keep a pulse on how the landscape evolves, and recognize that the tipping point of utility likely won’t hit every user at once—and that’s by design. For instance, at some distant point in the future, it’s possible there’s “not a surface area that generative AI won’t have some role in playing, as part of the Workday experience,” says Lannin. But jumping straight there isn’t the goal. Instead, Lannin advocates for a staged approach, focusing first on core groups, such as HR professionals, before progressing to managers and, ultimately, casual users of the product. “We’ll build in the ability for customers who want to be early innovators to do that in a variety of areas and experience and learn from the technology along the way,” Lannin says.
Watch the complete webinar “Generative AI: Darling, Devil, or Disruptor of HR Tech?” on demand.
