What’s wrong with automating this part of the often long, difficult hiring process, especially for highly specialized IT roles? While using AI might be a time-saver, according to many in the recruiting world, it also robs the company of the ability to think deeply about what a job requires, as well as an opportunity to connect in a more human way with candidates.

Paul DeBettignies, the founder of Launch Hiring as well as founder and strategist of Minnesota Headhunter LLC, said that he doesn’t have a lot of faith in the use of AI for crafting job descriptions.

“If we’re going to automate everything, then hiring, finding a job, and recruiting is going to become even more transactional than it’s already been,” DeBettignies said. “We all already say we don’t like it, so we’re just going to do more of it?”

DeBettignies added that recruiting has always relied heavily on tech tools. Many years ago, a time-strapped recruiter might have used cut-and-paste to slap together a job description from other job descriptions found online, such as Craigslist. AI might only make this trend worse.

“For years, job descriptions have always sucked, and now that we’re using AI, AI has been training on crappy job descriptions,” DeBettignies said.

Failure to launch. Creating a good job description relies on insightful questioning. Managers must articulate who they might need to hire and why. According to recruiting author, facilitator, and speaker Katrina Collier, “most of them get it wrong.”

Fortune reported last year that 66% of managers are “accidental”; Collier said accidental managers haven’t been trained in managing a team, let alone in replacing someone’s role within it.

“Unfortunately, the managers just want recruitment to go away, it’s their least favorite task,” Collier said. “When you’ve got the likes of any of the large language models, OpenAI, whatever it is, they can just type in…whatever, and up comes a job description and they go roll with that.”

Collier said the description generated by AI often isn’t specific to a company and team. Instead, she encourages recruiters to have an internal conversation to work it out.

If a company chooses to lean into the AI description, DeBettignies can ask the model why someone might not want to apply for the role. He often gets the same three answers: There are too many bullet points, there isn’t information on why someone would want to work at a company, or there isn’t enough information on salary or benefits.

“My advice is to not fully automate this,” DeBettignies said. “I do appreciate speed and I appreciate efficiency. Hopefully it does get us to…where we are now able to do the human things more and better and deeper than we’ve been able to do.”

AI as a spackle of sorts. To some, like Steve Visconti, the CEO of cybersecurity company Xiid, AI is a tool that could be used to help fill gaps in job descriptions.

Visconti said he believes AI is a good tool for help with job postings, “because you don’t want to overlook something that should have been obvious.”

“I would write the job description—which I do, by the way, I do this—and then I generate an AI version,” Visconti said. “Then I try and merge the two and see how I can make it better. So, in a sense, AI didn’t save me a lot of time, it just made it better in that specific case. I think it’s a great tool, very valuable.”

Visconti pointed out how AI could help fill in required skills for a vital IT position, including cloud native, Kubernetes, OpenShift, and so on.

Collier agreed that the tool could be helpful if “you really know who you need to hire” and AI is used to help flesh out a description.

“It can be amazing if you’ve done all the research, but often it’s just a case of, I need a quick win,” Collier said. “They just go and ask, and then [AI is] pulling in all the badly-written job descriptions that exist in the world and going, ‘Yeah, here’s a great one.’”

This report was originally published by IT Brew.