AI in Workplace Investigations: What’s Actually Changing

[Melanie Naranjo] (0:00 - 4:52)
Oh, hello. Chantelle, I take offense to that comment. What do you mean wind down?

We're still mid-deep into summer. Don't tell me it's almost over. I just got back from the beach, okay?

Don't tell me it's over. Hey, you know what? In California, our kids go back to school, or they've already gone back to school, and so that summertime feeling starts to end.

But that said, I'm from San Francisco, so our real summer is about to start in a couple weeks after surviving the fog. And for all of you, those of you that are in the Bay Area, it's the coldest summer on record in a couple years. So I'm looking forward to some summer in September and October.

Oh my, that is just wild. I don't envy you at all. I'm literally sitting here drinking like one of those IV, like mineral pack waters to rehydrate from all the sun I got this summer, or this weekend.

So hi, all. Welcome, welcome. For anyone who's new, my name is Melanie Naranjo.

I'm the Chief People Officer here at Athena, where we run monthly webinars focused on the HR audience, because one of the things that I have learned throughout my career in HR is just how lonely it can be, and how easy it can be to get siloed and not know what best practices are, because so many of these topics are so sensitive that you don't want to talk to other people about it, because you're like, what if my company's doing something wrong?

I don't want to ask, and then inadvertently give something away, and show that we were making a mistake. And so we kind of do all these things like employee terminations, investigations, all in a silo without talking to each other, and I don't want to do that. So one of the things that we as a company have done, being that we are a compliance training company, is we've decided that we're just going to build community, and we're going to share knowledge, and we're going to bring resources to the HR community, leaning on thought leadership and experts like Rabih and Chantelle, who I'm going to introduce in more detail in a second, to chat with you all about everything that you need to know, the latest and greatest, when it comes to these tricky topics. So with that said, as you all know, we are here to talk about AI and workplace investigations, because we've been hearing murmurings that more and more people are thinking about using it, or feeling pressured to start using AI in workplace investigations, and we want to talk directly with the lawyers around, you know, what is happening, what are you seeing, whether it be with clients or just in the news, and what's safe, what's not, how can you be safe about it. The usual disclaimer before I kick this off officially, is Rabih and Chantelle are lawyers, they are not your lawyers, so this is not legal advice, okay, all this is not legal advice, this is me, yeah, having a chat with two very awesome lawyers, and thinking aloud, but please don't take any of this as legal advice, and say, oh I saw this webinar, and Rabih said to do this, that's what we're doing, every situation is case by case, which requires, yeah, so much, so much more context and questions to give actual advice, and also you need lawyer privilege, which you do not have on this call, let me just put that out there, you do not have lawyer privilege, if anybody wants to sneak Chantelle a dollar or something, Venmo her, maybe that changes, I don't know, I'm not going to get into that. With that said, Chantelle and Rabih, so Chantelle and I go way back, we've known each other for like three, going on three years now, I think, introduced, I think you're right, yeah, introduced through Brandice, our own Athena in-house counsel, so Chantelle Egan is a partner at Medina McKelvey, and Rabih David is a senior counsel at Medina McKelvey, this is not my first time meeting Rabih, but our first time partnering with you on a webinar, Chantelle, you've been on several webinars, you two, why don't you all, introduce yourselves a little more, tell us about how you know each other, well, maybe it's obvious, how you know each other, your history as employment lawyers, and then we'll go ahead and kick off the session. Great, well, I will get started, as Moni said, my name is Chantelle Egan, I'm a partner in Medina McKelvey, and I am the head of our investigations practice group, I've been an employment attorney for almost two decades, and I transitioned from a large firm called Cypher Shaw about three years ago to Medina McKelvey, at Cypher Shaw, I started their national investigations practice group in crisis management, and then since coming to Medina McKelvey, we've really leaned into our investigations practice, where we're focused on not only providing neutral investigations, but bringing a sense of humanity to that process, and really helping it be something that is a, actually helps heal as part of the process, and then, of course, we do a lot of trainings with HR professionals to help you do the same things when you're getting tasked for it internally, Rabih?

[Rabi David] (4:53 - 5:45)
Yeah, thank you, Chantelle, my name is Rabih David, I'm a 15-year attorney at Medina McKelvey, came here about five years ago, doing employment law, and I work alongside Chantelle in workplace investigations, before that, I was at Oregon Sutcliffe, a large law firm where I did white-collar investigations, so similar, but securities-based and account regulations, so that was a little drier than what I do now, which is workplace investigations, and I get to work with HR professionals like yourselves much more regularly.

Before that, I was a JAG in the Air Force for a couple of years, I was a plaintiff's lawyer, and before being a lawyer, I was in the Air Force as a navigator and maintenance officer, so I've done all sorts of different things, we'll see what happens next year, but for now, this is what I do.

[Melanie Naranjo] (5:47 - 7:58)
I'm following you on LinkedIn for more, you're going to be doing this for me, that's what you're doing. That is exactly right, can't wait. For anyone who's new here, please note that these are interactive sessions, so I will be asking questions of the audience, and my ask is that you just respond to your comments in the comments section.

If you don't want to be called on, I don't want that to be a deterrent to participation, so in your response, when you are responding to the question that I ask, put an asterisk next to your response, and that will be my cue not to call on you, it's just helpful to get engagement, see what other people are thinking, and my recommendation to everyone is lean on the comments, connect with each other, this is your chance to build a network, learn from each other.

If you all are hearing us talk and you have a tip or you want to add on to it, drop it in the comments, let's get that conversation going. All right, so with that said, as I mentioned, we'll be talking about AI in investigations, the good, bad, and how to put it into practice. We'll do a brief Athena spotlight at the end that ties into employee investigations and how we are helping with that, and then we'll do a Q\&A.

All right, so let's get into it. First, first pulse check for you all in the comments, get ready. On a scale of one to five, how much do you agree with the following statement?

Five being totally agree, one being yeah, that's wrong. Nothing could possibly go wrong. Nothing could possibly go wrong by leveraging AI in workplace investigations.

So one being like, yeah, I don't think so. That sounds like a recipe for disaster. Five being like, yeah, I'm chill, everybody.

Why haven't we been doing this already? Wow, interesting. Okay, it's funny to me.

It's funny to me because a part of me didn't know how this was going to go, just given that people joined a webinar focused entirely on AI investigations. So I was like, maybe people are already doing it. They're curious how to do more, but we're seeing a pretty consistent pattern.

Okay, we got a couple people more, brave souls, love it, going a little higher, but mostly on the low end. Let me throw it to you, Rabih. As you see these responses, I heard a chuckle on your end.

Why a chuckle? Are you surprised by these responses? Sounds like maybe not.

[Rabi David] (7:59 - 8:17)
No, I'm not too surprised. I chuckled when I saw it. I think it was Bill Parkin who threw a negative five down.

So clearly a Luddite at heart and not interested in leveraging AI for workplace investigations. But I understand the concern and that's why we're here to talk about it.

[Melanie Naranjo] (8:17 - 9:28)
Yeah, totally. Chantelle, agree, disagree? Are you like, Rabih, what are you talking about?

I've been doing AI. I think the thing is that AI is very new and lots of times when things are new, our instinct is to be nervous about them and to be fearful about them or to be cautious about them. And so hopefully by the end of this conversation today, people will feel like they have better understanding of how they're going to see AI in investigations and also how they can use it as a tool in order to do their job more efficiently without it compromising the investigation.

I love that. Very optimistic. Nice, nice.

Okay, let's call on someone just so we can get a little more information about what's going on in people's heads. Someone without an asterisk, Laura Petrolito, P-E-T-R-O-L-I-T-O. If you could unmute yourself and tell me a little bit more about why you gave a one here.

What's scary to you when you think about AI and workplace investigations? Where do you think we could go wrong? And if I don't see you unmute yourself in a second or two, I will assume technical difficulties and call someone else.

Oh, there you go. Go ahead, Laura.

[Chantelle Egan] (9:28 - 9:59)
Hi there. I think for me, I saw a comment in here that says it's how you phrase the question to HR professionals, something can always go wrong. I completely agree with that.

It's a little bit of my risk eyesight that's here. But to me, I do really think it's just that it is so new. I also have just concerns about the privacy element of it and how are we making sure that the information is secure?

I think it was the biggest indicator of why I rated it so low.

[Melanie Naranjo] (9:59 - 10:17)
Laura, can I have a quick follow-up question for you? If it was up to you, would you be like, so then let's just not use AI at all? Is it like, are you yourself trying to push yourself or are you getting pressure from elsewhere to maybe try and explore a little bit more about how AI can be leveraged?

[Chantelle Egan] (10:17 - 11:01)
Yeah, I'm definitely, I'm all for it. I use it a lot personally. So I think I'm a huge supporter of using it in the workplace.

I think this is just a topic that would need to be taken pretty seriously. And for me, I wanted to learn a little bit more because again, I understand it is a serious security concern and privacy element. So we need to be really thoughtful and understand all of the regulations before we just go about kind of using it for this purpose.

I think there's a lot of purposes for training or leadership development or coaching that I think is a lot safer to start with. This is one I think is, to me, a little bit less safe use case. Interesting.

Okay.

[Melanie Naranjo] (11:01 - 17:29)
I'm going to comment on this, but in the meantime, if you all could do me a favor and also Laura, thank you so much for sharing. I really, I always appreciate people being vulnerable and open on these calls. Helps it feel more authentic and real.

In the comments, can you give me a plus one if you're with Laura and you're like, heck yeah, I want to figure out how to be using AI in workplace investigations and I'm driving this. And can you give me a negative one if you're like, no, if it was up to me, we would not do it. But my CEO or the leadership team or someone is pressuring me.

Okay. I'm going to let these come in. And while that happens, I want to acknowledge that I think this is a good thing that we are being, let me reframe.

I think that HR is often seen as risk averse and that is often a good thing. And sometimes it can also be limiting. And I think in this case, it really is about balance and it is good to be cautious, to be reasonably thoughtful about how we are leveraging AI in a reasonably high risk context and situation.

So I think that's a good thing. And I don't think this is the place for it to be like, oh, AI, you're so risk averse, you're slowing things down. You know what?

If we're going to slow things down somewhere, I think this is probably the right place to do it. Chantelle, I have a question for you because you and Rabih were talking about this on the prep call. I had the impression that a lot of people were getting pressured from CEOs to use AI and maybe necessarily weren't doing this for themselves.

It seems like that's not the case based on what we're seeing. It seems very few people gave a negative one here. Are you surprised by this?

I'm not surprised. I see, you know, AI is taking over a lot of HR functions and where it seems to be kind of focused right now is those tasks that seem more rote. So for example, you know, approving somebody for sick time or things of that nature.

This is a very nuanced part of an HR individual's practice. So I think this is going to be kind of a later frontier. And so I'm not surprised that people aren't getting a ton of pressure on it.

Where I suspect they may see the pressure is like, do we really have to have you do all the interviews? That's where I think is probably going to be the next step, especially with intake. So let's jump into that.

I remember when the first person flagged to me AI in workplace investigations. It might have even been you, Chantelle, that had mentioned like this is happening. And I was like, what?

It hadn't even occurred to me. Like the entire concept felt ludicrous to me because I was like, well, immediately I'm going to lose like lawyer privilege, right? Because everything I always do privilege and confidential.

I put my lawyer, our in-house lawyer on everything. I'm like, nobody can see this stuff. I was like, you can't do that with chat GPT.

Like what are you talking about? So my question to you is Chantelle, really? Is this really happening?

And if so, where are you starting to see people leverage AI in investigations? Well, I think something that you just said that I want to put a kind of a little star next to is when we think about investigations, there's the immediate need to figure out what happened, right? And then there's also, hey, we need to preserve everything because three years from now, there may be a lawsuit and they're going to scrutinize everything we did to see whether or not it's defensible.

So where we're starting to see AI is not necessarily in the actual investigation itself in terms of how the investigator or the HR professional who's serving as the investigator is using it. We're seeing it a lot more in the summary of the notes and putting together a report at the end. This is where we're starting to see people put their toe in the water in terms of AI.

Interesting. Okay. So quick follow-up question, then I'll throw it to you.

What I'm not hearing is things like AI transcriber or note taker listening in during the investigation, or why don't you just have somebody like AI do the intake, like AI do the investigation itself? Like it seems like that could be automated. Ask questions, they answer, moving it along.

I'm not hearing that from you. Am I misunderstanding? No, you're not.

Actually, I want to touch on the note taker piece very specifically, because there's a very kind of changing dynamic that is happening. It's very common now to have meetings and the meetings are recorded. It's a much different dynamic than even we had just a year ago.

The difference is that when you're doing an investigation, you really need to have thought go into what is your process? The vast majority of investigations, the processes, this is not going to be recorded in any capacity other than the notes that the investigator or perhaps another human joins and takes notes. You need to know that before you start an interview, because I've certainly had situations where somebody's otter AI joins the chat and I have to say, I'm sorry, this thought cannot join us because this isn't going to be recorded.

The other nuance here is what are the laws in the particular state you're in? In some states, they can use it and they don't need your permission. In other states like where we're BNI are in California, we need everybody to consent on the call.

I will just say from a practical note, where we're also seeing, I actually just did an investigation where it was required by law to have it be recorded. The transcript that was created through AI was inaccurate. What was interesting is that it was inaccurate based on who was talking.

It had pretty much the right words, but it said all the wrong people were saying the word. This is the other piece where we just have to wait for technology to catch up. Okay.

Very helpful. I promise for anyone who's tuning in being like, Melanie, get more specific. How do we know which tools to use?

We will in a second. I'm just curious to know what you all are seeing in the general landscape. Rubi, I'm curious from your perspective, same as Chantelle or different?

Are you seeing people using AI in different areas, whether you agree or disagree? What tools maybe are you seeing them use?

[Rabi David] (17:29 - 18:20)
It'd be same as Chantelle. Obviously, we work together, so we collaborate and we have a similar experience. That's no surprise.

We've seen people use them with interview notes to summarize at the end, summarize outlines for final reports or executive summaries. I've also had the experience of AI failing where I've tried to use it for summarizing witness interviews at the very end. Maybe once it gets it right and you dial it in, you have all the algorithm set up.

You say, okay, I want this information and impartial and everything else. Then the next one, it gets it wrong. In the end, people are using it to help them, but there's still a lot of fact checking that has to take place.

I'm sure we'll get into that. That's what I'm seeing.

[Melanie Naranjo] (18:20 - 20:32)
What I'm hearing is just straight up replacing an interviewer with an AI intake collector, maybe not the best approach here. Well, go ahead. I would also add to that because we have to remember that intake interview is actually one of the most important interviews that you take because that's what defines the scope of the investigation.

We don't want a situation where the claimant comes back and says, that's not what I complained about. I can certainly see a situation where we have AI in partnership, but where AI is right now, it's just not advanced enough in order to replace that human function. Let's get into some of the specifics.

We've heard from the people, they want to use AI. They just want to use it responsibly. As you all think about how to leverage AI, I'm hearing from you, even as lawyers, you too are pro AI.

It's just pro thoughtful and responsible AI. How do you go about determining where it makes sense, where it might make sense to use AI and where maybe it would be better off avoided? Chantelle, I'll start with you.

Well, I think about other parts of your job, where you're like, how can I leverage AI? It's usually things that are labor intensive tasks that have a good outcome, but it's not a good use of your brain. An example would be, I'm going to put in this data and then can you make me a chart?

For example, if you're doing a pay equity analysis, it's not going to be the same as if you're hiring an economist, but saying, okay, here's all the pay data for these key people that the allegations are about. Can you identify for me some trends? And then looking at that to help you jumpstart your synthesis.

It's really about getting rid of the tasks that are while helpful and help you organize your thoughts. At the same time, don't replace that human nuance and analysis that is so essential for an investigation. Got it.

Okay. Super helpful. Rabih, what's going on in your head?

[Rabi David] (20:32 - 23:19)
Yeah, I think it's important to use AI. I think this is a situation where this is like email, this is like internet at a much more grandiose level where those who don't use AI in their work are going to be left behind. And so it's a matter of figuring out, well, how do we use it to further our efficiency without compromising the end result and the accuracy of the investigation?

And so that goes into feeling comfortable with whatever you put in it. Because realistically, we can come up with some general uses right now. Six months from now, we're talking about complete other ways to use it.

This is a moving target. This is a rocket ship taking off and we have to just be willing to adapt and figure out how to use it. So I think something really important that everyone on this webinar should do is take a little time and do some research on how you can lock down your AI, whether it's ChatGPT, Claude, Grock, Copilot, Gemini, whatever you use, there are ways to lock it down.

With ChatGPT, for example, you use an enterprise account. So if you're using a free account because you're trying to save money for the company, the chances of that being secure, that's no bueno, right? That material being sent out is going to be trained on and used and it's compromised.

So you want a closed sandbox that you can play in. And buzzwords that you're looking for are MS35, Azure, different Microsoft ecosystem, Watson. There's lots of different competing services that I'm guessing most of the people on the webinar have more information on than myself.

But it's really important that you lock it down because once you lock it down and you think of it as, okay, this is my own private building, and then within that, only HR and legal professionals have access to this particular room where this is my account, where I get to ask the questions on ChatGPT or whatever AI you're using, then you feel a lot more confident. Then you're like, okay, it's like an intranet server. And then use the AI how you want, knowing that it's locked down and determine, hey, what works?

Because I know that witness summaries don't work very well right now, but six months from now, they probably will work at an exponentially amount better. So it's really about getting comfortable with knowing that you have the ability to play with and experiment what AI can do for you. And then we can still talk about how there's limitations and how you work around those, but having that comfort level is what's going to really make it a value add for you instead of something that is just a burden in your day-to-day affairs.

[Melanie Naranjo] (23:20 - 27:25)
Okay. I'm going to jump in with some very in the weeds questions here because I'm an HR person, so I imagine if I'm thinking this, other people are thinking this too. My first question is, what about privilege?

Privilege and confidential stuff? If I put something into ChatGPT, is that no longer protected by lawyer privilege? Does that mean that if we were to get sued, everything that I ever wrote in that ChatGPT conversation is now subject to scrutiny?

So to Ruby's point, this is why the levers that you're pulling on the back end with AI are so essential. If you're throwing it into an AI platform like ChatGPT or Cloud that's just an open platform, even if you're paying for it yourself personally, it may not have the guard rails. So that means you've now disclosed this information and where that's important is not just for privilege, but you may have in your policy that you're promising this to be as confidential as possible.

So it's a balancing act. So I'm sure many people on this webinar have cloud-based systems that you use. So it's like you have a payroll provider.

You're not supposed to disclose pay information, but you work with a vendor in order to give you services. You got to think about AI in the same way. So for example, ChatGPT has an enterprise solution.

Many of them also do as well. So making sure you're partnering with your IT counterparts to ensure that all the training modules have been turned off and things of that nature. Additionally, the other thing to check in with IT is whether or not the system that you have currently, whether there's a risk that when there's an update, the things that you have toggled to maintain privacy could be released.

And so you have to go back with each update and double check it. And frankly, the technology is different in that regard. So ChatGPT enterprise, pretty good at it being rock solid.

And there's an internal monitoring system you don't have to retoggle. Grammarly, they update it and you may need to retoggle it. Yeah.

Okay. And then let me just clarify because first of all, very good reminder that you want to partner with IT to make sure that the settings are correct and that they don't auto change when there's an update. But I actually am talking about, let's assume you have an enterprise account, right?

I have been trained, maybe incorrectly, tell me if I'm wrong here. I have been trained that even in a private Google Doc, if you do not write privileged and confidential and share it with your lawyer, it is not privileged and confidential. It could be forced to provide that information in the event of a lawsuit.

So my question is, even if you have an enterprise, totally private ChatGPT, I have no way of adding my lawyer to that and saying privileged and confidential. Is that a concern or can I use ChatGPT? So now I understand your question.

So the analysis is, so think about, let's actually use your Google Doc example as a solution, right? There is a way for you to share that with your lawyer and say, I'm getting legal advice from you as part of this process. And because not just because you put the words privileged and confidential and not just because you sent it to your lawyer, but you're actually getting them to weigh in on it.

So I think what we need to do is make sure that we have other platforms where we are memorializing that, for example, I brought in the attorneys. So in that Google Doc that you've created to say, attorney, I'm going to be running this through ChatGPT to create a chart. Once it's been created, I'm reposting it here.

And the attorney is like, great, great. Okay, that sounds like a great idea. You then have direction from counsel and your argument that this is being conducted under a privileged and confidential investigation at the direction of counsel is solidified.

You just may need more than one evidence source to prove it. Got it. Okay.

So I feel like what I'm hearing... Go ahead, Ruby, please.

[Rabi David] (27:26 - 28:17)
I would just add that I feel it's similar to just writing a Microsoft Word document. Like if you just write a document, yeah, maybe it will be discoverable. You don't have attorney-client privilege on that document.

Maybe that document, though, during the discovery process is not subject to discovery because it's not relevant or it ends up being overbroad or whatever it is. So just because, yes, there's a chance that what you're typing into ChatGPT is not privileged and it is discoverable, but is it something that's going to move the needle in such a way? And if it is, that's something that probably does need to be shared with the attorney to protect it under that privilege umbrella.

It's probably not practical to put the lawyer under all of ChatGPT because it defeats the whole process. And it's the sage old idea of, oh, we'll just CC the lawyer and everything, and then everything is confidential. And that's not how it works either.

[Melanie Naranjo] (28:17 - 29:11)
This is actually touching on something that I actually find really fascinating the further along I get in my HR career, which is this concept of risk tolerance and just like intentionality. For me, what I hear when you all say this as the underlying message is, make sure you are having intentional conversations with your lawyers, with your IT department to make sure that whatever process you put in place and wherever you choose to use AI makes sense and you've thought about the potential risk and you are okay with that level of risk because everything comes with risk. It is impossible to 100% mitigate risk.

But if you can have a clear and thoughtful discussion around what process and what risk tolerance you are comfortable with, then you can proceed in these ways that will ultimately save you time and potentially it sounds like strategically the correct direction. Does that feel accurate?

[Rabi David] (29:12 - 29:31)
Yeah, it does. And don't forget, you can focus on the role-based access. You can say, hey, I'm just limiting this ChatGPT or this AI, this Gemini account just to these four people or whatever it is.

It's not going to disseminate to the rest of the company in the first place. So there's ways to maintain that information.

[Melanie Naranjo] (29:31 - 33:56)
Another nitpicky question before I get a little more thinky-thoughty. My nitpicky question is, do you have to disclose to employees where in the investigation process you are using AI, even if it's just to summarize information afterwards? I would say as a general matter, no.

Your responsibility when conducting an investigation is to review all the information that has been presented and not to make any credibility determinations or to make any conclusions until the end. You do not have to give them all the details about how you are then analyzing that information. Likewise, I mean, there's other ways that we're thinking about AI tools.

So for example, using an AI as an assistant to say, okay, these are all the people that I've talked to. These are the people that I've called. Can you create a chart for me?

Or can you make me a list of everyone I haven't called and their phone number so it's easier for me to do? That's really about the functionality. Think about what you do now.

You don't tell the employee, well, we used Zoom for some people and Teams for others. You don't give them the details about that logistics. So it follows through here that you don't have to disclose it.

Where you do have to disclose it is if you're doing something where there's a legal piece of puzzle. So for example, if you're using AI like Otter, for example, to transcribe the notes, that's a recording. And you need to disclose that to the participant depending on what state you're in and let them know and get their consent to being recorded.

What I really love about that is this constant reminder I've been getting that AI is just another tool, like any other tool. You were talking about this with Microsoft Word, right? These have always been in place.

AI is just a new tool. And so if you kind of follow the logic of, hey dudes, like you never had to disclose which, if you were using Google Docs or Microsoft, if you never had to disclose those things, the likelihood is, as long as you have all the right settings and protections in place, you can think of AI relatively in a similar way. It's just another tool.

You don't have to disclose every step of the process. Go ahead. Right.

And I would also say it depends on how you use it. So one of the ways that we're seeing AI in investigations is actually people giving us evidence that may be manipulated. So they're like, oh, can you give me screenshots of these text messages?

Can you give me, oh, is the social media, can you take a screenshot of that? And, you know, I came across a situation where I'm like, this is not, this, something about it isn't quite right. Like it was a social media post.

The position was a little bit off. I can see a situation where someone uses AI to test an image that they get. Say like, does this look like it's AI generated or does this look like a real screenshot?

And that may be a situation where you want to disclose that you used AI to find those determinations because it goes to the reasonableness of your findings. Okay. Let me, let, let me add on to that.

One use case I could see for AI, I'd love to hear from Rabih and Chantelle, whether you think this is good or risky is, hey, I asked all these questions. What questions didn't I ask? Where are some gaps that maybe I am overlooking to help me connect the dots based on what I'm currently seeing right now or the set of information that I have collected up to this point?

I personally have the impression that that could help me just, as long as I'm still using my brain, I'm not just whatever it says I do. As long as I'm using good judgment, I personally think that could help close some gaps for me, spark some new ideas. I could also see the argument that that could lead to some level of bias.

What if it encourages me to ask more questions of some people and not others? What if it asks, encourages me to ask questions that are leading questions? And so my question to you is, does that feel like a reasonable use case of AI or is it, as with everything, the context matters and as long as you do it responsibly, yes?

Like what, what goes through your minds?

[Rabi David] (33:58 - 35:09)
I, I don't see any problem with that whatsoever. I don't even know if it needs that much context. That's the same thing as taking like a treatise and looking at it and saying, hey, what are the best top 30 questions you ask in workplace investigations?

And then comparing notes and going, oh, missed this one, missed that one. AI is just automating that for you. So that doesn't sound like there's that much, much wiggle room or even much gray area to speak of.

That seems like an, an ideal way to use AI in investigations and also the way I've personally used it. I use AI to frame my outline before I interview witnesses and say, hey, here's the general idea, right? I've got all the right parameters in place.

I've locked down AI and said, what are the best ways to approach this, this, this particular interview? These are the different areas I'm covering. What, how do I go about this?

And, and then it tells me its ideas and then I create my own outline based on it. It's, it's just like, that seems to be the way you should be using it while completely, like you said, Melanie, don't overly rely on it. We'll get into that, but use it with your brain.

[Melanie Naranjo] (35:10 - 40:01)
Yeah. And I would add to that, that it's another way too, if you address kind of a, a gap in your own training. So for example, to say, hey, you know, I am the, I know I'm going into an interview with somebody who perhaps could have experienced some prior trauma.

Let's say it's a very sensitive, you know, sexual harassment case. You're how, putting in your outline and asking AI, do you have suggestions on how I should change the wording of any of these questions to make them more open-ended, to make them, make sure that I'm adopting a trauma-informed approach. And it's in essence, as you said earlier, Melanie, sometimes being in HR can be lonely.

And this is a way for you to have a team member. Now we all know that we, when we work in teams, we don't take everybody's idea. We, you know, take the meat, leave the bones.

And that's exactly what you can do with AI in your corner, as long as you're doing it in a controlled platform. So you're not worried about the information being disclosed. I love that just because I think it opens up this whole new door of leaning on AI as, I'm going to be careful the way I say this, as a thought partner, but more as bouncing off ideas.

Sort of, it's, give me a plus one in the comments if you've ever heard of the rubber duck method, where you talk at a rubber duck, you just talk aloud. And just the act of talking aloud helps you think through your thoughts, recognize like, oh, I maybe will run into issues there. I didn't word that correctly.

Okay. Interesting. Not a lot of people know the rubber duck method.

I know. I just talked to myself in the mirror. Of the mirror method.

Yeah. There you go. But it's sort of like the rubber duck method AI, right?

It's the next level where you're sort of thinking aloud and saying, you know what? Last time I did an employee investigation, I got some feedback that it didn't come across as empathetic and it came across as too lawyery. How can I word these questions in a more empathetic way?

What kind of introduction would helpfully communicate or effectively communicate the goal of this interview while preserving a level of empathy for the employee to make them more receptive and comfortable as they share their thoughts and responses, right? Like I can see it just opening this new door. None of this is confidential information.

It's just sparking new ideas. So, okay. Let's summarize because I want to make sure people have actionable things.

So, what I'm hearing is one good way to use AI could be, hey, I've got some tedious stuff that I need to summarize or turn into a chart that I can then share with the appropriate people. Second is, hey, I could be using it to help identify gaps in my questions, in my thinking, that sort of thing. And then the third one could be leaning on it in terms of helping it set things up more effectively.

So, not just retrospectively, I did these things. What questions have I missed? But proactively, what questions should I ask?

How can I frame them more empathetically? What things should I be thinking about as I head into this? Are there any other things that come to mind where potentially AI could be a good use case?

Go ahead. I just thought of this after you went through the list. Yeah.

So, I think that another way to really think about is like, what is your personal interview style? So, maybe to prepare, you want to write a bunch of nitty-gritty questions, but then when you're actually in the moment, you're that person that needs to be not tied to the piece of paper and you really just need 10 bullet points to focus you. This is a way for you to say to ChatGPT or any other AI system that you have in a controlled enterprise to say, hey, can you take my outline and summarize it into the high-level points?

That's also a good way, too, for you to cross-check that your outline is complete because it's in making it shorter and in summarizing it that you realize, oops, I forgot that whole part about the retaliation complaint. It's a way for you to use it as a cross-check. Yep.

Love that. And plus, one to your comment, Laura, I don't think this is a replacement. I just think that it sounds like this could be just another pulse check to help round out the process.

Yep. Okay. So, all, in these last few minutes before we pivot, I would love to ask a question.

You touched on this, Chantelle, on, hey, are employees maybe using AI in ways that are potentially problematic, whether it be falsifying information or secretly taking notes that we're not consented to in the process. Rabih and Chantelle, what are you all seeing? What are the risks that people need to be aware of?

And how can companies most effectively go about trying to mitigate these risks?

[Rabi David] (40:03 - 41:25)
Yeah. Well, we are definitely seeing that employees more often, and it's probably going to just be standard moving forward, are using AI to frame their complaints. Their complaints look like they're written by lawyers, and they are people who don't have legal education or training.

You can tell when you start to talk to them. And then at least in my experience in investigations, I'll be interviewing the witness and the complainant, and the complainant will say, oh, that was AI. I didn't mean it that way.

It's like, wait, you didn't mean racial discrimination? You just meant that person was just like mean to you that day, and they're not even a different race? They're like, oh, yeah, AI just kind of took it to another level.

But because they're just relying on it so heavily, and because it makes it sound better for them than what they would have come up with themselves, they leave it in. And you, as the investigator, now have to tease out, hey, what is actually at issue here? And what needs to be removed from the complaint?

Because they have incorporated AI into doing it. And so that's one of the ways that employees have used AI in a way that disadvantages us as investigators. And then we can talk about how we should use it as investigators, but I wanted to get Shantel's take as well.

[Melanie Naranjo] (41:26 - 43:49)
So I think the other thing, just to put a point on what Rubi just said, is, and this will be applied to the old school complaints as well, is saying, like, all right, you put a bunch of words on the page. What do they all mean in your own words? So like you said, I mean, we see this all the time, like, I feel harassed, I feel discriminated.

I mean, like, what does that mean to you? Like, help me unpack that. And we now know that not only does there, you know, there's just this colloquial way that these words have been used, but we now know we have a direct experience that people are relying on chat GPT.

And they're saying, like, oh, I didn't mean it like that. Like, I just meant this. And it's helping you parse out exactly what they mean.

Frankly, this is that point where these are things that you were already doing. This is not new or different. This is just how the person created the document and complaint is different, but how you respond is the same.

Yeah. I really loved, you had given an example early on about updating policies to proactively address some of these things. For example, saying, hey, you can use AI for this, you cannot use AI for that.

Tell me a little bit more about that, Chantelle. So we're seeing a lot of that where AI is being specifically pulled out. And in particular, where we're seeing it is companies are starting to take like a greater control over AI and say, like, these are the platforms that you're allowed to use.

Like, if you're going to use anything else, you have to ask us in advance. And then on the flip side, when we think about that in terms of not just doing the investigation, but getting information, this is where that part two, for HR professionals that are thinking about policy violations at the end, I encourage you to think outside the box and ask yourself, is this person just violating, for example, the harassment and discrimination policy, or are they also violating our code of conduct?

Are they also violating our privilege policy? Are they also violating our unique AI policies? Really thinking about the policies that have been violated is also a different way that we have to think, and that could be expanded in light of how witnesses and claimants are using AI.

Yep. Love that. Rabi, anything you want to add there?

[Rabi David] (43:50 - 46:51)
No, I think that completely makes sense. Using AI, updating the policies, and holding the employees accountable with AI in mind. I would want to add before we start to the question and answer, we've talked a lot about the dos of AI, but we should also balance that a little with the don'ts, right?

And there's a lot of things that may be taken for granted, but obviously we're not using AI to generate findings, to say, okay, here's all of this. What do you think, computer? Is it more likely than not that this happened as a result, right?

Because it's that kind of thing that I think you were suggesting, Melanie, that you have heard CEOs are pressuring HR professionals to just cut out the middleman and just go straight to AI to just run the investigation. And we're not there, and obviously we may have a little bias there, but we don't think we're going to get there. The human connection is not replaceable.

And just as an example, I think a lot of it has to do with the paradigm shift of thinking of AI as a magic eight ball or a calculator where you type in the question and it answers it 100%, right? Like a two plus two equals four kind of thing. That's not what AI does.

AI is more like what Chantelle said, it's akin to being an assistant, a really smart assistant, a Harvard educated assistant, but an assistant that only relies on whatever you input. So if you have a slant on how you're putting in the facts and saying this person did this, and I didn't like how they did that, whatever, like, what do you think? Do you think they did it right or wrong?

AI is going to side with you. It's sycophantic in nature. And it says, hey, I agree with you unless you put in the algorithm and say, hey, make sure you challenge me.

It's going to say, here's all the reasons why, and it'll cherry pick. And then one last thing I'll say is AI is framed to give you a solution. It does not like sitting in the unknown.

So you have all probably experienced this. I experienced it even on my day to day, if I'm remembering a movie or something from the seventies, I'm like, hey, what was that movie where this guy did this? And it was Clint Eastwood, and then AI will say this entire blurb with all these facts based on the actor was this, and this happened, and it opened up this.

And then you'll look and go, that's not true. And then you'll say, no, you're wrong. And it's like, oh, my mistake.

And then it'll say a whole nother set of facts. And it does that a couple of times until you finally ping it enough. And it says, I cry uncle.

I really don't know. Sorry. But it doesn't do that right away.

So you're dealing with that expectation. AI will give you a beautiful, eloquent, poised response that is not accurate necessarily based on your current facts that if you over-rely on and you're not the keeper of the facts yourselves, you're setting yourself up for a fall in investigations.

[Melanie Naranjo] (46:53 - 58:56)
I keep thinking about, for those of us out there that have teens, the very confident answer with that teenager, it's like, well, yes, obviously, this is the answer. And your job is to be the parent and say, you said that with a lot of confidence, but this isn't quite right. What I love about this whole messaging is, if I'm hearing this correctly, AI is the future.

We are going to be seeing more and more AI incorporated into this process. We just have to be really thoughtful about the balance of what is thoughtful AI usage, what is irresponsible AI usage, and how do we go about training our employees on this, training ourselves on this to make sure that we can tease apart the right things. With that said, I'm going to briefly pivot into quick Athena Spotlight because this is something we care quite a bit about.

We are incorporating AI into compliance training and to the entire compliance process because we believe this is the future. We believe that we need to educate people on how to do this in the right, responsible way, not irresponsible way, and that we need to be addressing real problems, which are, to your point, Chantelle, the tedious nature of it. How do we make things a little bit simpler and how do we do it in a way that doesn't send people in the wrong direction?

I'll share with you all briefly and then we'll turn it back over to Q\&A. Let me know if you all can see my screen now. Can you all see this?

Okay, perfect. Here at Athena, we have created an AI-driven policy bot. Any of our customers have access to this.

What happens is you can upload any of your policies and use this as a place where your employees can go to and ask questions. The reason why I'm bringing this in with the theme of investigations and AI is because, one, we want to help you all leverage AI responsibly to help do your job more effectively and because we fundamentally believe that critical to mitigating some of these risks, even though you can't 100% eliminate risk, is to create a speak-up culture where employees understand, similar to what I can and can't use AI for, which tools I can and can't use, what policies exist? Is this problematic behavior? Can I say something about this?

Who should I go to? As an example, if you had an employee who was nervously sitting on a question around harassment but didn't want to ask HR because, as soon as I say something, it's going to turn into something, we fundamentally believe we'd prefer that an employee have access to say something or ask something anonymously than not ask it at all, and then we never learn the thing or give them the helpful information that could point them in the right direction. If you would put in, do we have a harassment policy? This takes a second.

It's connected through Chachapiti, and it's buffering because we're screen sharing with hundreds of people on this call, but it would give you a response, right? Yes, we have a policy. Here's what it is, right?

Once you uploaded it, it would link you to that doc, right? Maybe because dating has been quite a bit in the news lately, you might have an employee who asks a question like, do we even have a dating policy? What is her dating policy?

Same thing would happen here, right? You would get a response once it generates that tells them, yeah, we do have a dating policy. Here's what it is.

Now, two last things I'll say, just because I know this question comes up a lot and I think it's quite helpful. Something that I also really like about this is it only pulls from policies that you feed it, right? It's never going to pull from the internet.

It's not going to hallucinate or make things up. It only pulls from information that you give it. So, let's say in this demo example, I didn't feed it an AI policy, okay?

What I like about this is, let's say you don't have a policy on something. It's going to say, hey, I couldn't find any information. It's not going to make up information.

But here's what I love about this is through this process, you might learn about policies that you don't have that you should have, because your employees will ask the question, see there's not a policy, and then they will realize like, we should have a policy though, and they might reach out to HR, and that might then surface new things that you need to do. Obviously, this is my last note on this, and then we can move it along. People have also asked like, well, this is very cool, but can it also help me offload the very common peril and benefits questions, how to access my pay stub?

So, I will give you a sample one and say that, yes, this can also help you with that. Truly any policies, any processes, any docs that you upload, it can pull for information. So, if somebody asked, how do I change my benefits enrollments?

Nobody wants to answer that question 25 times a day. And so, you could have employees more easily access this information themselves by uploading the information. It's pulling from Athena's one, where we tell people, hey, outside of a QLE, can't do it, only open enrollment.

If you do have a QLE, here's the process. We use Namely, so here's how you do it. And it really walks them through step-by-step, and I didn't have to lift a finger to do any of this.

So, it's quite nice. I will stop sharing. If anybody has questions, a couple things to know is, this is actually accessible for free, this template.

So, if you want to play around with it, we'll link you to, in the follow-up email, the demo. It's preloaded with a few Athena sample templates, just so you can play around with it. You can try and break it.

It won't break, but you can try. Give us any feedback, just to see how it works. And again, if you're an existing Athena customer, you can start using this.

I think we saw something like 30% utilization in the first two weeks across our customers. They all just wanted to turn this on immediately, just because it saved them so much time. And then, we will also be sharing a template of our AI policy, so that if you don't have an AI policy yet, you can have that as a starting point.

Okay. With that said all, I've got a couple of things that I need to share with you. For anyone who is tuning in for the HRCI or SHRM credits, I'm going to screen share now.

And then, we are also, at the same time, going to put up a poll. If you could just take 10 seconds to answer these questions, and then we'll kick it over to Q\&A. Okay, all.

I'm going to go ahead and kick it over to Q\&A now. Let's pull and see what questions came in from the audience. Okay, let's see.

Melanie, as you're looking, can I make a quick comment about the Athena bot? Yeah, please. So, I really want you all to think about the bots that you are incorporating in your workplace as potential witnesses.

So, for example, you have somebody who has a leave claim, and they claim that they've asked their boss a million times, and they never got any information. Go make sure that you're checking in with your AI bots, the other infrastructure that you have at your company to see, did this person actually get this information? Have they been asking all these questions?

This is a rich source of information. It also can help you understand people's perceptions. Because, for example, what if your boss said, no, you can't take that vacation.

We've got, you know, a really big conference coming, and you have to go to that. And then you go to the time off bot, and the time off bot says, you have plenty of time in your bank. It's approved, and don't worry about it.

And helping you kind of understand how people got to a place of conflict. So, don't forget that as an important source of information to mine when doing your investigations. Oh, I love that.

That came up when we first rolled this out, Chantelle, and I'll answer Michelle's question directly here. Do you recommend having a disclaimer on the bot? We have a disclaimer on the bot, absolutely.

And our disclaimer is that we are not actively monitoring this, or the admin is not expected to effectively monitor this, and the information is actually not us. And so, I say that not to contradict. I actually think the two can work hand in hand.

Here's sort of like our stance on this, is an employee could put a question into chat GPT, and I wouldn't know that they put that question in there, right, because I'm not expected reasonably to monitor their chat GPT questions. And so, an employee could put in a question into the chat bot, and I wouldn't reasonably know, because we want this to be like a safe space where people can ask questions. Now, if, however, Chantelle, you as a company wanted to change that, right, like you wanted to say, disclaimer, we are monitoring these, or disclaimer, we do reserve the right to go back and track questions that have been submitted.

That, in my opinion, would be a different disclaimer that you would put into place. Potentially, you could create your own chat GPT, or your own like custom GPT around this. And so, I think, Chantelle, what really resonates with me is that like this needs to be intentional.

Like you need to figure out what your policy is around this. Are you proactively going to be monitoring this, or do you plan to go back and monitor responses? And if so, you should disclaim it.

If you don't, also, you should disclaim it so that employees know like what is the right usage of it. I can just imagine an employee thinking like, well, I put it in the bot. Shouldn't you have followed up with me when I put it in the bot?

And so, absolutely, there should be a disclaimer. And ours does have a disclaimer, just in case anybody's curious. Okay, cool.

We have just a couple minutes. I'll ask just the one question, just one question here. It was actually for you, Chantelle.

So, you mentioned AI can be used for charts during investigations. What sorts of chart did you have in mind when you mentioned that? And what are some helpful charts to create during an investigation that AI could help with?

So, there is. So, I love a chart. I'll just start by saying that I love a chart.

So, it's really about, once again, like your assistant. Imagine you had an assistant. What would you do?

So, I do a lot of investigations where people are working remotely and they work all over the country. So, I could say, okay, here's all my witnesses. Here's all my time frames.

Can you make a chart for me that everybody that's on East Coast time zone, everybody that's on the West Coast time zone, or not a chart, but can you make a schedule for me for if we're guessing it's going to be about two hours per interview, can you make a schedule for me with at least hour blocks between here and here and here, like helping you do those more kind of tedious tasks? And really, it's as simple as you feeding the information that you want in the chart and then telling them to organize it in a way. So, I want you to put this in a, you know, here's all their names, here's all their phone numbers, and put this in a chart for me so that with check marks on it so that I can check them off once I've contacted them.

I love that. I want to be respectful of time. Chantelle, Ruby, you all have been amazing.

Any final words for the audience before we leave them all today? Well, I would say let's just not forget the human side of this. In all things of AI that we're seeing now in every aspect of HR, the key is the human element.

These are amazing tools, but you are the smartest person in the room. And, you know, take stock in that and make sure that you are leveraging these AI tools so that they are a help to you. And if at any point they are a burden, speak up.

And so that hopefully the organization can work through it. Thank you so much, Chantelle. Ruby, any final words?

[Rabi David] (58:57 - 59:22)
No, I second that. Don't over rely on it. Use it as a tool.

But work with your administration. Don't do things in secret and behind closed doors just because you think AI is working for you and your company is worried. You're never going to go very far that way.

But I really appreciate it. And thank you so much, Melanie. It was a pleasure.

Thank you for having me.

[Melanie Naranjo] (59:22 - 59:49)
Thank you. I always love learning from the experts. And I will say, Ruby and Chantelle work for an actual law firm.

So if you are looking for advice, actual legal advice, and you liked what you heard today, please feel free to follow up with them and partner with them directly. Big fans over here at Athena. We partner with them.

So highly recommend. All right. Thanks so much.

Bye.