OPM applies AI to modernize federal job description, retirement processes
The Office of Personnel Management wants to apply artificial intelligence across the lifecycle of federal human capital management, launching new efforts to modernize the writing of position descriptions in the hiring process and automate customer service when federal workers apply for retirement benefits.
OPM Director Scott Kupor touted the agency’s new USA Class tool during an interview at the UiPath Fusion conference, presented by FedScoop, as a way to streamline notoriously slow and complex federal hiring.
“When you’re hiring people in the federal government, it’s a very complicated process of writing job descriptions, making sure those job descriptions deal with their own classification standards, which is what level should that job be, what job family does it belong to,” Kupor explained.
The federal government “has a lot of jobs,” the director said, with more than 600 classifications and a workforce north of 2 million civilian federal employees. “So the ‘n factorial’ is pretty significant.”
Kupor said OPM sought to leverage AI’s strength in digesting large volumes of information — in this case, thousands of existing job descriptions — to train a model, and then prompted it to create new position descriptions aligned with OPM’s classification standards. Federal hiring managers then review the outputs to ensure accuracy, further strengthening the model.
“If you feed AI a bunch of job descriptions … it’s a lot easier to create the next job description by you being able to type in and say, ‘OK, I’m hiring for a financial analyst, tell me what’s needed,’” Kupor said at the conference. “And then we can kind of spit out at least a draft of a position description.”
On the opposite side of the federal human capital lifecycle, retirement processing has remained a manual, paper-based process for roughly the past 50 years, according to Kupor.
“We send documents by ‘snail mail’ from HR departments and payroll departments to ultimately shove into an underground mine in Boyers, Pennsylvania,” Kupor said of the current process.
The director called the digitization and automation of that process “basic, obvious stuff,” instead pointing to OPM’s retirement services call centers as where the biggest gains stand to be made.
“Not surprisingly, we have way more calls than people who can actually service them,” he said.
To offset that, OPM wants to provide AI bots for retirees to communicate with, offloading the toil those call centers face.
In both cases, Kupor stressed the application of AI is not meant to replace humans. Instead, he said, it’s meant to complement their work, freeing them up from overload or arduous manual work to focus on the things they specialize in that an AI bot can’t do.
“It’s literally about making sure the people we have are there for the things that only the people can do. Right? So if you’ve got some complicated problem, then yes, I want you to talk to the individual,” he said. “But, if you’re trying to change your address on a retirement form, it’s just so obvious that we can do that in a way that is modern,” and not tied to the business hours of a call center.
Similarly, with the USA Class application, OPM is not looking to rely on AI to make high-level hiring decisions, Kupor said. “What we’re saying is if the AI can actually work from what we’ve done in the past, and it can streamline the process, I would much rather have my HR hiring managers focused on recruitment or talking to candidates or stuff like that, rather than having them spend so much time writing a position description.”
