AI Tools and Labor Law: Legal Guardrails and Looming Minefields

Episode Notes

Transcript

Nina T. Pirrotti:

The guidance, as far as I’m concerned, as a plaintiff’s lawyer, I would just say ignore it because nothing has changed in terms of the law. The Supreme Court has not overruled Grig’s.

 Daniel A. Schwartz:

Even if you decided that federal law, I don’t really need to worry as much about it, you still need to worry about various state laws. And with remote employees working in various states, all bets are off.

Matt Greer:

How is AI reshaping the workplace? In this eye-opening episode, Nina Pirrotti and Dan Schwartz discuss AI’s use in the hiring process and when evaluating performance and discipline. They highlight judicial, legislative, and regulatory responses, especially when AI creates risk of discrimination and offer tips for navigating this fast-changing technology. Hello, and welcome to the ABA Labor and Employment Law Podcast. I’m your host, Matt Greer. In my day job, I work for the Washington State Public Employment Relations Commission, where I spend most of my time serving as a public sector labor relations mediator, hearing examiner, arbitrator, and trainer, always as a neutral. Probably like you, my focus area is one small piece of the bigger labor and employment law universe. I enjoy broadening my horizons to this podcast and learning about what’s going on in the wide world of labor and employment law from experts in the field.

On today’s episode, I’m happy to welcome Nina Pirrotti and Dan Schwartz to discuss the fast developing use of artificial intelligence throughout the employment relationship and the legal landscape forming around that use. Nina represents employees and is a partner at the law firm of Garrison, Levin Epstein, Fitzgerald, and Pirrotti in New Haven, Connecticut. Dan represents employers as a partner at the law firm of Shipman Goodwin based in Stanford, Connecticut. Dan and Nina are both experts on this topic, and I really appreciate both of them joining us to share their wisdom on the podcast. Be sure to check out their full bios in the show notes. Thank you so much for joining us, Nina and Dan. I think this might be the first time where both of our guests have come from the same state, so it’s good to have a couple nutmeggers on the podcast. And yes, I did have to look up what the name was for people from Connecticut I did not know.

So apparently that’s the official unofficial way to describe Connecticut people.

 Daniel A. Schwartz:

We have to have you out to Connecticut. We can show you some of our New Haven style of pizza.

Nina T. Pirrotti:

Absolutely.

Matt Greer:

Yeah. I look forward to it. It sounds like an invitation to me. All right. Well, thank you so much for both being on the episode. I’m really looking forward to this conversation. I think it’s definitely a topic that is on the forefront of a lot of people’s minds, even if they don’t deal with it on a day-to-day basis. They’ve certainly heard about artificial intelligence and thinking about how that works in the labor and employment law world I think is really going to be interesting for folks. So I though we’d start off though, for those people who might be listening to this episode who don’t have a lot of experience or knowledge about what artificial intelligence is, maybe they’ve only seen it in some of its less serious applications like creating funny photos or videos or memes out there. I don’t know. Dan, would you mind giving us a quick kind of a 30,000 foot overview of what artificial intelligence even is?

 Daniel A. Schwartz:

Yeah, I’ll be happy to level set for our listeners here. So when we talk about artificial intelligence or AI, particularly in the employment context, we’re really talking about any software system that can make predictions, recommendations or decisions without explicit human instruction for each step. So in hiring, it might look like something that screens resumes or ranks candidates by algorithms or video interview tools that score facial expressions or tone of voice, anything that sort of takes inputs and produces an output that influences an employment decision. There’s a subset of that though too, generative AI or GenAI, and that’s your sort of new layer and that’s your headlines, the ChatGPTs, the co-pilots, the clauds. And those systems don’t really score or sort. They create content. So they generate, that’s why the term generative AI. So they can draft job descriptions or interview questions or write rejection emails, also summarize documents pretty well.

So both of those things can really have an impact in the employment relationship.

Matt Greer:

Yeah, no, it’s going to be very interesting to dig down into that a little more as we get through this, but I appreciate the overview. Nina, would you add anything onto that definition or does that kind of capture it from your perspective as well?

Nina T. Pirrotti:

I think it’s pitch perfect. As usual, Dan is spot on.

Matt Greer:

It’s good to have experts for sure.

Nina T. Pirrotti:

Except when we disagree.

Matt Greer:

Yeah. Well, I’m sure there’ll be opportunities and further parts of this conversation where we can distinguish things here. But in terms of defining what the actual artificial intelligence is, it’s good that we have that baseline for folks. I know I’m learning along as we go along with this. So I don’t know, maybe it seems to make sense to start off with a more substantive part of our conversation about artificial intelligence and the employment relationship at the inception, the hiring process. So in planning for the episode, Nina provocatively described the current dynamic in the hiring process with AI as being a bit of an arms race between job applicants and potential employers. And I thought that was a really interesting way of framing it up. I don’t know. Nina, can you tell us a little more about what you mean by that?

Nina T. Pirrotti:

Sure, Matt. We have found that as employers are deploying AI to screen candidates more and more widely, candidates are fighting to keep up. And in some respects, they really are doing so. They’re using tools to tailor their resumes to match job descriptions. They’re using keyword optimization. And here’s the really interesting part. They’re actually even embedding hidden instructions in their resumes. So I will tell you, recently I was having a very interesting conversation over a drink with a colleague who is an expert in AI, and he shared with me that he knows of applicants who are interviewed by AI bots who are actually in real time employing AI to make sure that their answers comport with what they expect that the AI bot wants to hear, which is taking it to a whole new level. But it really is in that respect, truly an arms race.

And what I wonder is, and it’s just a concern, I haven’t formulated a hard and fast opinion yet, but are we at the point where we’ve made the entire process faster, but in some respects really less trustworthy?

 Daniel A. Schwartz:

Yeah, if I can add, Matt, I think the AI arms race, particularly in hiring, it didn’t start with employers, it started with sort of volume. Employers can post a job and they would get hundreds if not thousands of applications per opening, and now a lot of those are AI generated or AI optimized. And so the systems can come in because really no human can process that type of volume manually. So you get better processing, some standardized screening criteria. Instead of different people looking for different things, you could have some of that same criteria. But in the notion of improving efficiency, these tools only measure what’s easiest to extract either, right? Keywords or credentials and not necessarily what’s most predictive of job performance. So it’s a learning process, I think, for both, and we’re still only in the early stages of this when you think about it.

Nina T. Pirrotti:

Yeah. Yeah. I think Matt, the question that ultimately is going to be asked is managing volume actually be getting more volume or is it actually producing a result that is manageable for the employer? And it really reflects the universe of candidates who are most suited for the position?

Matt Greer:

Yeah, no, a tool that’s driven by efficiency and dress towards volume. I mean, it’d be ironic if that ended up creating more work for people in the end as the tools are being developed. It’s an interesting concept. And it’s been a long time since I’ve been in the workforce. I was shocked to hear that there is AI being used in some of these ways, including interviewing applicants. I can’t even imagine how that would actually play out in a real-time kind of interview context. So it’s really, really fascinating.

Nina T. Pirrotti:

I will say this, Matt, it actually could be very funny. When I spoke at the ABA conference in Nashville on this very topic, AI and hiring, we started out because we like to loosen up our audience and make them laugh a little bit. We started out with this montage of real applicants being interviewed by AI bots and all the potential things that could go wrong, including at one point one AI bot is interviewing the applicant and then another one just magically appears. And then they’re asking competing questions and they’re talking over one another and it’s just the applicant looks absolutely florid and flummox. How am I supposed to handle this now?

Matt Greer:

I would certainly be confused. Yeah. So Dean, I’m curious from your perspective as an employer side person, I mean, have you seen the efficiency in cost savings? Has it played out in the short term in the practical sense with your clients?

 Daniel A. Schwartz:

Oh, I think so. I mean, I think they’re able to really get to candidates quicker and find what they’re hoping for, which is candidates who will be successful. Now that’s not to say that they’re successful all the time, but I think they’re increasing their odds of doing so through the use of these tools. The problem is you could potentially be excluding a class of employees, particularly those who may have disabilities from that applicant pool if you’re not particularly careful. So for example, some of these video interview tools might analyze facial expressions or tone or speech patterns to see if someone’s telling the truth or not. And I use truth in sort of quotes there. And if an applicant has a speech impediment or they’re neurodiverse or maybe has a visual impairment that interferes with their ability to use the portal, the tool may screen them out and a human may not even know much about it, and that’s not ultimately a successful use of the tool.

So there’s got to be a balance there. I think in the drive for efficiency, you can’t throw out all of the legal guidelines that still exist.

Nina T. Pirrotti:

And not only legal guidelines, but Dan, you just pointed out they could exclude applicants who were otherwise pitch perfect for the job, and what a loss for the employer that they didn’t have that opportunity to consider that person. So it’s an evolving process. We’re certainly not there yet, but I do agree with Dan that it’s playing the odds, and I understand that these are businesses that need to do that. And the idea is just to do it in a very thouhtful and deliberate manner and not to play fast and loose with AI because that’s when the employer gets into trouble and that’s when the employee is harmed.

Matt Greer:

Sounds like there’s a lot of important decisions in terms of rolling these out in terms of, I guess, which tools you have to use and the upsides and downsides of those various approaches. And I’m guessing that the easier and cheaper approaches may not be the best in the end, even though maybe some folks, that’s what they’re looking for. I think the AI is the answer for that is my guess, right?

 Daniel A. Schwartz:

Yeah, there are a lot of free tools that are out there, and those are the ones that we’re seeing getting people the most in trouble because they may not be as developed. But really, if you’ve been playing around, particularly on the generative AI side with some of these frontier models, the ones that are really at the cutting edge side, they’ve really gotten quite advanced in what they can do and the amount of data that they can process. So I try to remind myself when I talk with clients about this, the AI we’re using now is probably the worst it’s ever going to be. So if your only experience with AI was something from two years ago, it’s really quite different in just a very short amount of time.

Matt Greer:

Definitely fast moving, it seems like for sure. I know Dan, you gave a little bit of a heads-up about the legal implications, and obviously most of our listeners are people with legal backgrounds or legal interest. And I’m kind of curious, so you mentioned the disability aspect of it and how these tools might inadvertently, or maybe even advertently, who knows, I hope not, but be used for those purposes. I’m curious, what are some of the risks for employers? And maybe Nina as well on the employee side when they’re trying to navigate these tools at that initial hiring process stage where those disability issues might come up? Have there been ways to address that? And what are the legal liabilities right now that you see?

 Daniel A. Schwartz:

Yeah, so I’ll start tackling that. So the funny thing about this is there are really no new federal laws on this. The laws that have been in place in the employment context, and the listeners will be familiar with a lot of them, which will be sort of Title VII, your Federal Anti-Discrimination Act, the Age Discrimination and Employment Act, the Americans with Disabilities Act, all those laws were drafted well before AI, but those parameters still come in. And so if you are excluding a class of individuals because of a protected category, you could be violating these statutes. Now, the Biden administration had some agency guidance that started in 2022 with some technical assistance on the ADA. There was some Title VII adverse impact guidance in 23, but the Trump administration really repealed a lot of that. Now, that guidance wasn’t binding per se, so it never had the force of law, but it did provide employers and others with some thoughts of what the government might do in those instances.

But with the Trump administration de-emphasizing disparate impact enforcement, it’s hard for employers and people in the public to understand what’s still in play. And the bottom line is the statutes haven’t changed. Plaintiff’s attorneys like Nina are looking for these types of claims and are finding a few of them out there, so really need to be mindful of the existing laws.

Nina T. Pirrotti:

Well, Matt, fortunately, just because the EEOC and its current incarnation has saw fit to de-emphasize disparate impact, the courts, as Dan rightfully pointed out, are not bound by that guidance. In fact, Griggs versus Duke Power is still good law. And remember that that stands for the proposition that practices that are fair in form, but discriminatory in operation are unlawful. And you can see how that could potentially apply in this format because of course, AI tools are supposed to be neutral. They’re neutral, and certainly the AI tool itself doesn’t have any intent to discriminate. So you don’t have a bad actor the way you do in disparate treatment cases where somebody has an intent to discriminate and effectuates that by use of their power, control, et cetera. But here in the world of AI, you could potentially have a situation where the algorithm is fair in form, but discriminatory in impact because here is where AI is vulnerable, particularly vulnerable, and it’s in disparate impact law where algorithms are used that examine addresses that examine where you went to school, your educational forum, or other criteria that may be very neutral.

That sounds neutral, my address, where I went to school. But if the plaintiff or the class can show that there’s a discriminatory or disproportionate impact, negative impact upon a protected class, that is potentially a live and viable cause of action without any intent needed on the part of the employer.

Matt Greer:

So for those folks who are paying attention to the employment discrimination area of our practices, the EEOC, the Equal Employment Opportunity Commission, has made some recent pronouncements changing some of those guidances that Dan was mentioning around disparate impact. So maybe to follow up on that in your perspective as a plaintiff’s side attorney, I don’t know, for those who are less familiar with that, so the disparate impact is the bottom line that those don’t generally require a showing of intentional discrimination on behalf of a human being in the traditional kind of sense when we look at these cases. And so if that’s the case, then do you have any initial thoughts? I realize it fairly assumed about the new guidance that the EEOC is putting out there and how that might impact how you evaluate these claims and when clients come to you with these concerns.

Nina T. Pirrotti:

Well, the new guidance simply is saying, “Don’t place a lot of emphasis on pursuing this type of claim. We’re not interested in it. We’re going to focus on other things.” The guidance, as far as I’m concerned, as a plaintiff’s lawyer, I would just say ignore it because nothing has changed in terms of the law. The Supreme Court has not overruled Greg’s. So I would consider it to be so much noise as I do a fair amount these days. And I would just say, “Hey, this is the law. It has not been overturned, and this is how we evaluate it.” And it comes with a whole host of other issues in terms of maybe it’s not just the employer who’s on the hook. Maybe it’s the vendor. And isn’t that mind-blowing that the vendor, the innocent vendor, all the vendor was doing was putting together the algorithms to help out the employer.

But in disparate impact cases, and there’s particularly one which is Mobley versus Workday. Workday, as Dan will tell you, is one of the primary vendors of AI software in the employment context. They’re considering that right now. Derek Mobley alleges that a Workday applicant tracking system, which is used by hundreds of employers, discriminates on the basis of protected classes like race, in his case, disability, age. He applied to over a hundred positions with employers who use the Workday platform, and he was rejected by all of them. And his argument is just that. The questions were facially neutral, but they’re disproportionately impacting protected classes like the ones that he was in. And under that theory, Griggs is good law. Disparate impact, fair inform discriminatory in practice is good law. So in that case, Workday as a software vendor can actually potentially be held liable as an agent of the employer who’s using it, which is kind of wild.

Matt Greer:

Yeah, that is amazing. So I guess going back to that initial question though, so Dan, regarding the new EEOC guidance, I mean, are you advising your clients to rest easy that this new guidance is changing things? Or do you have a similar perspective to Nina in terms of how you view that? And then I think it is interesting given to some of those cases that are being developed here too, but maybe get your take on that first.

 Daniel A. Schwartz:

I tend to agree with Nina that the EEOC’s pronouncements here don’t really change the underlying law. And I think employers would be wise not to shift gears. Not that frankly, employers want to discriminate. I mean, that’s not really a good business practice. You’re not going to get good employees. You’re not going to get customers as a result of that. But I think the other thing to keep in mind is that even if the EEOC decides not to pursue it, you have individual plaintiffs and their attorneys just looking for class actions that are out there and you don’t want to be a target. Those are expensive, expensive claims to defend against. And then the other thing to really keep in mind is that states are coming out with their patchwork of laws that are in play here. So even if you decided that federal law, I don’t really need to worry as much about it, you still need to worry about various state laws.

And with remote employees working in various states, all bets are off.

Nina T. Pirrotti:

Good point, Dan. That’s important to keep in mind. And as we only have a smattering of them now across the country, but as time goes on, obviously individual states are going to be more and more proactive about addressing those issues.

Matt Greer:

We actually are about halfway through our time, believe it or not. Why don’t we go ahead and take a quick break here? And when we come back, let’s dig into if you have any insights from what we’ve seen from the state level and also some of the cases that are being developed in addition to what Nina’s laid out there. But first off, let’s take a quick break and give it an opportunity for the ABA labor and employment law section to give us an announcement about some upcoming events. All right, and we are back. So thanks. As we left off, we started talking about the state level approaches that are happening, and I think that might be interesting for people to hear about. Do you have any examples of what’s happening at the state level? And are we seeing any kind of trends in how states are handling these issues?

Nina T. Pirrotti:

Sure, Matt. Well, the first one that I know of is New York City with Local Law 144, which went into effect in 2023, which requires employers using AI, automated employment decision tools, to conduct annual bias audits and actually inform applicants of their use of AI. You also have Illinois, which requires notification when AI is used for employment decisions and prohibits AI-driven discrimination. And then there’s Colorado in 2026, which requires pre-used consumer notice and impact assessments for high-risk systems that has been interpreted to be inclusive of employment decisions. But most exciting, I think, is the developments that are happening in Connecticut, which I think Connecticut’s often on the cutting edge of things. And here they are again, and I know, Dan, this is an issue near and dear to your heart. If you wanted to share with our audience some of the developments in Connecticut with the hope that maybe others might follow.

 Daniel A. Schwartz:

Yeah. Connecticut is, depending on your perspective, maybe a canary in the coal mine of what could be coming down the pike from other states. And they just passed a law that was signed on May 29th of this year that’ll start going into effect on October 1st of this year. And it’s really probably the most comprehensive of the AI employment laws that we’ve seen. And so you’re seeing things like employers will be required when they’re issuing more notices to disclose whether the reductions are related to AI or other technological change. And that’s really one of the first of its kind reporting requirements. And then the following year in 2027, employers who are using these automated employment-related decision technology, as it’s called, have to give pre-decision notices to employees if they’re not going to hire them. Very similar to what the Fair Credit Reporting Act does for those background reports.

And then one last, I think, interesting tidbit is the law says that the use of artificial intelligence is “not a defense” to a discrimination claim. And so employers can’t just say, “The AI made me do it,” but you can consider anti-bias testing or other relevant information to show that the AI that you’re using is free of discrimination and is a reliable tool. So that’s a really interesting law to dig into, and I suspect other states will use it as a model. California, for example, has one that’s under strong consideration as well.

Matt Greer:

Great. Yeah, thanks. So it sounds like if there are kind of trends, I’m hearing the theme that disclosure is really a key piece of a lot of these laws is just not necessarily restricting it or prohibiting the use of issues that when it’s being used or how it’s being used, sometimes there’s going to be some legal requirements to disclose that to folks or to regulatory bodies.

 Daniel A. Schwartz:

Yeah. And I think in some part, many people know when you’re interacting with a chatbot or a video tool. But as these tools get increasingly more sophisticated, I think there is a concern that people may not realize that they’re not speaking to a human. And so some type of notice is now required, and I think that makes a lot of common sense. I don’t think that’s an out there proposition. I think one of the other trends that we’ll probably see is the disclosure of the use of AI when doing a reduction in force. As I said, Connecticut has passed that. California is considering that in their proposed law, which would be a 90-day advanced notice. So I think that’ll be something else that we see, particularly because the Federal Warn Act that requires employers to notify departments of labor if they’re doing reductions in force, it doesn’t have a similar requirement.

We’re probably not going to see something like that from the federal government. So I think it’s up to the states to at least track this and figure out, is this a widespread issue? Do we have to worry about this or is this really an issue in search of a solution?

Matt Greer:

Very interesting. It’s definitely something to keep your eyes on. I know that’s not the focus of our conversation, but since I do work in the collective bargaining realm, I do know in Washington State there have been proposals, I don’t think anything’s passed yet that would make the use of AI, if it was going to have some reduction in force or we take some workload away from humans that there will be an affirmative obligation to bargain over that with unions if there’s a unionized environment. So I know there’s some efforts like that along the way around the country too, I believe. I’m not sure how may have been passed or not, but another interesting global development there. I don’t know, I thought maybe we could go back to the whole the proxy theory that Dina gave us a little heads-up about in that the Mobile versus Workday case.

Are there other developments on that front? I mean, it is kind of interesting, and I guess the idea that these vendors or other third parties might be liable for some of these actions when employers use them and they have that discriminatory impact. I’m curious, is there anything else that’s interesting to share on that upfront and cases that people should be keeping an eye on?

Nina T. Pirrotti:

I think Harper versus SiriusXM, which was filed in August 2025 in the Eastern District of Michigan is one that we should all have our eyes on. Harper alleges that SiriusXM’s applicant tracking system discriminated against him on the basis of his race. Like Mobley, he applied to hundreds of jobs, or I think 150 actually, and was rejected despite having strong qualifications. And what’s interesting about Harper is this proxy theory. Mobiley, for the first time, you’re thinking, oh my gosh, vendors might actually be on the hook for this. Harper’s focusing more strongly on employers be on the hook via this proxy theory. They’re claiming that SiriusXM’s use of these AI systems used educational institutions, attended zip codes and employment histories as proxies for race. And if you screen graduates for certain schools out, residents of certain areas and employees of certain employers, and those criteria correlate with race, well, you may very well be engaging in disparate impact discrimination.

And guess what? It doesn’t matter what your intent is, you could have the best of intents. You could want the most diverse population possible applying for your work, but if that’s the impact, you potentially are liable.

 Daniel A. Schwartz:

Yeah, that one sort of keeps me up a little bit at night thinking about it because all of these criteria, schools, geography, zip codes, they all seem sort of facially neutral. Maybe individually they might be defensible as job related, but when you stack them together in an algorithm at scale, the cumulative effect of that can produce sort of pattern outcomes that look an awful lot like intentional discrimination there. And in the Harper case, the plaintiff is alleging both disparate treatment and disparate impact. But for employers that are considering using these automated tools, I think understanding what they’re doing is really important For employers to not just take it at face value. And that could be hard because some of these AI tools are black boxes. They sort of say, “We don’t know how it comes out, but just trust us.” That’s not going to be a great defense when you have to defend a disparate impact claim down the road.

Nina T. Pirrotti:

And there’s going to be lots of discovery that is going to get into trying to figure out the ins and outs of this so-called black box, because it’s not only employers that are in the dark, and they often are about how these algorithms are put together, but the vendors themselves haven’t so far been able to articulate with precision how they are. Some of that is because they’re claiming trade secret and they want to protect themselves, but others may be that perhaps they need to be a little bit more transparent all the way through the process in figuring out how these algorithms are created.

Matt Greer:

Yeah, that’s interesting. Maybe we should just go there. There’s some curiosity about litigating these cases and the unique challenges there are. And yeah, discovery seems to be one of those pieces, but are there things unique about the litigation of these types of cases when AI comes into play in terms of litigation practice?

 Daniel A. Schwartz:

What we’re really only at the starting point is what’s going to ultimately be discoverable? What’s the scope of it? In federal court, you have this whole proportionality side of things, but you have employees and managers and supervisors, everyone using these AI tools, sometimes the official AI tools and sometimes the off the books ones, the ones that are just commercially available there. How much discovery is going to be allowed? Can you get the search history of these tools? That’s going to be a big aspect. And then for the AI tools that are used in the hiring process, or we haven’t even gotten to the monitoring and all the other tools that are out there, how much is going to be allowed and who goes through that data? If you’re monitoring employees to see if they’re being productive, and you have computers that are recording video, that are then using an algorithm to see if the employee is using click strokes or is putting things into their proper format, how much data is out there?

I mean, we’re really at the beginning stages of an explosion of data. And so I don’t really see the courts yet fully tackling these issues. And I think they’re sort of hoping the parties work it out, but I think that’s going to quickly be the next frontier in these discussions.

Nina T. Pirrotti:

Matt, I’ll tell you one thing. Rule 30 depositions are going to be our best friend.

 Daniel A. Schwartz:

And our worst nightmare. Well,

Nina T. Pirrotti:

Exactly. I’m sorry, but I actually did. You took the words right out of my mouth because once we take the time to make sure our rule 30 is buttoned up and all of our topics are clearly delineated with precision as they must be in order to provide the employer with ample notice of what exactly was searching and what they have to do to go out and find it, we will have a one size fits all with some tweaks that we can use in any case in which AI is a decision maker and make sure that we get the answers. And I anticipate that that’s going to be followed by motions for protective order. They’re going to be motions to compel. They’re going to be unfortunately quite a tension between employees and employers and their lawyers on how these issues get resolved.

Matt Greer:

All right. So some discovery disputes that maybe are unique in this area down the road here. It should be, I don’t know if that’s fun to watch, but I guess it’ll be interesting for those who practice in that area. We’ll let you all decide if you think it’s fun or not. Maybe some parts of it are, I don’t know. You’re on the forefront, a forefront of a developing area. That’s kind of exciting, I guess,

Nina T. Pirrotti:

Right? Exactly. If we can look at it that way.

Matt Greer:

Yeah. Well, I am curious, so Dan kind of mentioned that we have been focused quite a bit on the hiring decision piece of this, the applicants and evaluating job applicants and hiring folks. How has AI been used in other aspects of the employment relationship, evaluating folks, maybe being used to determine discipline issues and termination issues and those kinds of things? I’m curious what’s going on out there that people should keep an eye on.

 Daniel A. Schwartz:

When you think about it, the employers are really using it to evaluate performance. Now, performance can be measured in a lot of different ways, but employers are using it for productivity monitoring, for schedule optimization, for discipline recommendations as well. For employers, my advice to them has always been keep the humans in the loop. Do not let the AI be a substitute for your judgment, for your observations. And I think there’s a push and pull there because the appeal of these AI tools is really hard to resist. They do save time. They can be useful in a lot of respects. If you’re evaluating 20 employees and having to write up their performance reviews, that’s a lot of time versus asking the AI tool to do a first draft of those evaluations based on various criteria.

Nina T. Pirrotti:

And Dan, I would say the operative words you just uttered were first draft.

 Daniel A. Schwartz:

Yeah.

Nina T. Pirrotti:

Because that’s where the human element comes in that you’ve been talking about.

 Daniel A. Schwartz:

And I think different employers are going to find different lines to draw there. And I think that’s ultimately where some of the litigation is going to arise from when the evaluations that employees are getting don’t seem fair to them, don’t seem based on their logic, and maybe they’re even hallucinated, sort of made up. That’s when employers will face those legal claims.

Nina T. Pirrotti:

And there’s just no phoning it in. Just like in our work, Dan, with generative AI, there’s no just letting AI take over and do our thinking for us. This has to be, as I said at the beginning of this podcast, a thoughtful, deliberate, and intentional process at every level. And that includes because yes, these legal platforms do hallucinate. That includes going in and making sure that all of the information that they’ve synthesized and so helpfully saved us time is actually accurate and comprehensive information. So yes, we can be efficient. We can use that as an initial tool. And yes, we also have to very much be there in the mix and making sure that the tool is being used responsibly. And I think that’s what we’ll be looking at, Dan, when we evaluate your client’s use of that product.

 Daniel A. Schwartz:

Every technology that comes about has this sort of appeal of, it’s going to save time. It’s going to make all of our lives easier. Like email, it’s going to save time. Does anyone think AI has made lawyers’ jobs easier? It just has increased the volume of it. And what I’m seeing is yes, AI is solving some of the grunt work and solving some of the issues, but it is creating a whole host of other things because people can now generate a lot more content more quickly. It used to be that you were limited in time based on what you can do. When some of those limits go out the door, now it’s more work, ultimately not less.

Matt Greer:

Yeah.

Nina T. Pirrotti:

That is so true, Dan. That is so true.

Matt Greer:

Bit of a paradox there. There’s a bit of a paradox in a lot of the things we’re talking about here in terms of how this is developing, it sounds like. So yeah, very fascinating stuff. We are getting close to the end of our time. I wanted to make sure we had a little bit of a time at the end to get your tips. I think you’ve been giving a little bit of tips along the way, but maybe just synthesize, what should practitioners on both sides be doing right now to stay ahead of these trends? And for those who are listening to this and are just like, oh my gosh, I am way behind here. What can they be doing? What kind of advice would you give them right now?

 Daniel A. Schwartz:

I think one is that employers and their counsel really should figure out who owns the AI decisions, designating personnel and clear authority for this, that AI governance is really a core compliance responsibility and you can’t outsource that. It’s your responsibility. And that means overseeing the vendors and making sure your vendor contracts are in order as well. And I would also think that that includes some ongoing monitoring of hiring decisions, for example, running your own disparate impact analysis to make sure that you’re not impacting a protected group. And then I think it’s trying to keep up with the state laws, having your notices in place, having your record keeping in place as required by some of these statutes. Those are the big tips. And then just learn and learn as much as you can about AI and use it even for your simple things. Because I think the more you understand, the less foreign it will be.

And just like when attorneys needed to figure out how to use Google at the beginning of this, you’ve got to figure out how to use AI because everyone is going to be using it in the next few years, if not already.

Nina T. Pirrotti:

I would just piggyback on that and say abandon your resistance if you do feel resistance. And I understand why. You have to abandon it because there’s no question that AI is here to stay. So what can you do to make it work for you both as employer and as employee? How can you approach it in a really responsible manner that enables you to maximize a positive impact? So I think the watchwords are vigilance. And listening to podcasts like these, Matt, because every time you expose yourself to a new aspect of AI, you’re learning more about it, you’re evaluating it, you’re talking with your colleagues about it, with your law firm about it, and you’re staying ahead of the curve, which I think is really going to be essential because that’s not going to be an excuse in any context. I didn’t know, or I didn’t mean it.

Unfortunately, the expression ignorance of the law is no excuse. Well, here, nowhere does it present itself more powerfully than here because there are plenty, I’m sure, despite how busy I am, Matt, there are plenty of well-many employers who do not intend to discriminate. And it won’t matter because if they don’t do their due diligence and they don’t stay on top of things and they don’t do that second layer of review, they’re going to be on the hook regardless.

 Daniel A. Schwartz:

Yeah, something tells me, Nina, we’re going to remain busy over the next few years. These issues are not going away. They’re really developing. So I think there isn’t a better time than now to just get a base knowledge of it and just keep learning a little bit. We didn’t figure this out overnight, but it’s not so complex yet that you can’t understand it either. If there’s a takeaway for a listener here, it’s be curious. I love the Ted Lasso show. If you fail on something, be a goldfish. Just be curious and try something new.

Matt Greer:

That is a great note to end on. I think we could probably spend at least another hour talking about this, and I’m guessing we’re going to have future episodes of the podcast as these things develop in the future, but we are at the end of our time. I want to thank you, Nina and Dan, really very much for this conversation. Definitely was eye-opening for me. I’m guessing for a lot of our listeners, they learned a lot of new things as well. And also some reassuring words about how maybe the problem is or that the issues aren’t so big that we can’t tackle them. So I really appreciate that. I also want to thank our listeners as well for tuning into this episode and hope you learned as much as I did and appreciate that. So if you do enjoy the podcast, make sure you give us a like and share it with your friends and other colleagues and listen to future episodes as they come out.

All right, thank you very much, Nina and Dan. Really appreciate you being with us.

 Daniel A. Schwartz:

Thanks, Matt.

Nina T. Pirrotti:

Thank you, Matt.

Continue Reading