AI IN HR: WHERE IT’S HEADED AND WHAT PEOS SHOULD KNOW

Every HR department has always had the potential for a lazy corner. The annual reviews that get written the night before they’re due, the copy and paste roll-forward of comments with numbers updated for the current year. The candidate screening process where resumes aren’t read past the key word matches. The HR policy questions that get answered with a ‘let me find out’ and the follow-through never happens. That corner of HR has always existed. AI didn’t invent the corner, but it did provide a better hiding spot.

AI tools have made it possible to be better at the judgement heavy work of HR as well as mask the appearance of effort without the mental investment. PEOs need a way to figure out how to tell these apart, not only for their own operations, but their clients as well. PEOs’ risk lies in the appearance of thought without the underlying effort. There are two very different things happening under the cover of AI in HR and they can look the same from the sidelines.

When humans use AI to supercharge their judgement-heavy work, efficiencies can be gained. Writing an email starts with a draft that a person then edits with the context and nuance of the specific situation. A manager uses AI to draft meeting minutes and reviews and corrects facts and context as well as adding details that occurred outside the meeting that the AI tool does not know about. This is using AI to jumpstart the judgement work.

When humans abdicate their responsibilities, it can look like not reviewing candidate resumes screened by AI unless there’s a complaint. It can take the form of performance reviews passed along from person to person with no real review. It can look like a chatbot that removes the responsibility of HR to know the policies. It looks like dozens of ghost jobs posted to recruiting websites to give the appearance of a healthy business to potential investors. These are the kinds of scenarios that PEOs should be concerned about and almost nobody is writing about them. It looks like progress and efficiency, until it isn’t.

There’s a lot of coverage regarding bias in AI hiring decisions. While that’s a real concern, the underlying assumption is that there’s an accountable human in the loop and often the AI solutions on offer are marketed and sold on the premise that with AI a less skilled human can do the process or may not even be needed at all.

The risk exposure to PEOs is that bad habits can lead to real exposure. Recent and pending legislation assumes the there’s a human in the loop. When an organization’s hiring process is abdicated and hollowed out, there’s nothing to back up decisions and no leg to stand on. The EEOC has already signaled that fairness requirements do not disappear just because AI was involved. The organization is still responsible for the decisions and outcomes.

Given the stakes the following are some litmus tests for identifying legitimate AI use:

1. Can the person responsible for the system explain the decisions the AI made, not just what the tool did, but why it did it?
2. Did the process get faster or just lighter? Reduced human involvement after AI is a giveaway for abdication.
3. If personnel had to explain AI’s decisions to a candidate face-to-face, would they feel comfortable backing it up?

If any of these cause discomfort, there’s a good chance the AI is abdication disguised as efficiency. The compliance checklist will look the same all the way up until there’s a problem.

The challenge for PEOs isn’t in evaluating technology, it’s in the role of what will the human still do after the technology has done its part. When we remove the human in the loop, we remove the main reason the process is defensible.

That said, evaluating technology is still important. What data was it trained on? What bias testing has it undergone? What kind of chain of thought support is available upon demand? Additionally, informing candidates where AI is used in your hiring process is just as important as laying out for them what constitutes legitimate and off-limits uses of AI on their side.

Asking good questions both before adopting AI and after it’s running in your environment can help manage the risk of the lazy corner quietly growing. AI can be responsibly used in HR, but must support the human side, not replace it.