AI and employment law: Trends, predictions and compliance

As health care organizations adopt artificial intelligence (AI) tools in human resources, employers must ensure that reductions in force — even in small practices — comply with federal protections, including the Age Discrimination in Employment Act, which requires specific disclosures when employees over age 40 are affected. Christopher S. Mayer, J.D., an employment law specialist with the firm Frier Levitt, explains why physicians and practice leaders cannot blindly rely on AI recommendations for layoffs and must review potential disparate impact across protected categories such as age, race and gender to reduce legal risk.

Medical Economics: This is still very new, but what is exactly the intersection between human resources and employment law and AI? What are some trends that you're seeing? Or what would you predict?

Christopher S. Mayer, J.D.: Well, I mean, certainly any sort of service facing industry or business, they have to carefully address and regulate how their employees use AI, so that a lot of that's going on in terms of adopting policies. But I do think in health care in particular, there's going to be an impact on, as we talked about, folks who do coding, claims, there's, there's going to be a huge impact on that. And that's where I think AI is really going to come into play over time and start to affect it.

But from the HR perspective, what HR people have to be mindful of when they're doing a large-scale reduction in force — and for a small provider, a large-scale reduction is three or four employees. It doesn't have to be 50, 60, employees. But the main thing they're going to have to be mindful of, just like you have to be mindful of in any reduction, this is no different than any other reduction that you're doing, is the is the impact that it's going to have across protected categories. What that means is, you still have to make sure, before you go through the layoff, and if you're using an AI tool to tell you who should be eliminated, you're going to have to sit down carefully. And sometimes this should be done through counsel so you can, have the attorney client privilege attached to it, but that's kind of another matter. But look at the impact that it's going to have, and the layoffs that, AI say, has suggested that you do, if it's going to eliminate five or six people over the age of 40 and no one under the age of 40, that's not a good look, and that's a problem. And not only is that a problem in terms of potential liability, but if you're going to do a reduction where you're trying to get a release in exchange for severance, which is a very good business practice, and something that any sophisticated employer does, you're going to have to disclose the age or ages of the people who are impacted by the layoffs, and the ages of the people are not impacted by the layoffs under federal law to get that release. The Age Discrimination Employment Act, which is the federal law I'm talking about, sets forth very particular requirements for a reduction in force that impacts, really more than one employee, a group of employees, and that is one of the things that you have to do. You have to show who by age. You don't have to identify them by name, but by job title and by age, is being impacted. And so if you're seeing as an employer a disparate impact, your employees who are being impacted by that are going to see it too. And not only that, you have to tell them they should go and consult with a lawyer before signing the agreement. So they're going to take that, the lawyer is going to look at that and go, don't sign it, we might be able to sue over this if they're going to go forward with it. So that's a long way of saying, you just have to be very mindful of the impact, and that it's not limited to age. I mean, obviously any category that's protected by law, race, gender, all the protected categories, you have to make sure you're looking at it and determining whether or not it's going to have a disparate impact on your workforce, and if it is and statistically, it doesn't look good. And there are tools that you can plug that data into that will tell you whether it's going to have an improper statistical impact. You might have to make adjustments to the layoff. Now this analysis is more appropriate, as you can imagine, for a large-scale layoff, whereas in a large-scale layoff, if there's a statistical anomaly, that's a problem. You know that. So any sort of statistical anomaly would be sorted out in a large-scale layoff, like, if you're doing five people and three happen to be over the age of 40, well then there may not be much you can do. That's a little different. But if you're doing a large-scale layoff and there's a huge impact on one protected category, that's a problem. But even in a small way, you just have to be mindful of the impact it's going to have, who's going to be affected by it, you can't just blindly accept what the AI tool has told you to do. You have to make sure that you protect yourself from liability.