S6 E6: Human in the Loop: Using AI in HR Without Losing the “Human”

Podcast August 20, 2025

Pete Wright
Welcome to Human Solutions, simplifying HR for people who love HR from AIMHR Solutions on TrueStoryFM. I'm Pete Wright. We are glad you are here. Today, we're talking about AI in HR, what it's doing well, where it's going off the rails, and why the smartest companies are figuring out their AI policies now, not after the headlines hit. From performance reviews drafted by ChatGPT to job postings fine-tuned in seconds, AI is very much shaping the way HR teams work. But what happens when the chatbot gives bad advice or an algorithm bakes bias into your hiring process? Joining me today once again are our own Kyle Pardo and Terry Cook. They are watching firsthand how AI tools are showing up in everyday HR workflows. And they're here to unpack the legal, ethical, and don't-try-this-at-home side of AI adoption. Together, we're going to dig into the good, the bad, and the real world steps you can take to make the AI help, not a hazard at work. All right, Terry, Kyle, welcome back.

Terry Cook
Thank you.

Pete Wright
We have a minefield of a conversation today, I fear.

Kyle Pardo
Absolutely.

Pete Wright
How are you both using AI? Just in terms of a table setting, how are you using AI day to day? Do you find that you have a persistent AI chat companion, Kyle?

Kyle Pardo
I think all the time, using AI a lot. But I'm trying to put it into three buckets. When to use it, when to be cautious, and when to stay away. So if I think of it, if I'm going to put together an onboarding program, I'm absolutely going to use AI to help me create an onboarding checklist. Help me think about what I should put into my onboarding plan, talk about benefits, welcome employees, you know, kind of, you know, day to day things going on in the office. I'm going to put that into AI. I'm going to ask it to help me create a checklist, you know. The area where I'd say, I don't know if it's going to work or not. I'd be cautious about using it to explain employee benefits, for example, because I want to really make sure it's going to give me the right information. And it might, it might give me some good information, but, you know, I'm going to be cautious. I'm not going to use it for the personal touch. You know, I still, as you know, want the human and human resources. I want to give the employee a tour. I want to take them to lunch. I want to tell them about the company culture. So, you know, I kind of think of it in those three buckets of, you know, absolutely use it for the checklist. But think about it, how you're going to use it otherwise.

Pete Wright
I think that's a really good perspective. It is HR. It's not AIR.

Terry Cook
Thank you.

Kyle Pardo
Yeah.

Pete Wright
And I think there are very real concerns from AR pros.

Kyle Pardo
Yeah.

Pete Wright
Oh, we're in trouble. from HR pros who are who are looking at what many small to midsize businesses are doing and thinking, well, maybe we can outsource a lot of our HR process to AI. And, you know, I think I think I can safely say the reality is, from our perspective, be very careful. You know, there be monsters. Terry, how are you using AI? Bye.

Terry Cook
I think similar to Kyle, you know, I think anytime I'm thinking about something that I might be creating, you know, trying to look into something that I might be working on and I get stuck with And I'll say, gosh, I feel like I'm not saying this right. I'll tend to look to the AI and say, here's what I'm trying to say. Here's my information. Help me say it differently. You know, and then tweak it from there. I find it extremely useful to do things like that. You know, like Kyle, I'm always careful about some things. Like I work a lot with compliance. So I would say compliance is an area where AI cannot always be trusted. So you can tell it to do something and it comes back with the wrong answer. And sometimes it'll be me just trying to test it to say, you know, tell me this. And then it comes back and it's not the correct answer. So I'm always very cautious. And again, also agreeing with Kyle, when you're trying to talk to somebody or trying to do something with an interaction with a person, you don't want to sound like a robot. You don't want to sound like a computer. You want to sound like you and you want to sound like you're trying to develop a rapport. So I would say that's how I use it and don't use it.

Kyle Pardo
I was going to just add to that too, Pete. I mean, one of the ways that I'm using it is for performance reviews. You know, the studies show that people write performance reviews just based off of the past three months, just because you forget what happened earlier in the year. So I use it. I take notes throughout the year when I have a conversation with my employees, what's going well, what's not going well, what projects they're working on, things like that. And at the end of the year, put those notes into AI, not using individual names, not using company names, client names, keeping all of that out of it. So it's kind of a generic information. But asking AI to help consolidate and summarize that information for me, but absolutely has to be the human delivering the review, putting the personal touch on it, putting specific examples in there, things like that. So, again, AI is going to be helpful, but not to rely on it exclusively because you need to have that human conversation giving feedback.

Terry Cook
Продолжение следует...

Pete Wright
Before we started the conversation today, I gave you both an acronym that I believe was new to you. And you very falsely gave me credit for it because it's a ridiculous acronym, but I think it's important. It's HITL, H-I-T-L, human in the loop. And I think that comes into this conversation in probably every point that we're making, right? Kyle, your example about performance reviews, you are absolutely in the loop.

Kyle Pardo
Thank you.

Pete Wright
And you have the original notes that you can skim and make sure that all the points are accurate to your experience with that employee. Where else do you find those sort of day-to-day support tasks, the day-to-day sort of tool come into play where you find a good balance between your own sort of cognitive mastery and the AI?

Kyle Pardo
you know we do a lot of in-person training or virtual but instructor-led training you know you

Terry Cook
Продолжение следует...

Kyle Pardo
can use AI to do training you can use e-learning to do training but but it doesn't have the human touch of answering questions and using personal examples and getting conversation amongst participants and so I think just when you think about any of these airways use AI that that human touch adds another layer to it so it doesn't mean you can't use e-learning e-learning is great and has a certain purpose. But the human touch adds that next layer to really, you know, make it personal to the people in the course.

Pete Wright
Are you doing things like job postings, offer letters? Are you doing some of those more quotidian tasks with AI? Are you seeing that happen a lot?

Kyle Pardo
Yeah, I know what Terry answered. She, you know, manages our recruiting function and absolutely.

Terry Cook
definitely for job postings um i you know i think too many people try to work too hard to make a job hosting when you can really just start it up with ai as kyle mentioned you take your own information for the job posting plug it in so that you know without personal company information and then just let it start to generating generating that job posting and then you're you're you're dealing with more tweaking adjustments for job postings versus somebody that might spend hours in front of a computer trying to go line by line through a job posting. So I think definitely for job postings, I've used it, even job descriptions, there's portions of that that you can do as a general use for AI and then add your own specifics to. But it is about putting the right information in to AI to get the best result that you can at least use to tweak. I think that's the important part. If you're very generic and say, can you create a job posting? I'm sure it can. It doesn't have any clue what it's doing, but it will create something. But then the more you kind of dive into it and like Kyle said, putting your own information in there so that you can tell it what you're really looking for, then you get the good results. And it's definitely a time saver.

Kyle Pardo
I think we're seeing HR people also use it even for kind of, I'm going to say some company policies, meaning that if you want to write a dress code policy, you might want to put information in there and say, you know, here's our company culture, create, tell me what, you know, create, help me create a policy for casual Friday. We don't allow, you know, jean shorts, but we do allow this. You might get a really good policy, but be really careful if you're going to use it to create a, you know, earn sick time policy or something like that that's really compliance-based and, and, you know, thousands of types of policies you want to have really vetted or done by a professional. But, you know, to do a dress code policy, sure, it's going to give you some really good information.

Pete Wright
I think that's really important. And I also like the approach that both of you are discussing, because there's one thing where you type in, hey, create a dress code policy. It'll create a dress code policy. I would just like to offer for those who are exploring with it to remember that you can talk to it like it is a brainstorming partner. After it gives you a policy, say, look, That doesn't quite work. Now I want you to make sure you include something about jean shorts and I want you to include something about, you know, funny hats. And I want you to include something about open toed shoes. And then it'll give you a revision. And now you get to read that and you get to identify that. It does sort of offer an interesting conundrum. How do you verify all of this AI generated content without spending the same amount of time that you'd spend writing it yourself? Do you both run into that that challenge?

Kyle Pardo
Yeah, absolutely. Again, if it's compliance related, really taking a look, and Terry was talking about this earlier, of looking at the source. And, you know, we would typically, we were going to write a policy, we would spend a lot of time looking at the Department of Labor's website. We would spend a lot of time looking at the Massachusetts.gov website, you know, really looking at the official sources. So when you start to see things that are pulled from, you know, different blog postings or, you know, other resources, you know, it's something to consider. You know, I heard somebody explain AI as really being like an intern. And so an intern starts, they really don't know a lot, You have to take some time. They might come with a first draft and you say, well, that's pretty good. But next time I might add something else into it. And then, as you said, Pete, it really becomes an iterative process where you're going back and forth with an intern in teaching them. And at some point they're going to get better and better, but still not going to be quite the level of the department director or, you know, who's going to customize it for the company.

Pete Wright
Thank you. be testing almost every time you use it, right? Oh, interns, slow your roll. All right, let's talk

Kyle Pardo
Yeah.

Terry Cook
where you go

Kyle Pardo
Yeah.

Terry Cook
where you go

Kyle Pardo
Yeah. Yeah. Yeah.

Terry Cook
you

Pete Wright
then about, you know, we know AI can save time, but let's talk about where it has gone wrong, particularly in HR, right? We want to throw up some red flags of experiences that we've come across that indicate AI can be truly troublesome, the chatbot sort of initiatives. Kyle, you want to kick us off?

Kyle Pardo
Sure. So, so again, a chatbot is, is another intern, right? And so it is going to gather information, possibly from your website, possibly pulling it from other sites. So you can tell AI, I want you to look at this document. And so let's say you want to use AI to help summarize your employee benefits. You could put in the summary plan description, which is a really long, detailed document that we all get with our health insurance that explains what's covered and what's not covered. And typically, nobody reads this full document until you need it. And then you feverishly look through it to see what's covered or not. So you could upload that to AI and say, help summarize this for me. But again, they're going to keep talking about this caution. You know, if I'm looking to see if a chiropractor is covered or some type of medication is covered or something, I'd be really careful in terms of what AI is telling me. And I'd probably want to double check that with the insurance company. Or I'd want to go back and look in the document myself to see if I could find it actually written in the document. Because if I go ahead and I have surgery, assuming it's going to be covered and then find out it's not, and I think, oh, I based my decision on what the AI search told me, that would be a challenging situation. So something like that, I would really probably want to try to verify the information I'm getting back.

Pete Wright
Terry, I have a note here that you specifically want to talk about an Amazon AI recruiting scandal.

Terry Cook
Oh, that was some time ago, but I think it applies. I've read other cases since. But essentially, if somebody's looking at what can be a good employee for me, you can talk to you know, chat GPT or you can talk to AI in general about it. And sometimes it'll say, well, show us what your current, you know, your current employees are your best guess of what a future employee might look like. So in this particular case, they showed AI all of their current employees who happened to have been male. And so anytime AI started looking at future resumes, the company was thinking they would pull like experience, things like that. But because they were male, anything that had any relationship mention of women, women's college, women's activities, women's groups, it was automatically eliminating the women from that, from that grouping. And again, certainly inadvertent, it wasn't like Amazon was telling it, please don't show me any female applicants. But I mean, it's happened a lot. You know, I have heard a lot of cases. I went through a diversity recruiting program and they just said that you, you as a person

Kyle Pardo
Продолжение следует...

Terry Cook
can inadvertently teach AI the wrong things. You're not going to try to be biased, but if you tell AI, here's the type of candidates that look like the right candidates for our company, they're going to take that literally and they're going to look at those candidates and think that they have to compare any future applicants to the candidates. So again, is it an example where a company wants to do something wrong? Of course not. They're just inadvertently happening because AI is reading into it the wrong way.

Pete Wright
because your intern is incredibly impressionable and also quite sycophantic.

Terry Cook
that's right there you go yeah oh sure yeah well that just happens pete yeah yeah that's right

Kyle Pardo
Yeah.

Pete Wright
And it will tell you you're right and so, so wise, Terry, all the time.

Kyle Pardo
Yeah.

Pete Wright
This policy is, it's the best policy we've ever seen.

Kyle Pardo
Yeah. Yeah.

Pete Wright
Yeah, right.

Terry Cook
absolutely yeah

Kyle Pardo
Yeah.

Pete Wright
Can we do a quick sidebar on specific tools? We're using AI in kind of a generic capacity right now. And I'd like to talk a little bit about the big tools, because they all have different sort of strengths. Do you, what do you guys use right now? And what do you like about it? Terry, you want to start?

Terry Cook
I find myself going towards chat GPT a lot. I mean, it just seems like it's something maybe I'm getting comfortable with. Um, certainly there's others. Um, I know Claude, I've used that on occasion. Um, there's one that's escaping me that I've used in training before that was pretty reliable, um, to help me start generating training programs. Um, but I do, I think probably because of, again, comfort level, I tend to go right back to chat GPT. I don't know about you, Kyle.

Kyle Pardo
Yeah, the same two. And then the third one I would add is Gamma, which I've used for presentations. And so it, you know, can take an outline, which maybe create the outline in a chat GPT, and then taking that outline and put it in gamma, and it's created a really nice presentation, which then you can customize with your own, you know, company colors and logo and all those types of things. But it'll help create images on there and, you know, kind of give you something pretty quickly that you can use.

Pete Wright
Okay. So ChatGPT is from OpenAI, the company OpenAI, and it is available at ChatGPT.com. As we record this, I mean, two days ago, they released the latest iteration of their model, ChatGPT5. Ergo, you should be testing it around the things that you used to use with ChatGPT4 because it's different. It does things different. It thinks, I'm saying that in heavy air quotes, it thinks differently.

Kyle Pardo
Thank you.

Pete Wright
The other that Terry mentioned is Anthropics Claude. Claude, it tends to have, I think it's regarded in the AI industry as having stronger guardrails around sort of what it can do and what it will do, where some of the other players don't necessarily. It's also a bit of a scrappy upstart. I think people who use it, what I hear all the time is you run into limits faster than you do with some of the other tools. So Claude is a really good tool and it has some capacity issues. The other that I'm surprised neither of you have run into is Google Gemini. Google's Gemini is an incredibly powerful tool and it is baked into one of the best tools for a specific purpose that I've run into. And that is Google's Notebook LM. Have either of you heard of Notebook LM?

Kyle Pardo
Thank you.

Pete Wright
it can oh the fact that you can hear two people breathing and thinking and making mouth noises is

Kyle Pardo
good as this, but it's interesting.

Pete Wright
unsettling for a guy like me but but yes it's true you can the the reason i want to bring it up is because we talk about using AI on your own data. And Google Notebook LM is an incredible tool for this. So imagine you take all of your policy documents, right? And you take that text and you put them in as resources in Google Notebook LM. You can then effectively interrogate just with the Gemini chatbot, interrogate those policies. And every response it gives you has a little link back to exactly where it found that information. That sort of recursive nature of being able to effectively interview your policies is incredibly powerful when you're dealing with long, long documents or sets of documents. You can put up to 300 separate documents in an individual chat with Google Notebook LM and have just a fantastic hyper-centralized brain of your policies that can be very, very useful. Very early days for this, but you can imagine as, you know, co-pilot from Microsoft and, you know, that increasingly invested in SharePoint, all of these repositories become just incredibly accessible to a degree that we haven't had before, all because of these LLMs.

Kyle Pardo
Yeah.

Pete Wright
So there's a big three. I think perplexity is another one that is that that has some baggage to it. But it's another one that offers great retraceability of steps. All of the models now effective this week now do web search. So when you do a search in Claude and ChatGPT5, you can tell it like make sure to to search the web for these resources and it will tell you the sources it found. Do that. Make sure you do that if you are listening.

Kyle Pardo
Yeah.

Terry Cook
Thank you.

Pete Wright
If you're just relying on text and not looking at resources, you are not Hiddle. You are not the Hiddle.

Kyle Pardo
One of the things Terry and I have been talking more about with AI, too, are the potential scams and the challenges for HR people because we deal with so much confidential information. And, you know, we hear more and more stories of, I don't know how many people have seen like emails coming out with like direct deposit requests. And, you know, and you're not really sure if it's from the employee or not. And you're trying to figure out, you know, are they looking to change their banking information? But we're also seeing where kind of the future of, you know, you look like you're, it looks like you're interviewing somebody, but it's not a real person.

Pete Wright
it's not a person yeah

Kyle Pardo
It's not a person at all. And so, you know, will companies go back to in-person interviews just to verify that you're getting a real answer from a real person? So I think those are the types of things as an HR person that, you know, we don't know what that's going to look like in the future. But I imagine we'll have to put some more safeguards around some of those processes.

Pete Wright
Well, and that's where I want to go for our sort of final segment today, which is around those guardrails with all of these new tools. What are you seeing in terms of formal AI policies? And, you know, what should they include?

Kyle Pardo
Yeah, I mean, so first of all, I think it's going to be a policy that you're going to have to keep reviewing. This is not going to be put it in your handbook and leave it and check it in five years because we just don't know what it's going to look like with AI. So I put this on your system, you know, review it every six months and see if it still makes sense. But I think companies are going to have to put something, some policy together in terms of confidential information. What are the safeguards and what, you know, can you put a search in there with client information or employee information or, you know, is that prohibited? So I think companies should be should be thinking about that. I think what to do if your company does get involved in a scam or somebody's information is jeopardized, making sure you have good. And then that's, you know, kind of data privacy. Companies should have that information anyway, but should really kind of make sure that's wrapped up in their AI policy. Terry probably has more to add to that, though.

Terry Cook
No, I think you're right. I think it's about the security. So it's security of confidential information. It's letting people know how and when to use it. So there should be some kind of a structure or guideline from the company. Definitely not only just the policy, but the follow up and the training. I find a lot of people misunderstand how reliable something might be and might not be. So they might not be aware of the caution areas. But saying it to the other way, they don't realize how they can use it to make life easier either. So I think it's about not only just having the policy, but I do agree with Kyle having a policy, being very clear on what the company expects you to do to protect information, when you're allowed to use it, when you have to check before you put some information into chat, GPT or an AI. format to make sure that it's okay to do. I think all of those things are going to be important. And as Kyle said, with more and more scams coming up, just really being aware that you may have to keep adding to that policy and changing it. And again, with that training and discussions, communication, again, not computer, people, talking to people to explain like what can happen and what's safe and what's not safe.

Pete Wright
I want to talk just briefly about culture and technical culture at work, because I feel like We have two we have two ends of this spectrum, and one exists with teams that are anxious or skeptical about using AI tools, refuse to use it, don't want to get involved with AI, tend to be sort of paralyzed around the conversation at the extreme. And the other is the side that says, OK, look, I was a bring my device to work guy. Like I brought my iPhone to work in 2007 when my company was making me use, you know, BlackBerrys because I like the newest, latest stuff. And our AI policy is too restrictive. It's not allowing me to get my work done. So I'm going to be here on my personal device using AI for company stuff. Right. Both of those things have different sets of risks. And I would love it if you would comment on on how to approach teams that way. And it might involve talking to I.T., right, and making some policy choices. How do you how do you do that? Who would like to take the skeptical employee angle?

Terry Cook
I love how you love that one over here, right? But no, I do.

Kyle Pardo
Thank you.

Terry Cook
I agree with you. I think it is a matter of letting people know what they can and can't use. So it is IT getting involved to the person like Pete who decides to use his own, his own, you know, equipment because you're, oh, I don't know, Pete.

Pete Wright
that was a total hypothetical terry i feel i'm feeling a little bit targeted

Terry Cook
Never, Pete, I would never do that.

Pete Wright
you

Terry Cook
But now I do think that that's true. And some of that is, you know, companies don't want company private information being put on personal equipment. And again, to your point, Pete, that's got to be clear. I mean, because not everybody would see that, you know, using your example. that you said, somebody has a preference of what they want to use. And even if it isn't the Blackberries from years ago, you know, there are people that are like, I want, I only want Apple, I only want Android. You know, there's always people that have preferences and what they're comfortable with. But again, having that information, private company information being transferred over to a personal device can be a problem and it needs to be specific in a policy from IT, from HR, that that's, these are the areas that could be a problem by using personal equipment.

Pete Wright
Okay, Kyle, other side. I think Terry just addressed the non-skeptical employee side, but the Pete side. We'll just call it the Pete side. So I guess, Kyle, you get the de facto skeptical employee side who's afraid to use who's afraid to use A.I.,

Terry Cook
You get so skeptical when.

Kyle Pardo
You know, so I think people are afraid to use AI because they're afraid that they're going, their job is going to go away, right? We hear that. We hear that a lot. And I'm going to go back to acronym of, you know, keeping the human in the loop. Because if we can get all of those day-to-day, you know, mundane activities off of our plate, the, you know, writing letters or creating forms or doing some research or things like that, if we can get that off our plate, as HR professionals, we can get the human back into the work that we got into in the first place. So we have the time then to take the new employee out to lunch. We have the time to plan, you know, a town hall with the CEO and have a company get together just to talk in person. And so very often those things get pushed aside because we have to do those day-to-day kind of tasks. So I, you know, I, I think the people who are afraid of it kind of thinking about like, how can we, how can we embrace this? And it is super exciting and, you know, but get those kind of those things off our plate so we can, we can do the human, human touch pieces.

Terry Cook
A lot of times it's just about, it was a very good answer, Kyle. But it is all about educating them because I think it's either fear because they just, it's the fear of the unknown or it's fear of what Kyle's saying.

Pete Wright
Продолжение следует...

Terry Cook
They're fearful that they're going to be replaced or something's not going to be needed anymore because they're going to be able to show that somebody else can do what they're doing. So I think it is addressing both sides.

Pete Wright
I think so, too. And I think that that figuring out where that happy sort of middle ground is does involve IT coming up with a policy, compliance coming up with a partner policy. All of these bodies coming together to say, look, we're going to give you enough that you can do the work without being overly restrictive as long as you meet us in that middle ground, too, and not start using your own stuff and putting company confidential information into public AIs.

Kyle Pardo
Продолжение следует...

Pete Wright
We can meet in the middle. And I would say as a corollary to the skeptical employee, it's OK to be skeptical. Part of the fear is just not knowing what questions to ask. I don't know which parts of my job I can streamline, so I'd rather just not think about it. And I think that's OK. It's OK. You can get to the point where you'll figure out what questions to ask.

Terry Cook
Продолжение следует...

Kyle Pardo
Продолжение следует...

Pete Wright
It's not, I think, in spite of what some technological skeptics are saying, it's not going away. It's not going away. It's here.

Terry Cook
No. And letting them know who to ask, I think, to your point, right, Pete? You know, I mean, a lot of people might say, well, I don't know. Is that my supervisor that I have to ask that? Is that HR? Is that IT? So really give directions going to be important.

Pete Wright
Huge, huge. What is AIM working on any sort of courseware, any sort of complementary resources to put together for folks that we can share?

Kyle Pardo
Something that goes out to people who are looking to update their handbooks and keep it updated. And as I said earlier, probably one that will continue to update as we learn more about AI. So you'll see that change. And I think you're going to start to see it woven into more and more of our training classes. So, you know, how does, and not just like maybe just a standalone class, but really into the day-to-day HR activities.

Terry Cook
Thank you.

Kyle Pardo
And so I think it's something that we'll all be feeling going forward. We talked about, you know, what will this conversation look like in six months? you know, we'll be beyond to some other new platform, but you'll start to see that woven into more and more of our programs that we offer.

Pete Wright
Excellent. Well, I think this is a great conversation. I think you acquitted yourselves perfectly. And I don't at all believe that you are AI bots having this conversation with me.

Kyle Pardo
um

Terry Cook
Or not.

Pete Wright
I am gratified to say that. Thank you, everybody, for downloading or for being a part of this live stream. However you came across us, we're glad you're here. We hope this gives you some insights and some resources. Make sure to swipe up, check your notes. That's where you'll find links to the to the handbook subscription that Kyle mentioned. And we will keep you updated. You can listen to the show if you missed the live stream. Hopefully you're subscribed to the Human Solutions podcast, available wherever your podcasts are. That's where we put the archives of these in a easy-to-digest, easy-to-drive or carpool audio format. We'd love you to check that out, too. On behalf of Kyle Pardo and Terry Cook, I'm Pete Wright, and we'll see you next time right here on Human Solutions.