Can HR Trust AI in Front of the Board? A Guide for PE-Backed HR Teams

Every HR leader at a PE-backed company is hearing the same question from their operating partner: what are you doing with AI? Underneath sits a sharper fear: a wrong number, produced by a model, reaching the board with your name on it. Greg Doss, Director of Recruitment at Community Medical Services, a HIPAA-regulated addiction treatment provider of about 1,100 employees, has spent eight months answering that in one of the most sensitive data environments there is.

‍

How do you put AI-generated numbers in front of a PE board without getting burned?

‍

The safeguard is not the model. It is the process around it. Every report Greg builds gets a human review before it moves. When a three-minute report replaces an eight-hour one, the pull is to send it fast. He reads it like a skeptic instead.

‍

There still has to be a human aspect, a human check.

‍

What happens when the AI and your HRIS disagree?

‍

You have probably found a real data problem. Greg treats HRBench through Claude as a second set of eyes on the HRIS. When a headcount figure looked too high, he cross-checked Paylocity and found a gap of 35 to 40 employees, traced to a coding change the HRIS reporting had missed.

‍

We're actually using it as a checks and balances to our HRIS system.

‍

Can HR run AI safely in a HIPAA-regulated company?

‍

Yes, when the guardrails come before the prompts. CMS's IT and HR teams wrote an AI policy, then issued Claude for Healthcare to about 50 staff, which rejects any file with patient information. The rule is flat: no PHI, nothing that identifies a person. The team studies groups, never individuals, and HR owns compliance, not just IT.

‍

Claude speaks like me. It doesn't speak like an IT person.

‍

How much time does this save when the board is in the room?

‍

Enough to change what HR can promise. A report that used to take eight hours takes about three minutes. When the CEO forwarded a pitch from an outside AI HR vendor, Greg's reply caught him off guard. On a recent board day, a question that once meant an hour of pivot tables took about five minutes.

‍

I reported back to him 20 minutes later and I had a report. He's like, where'd you get this from? And I said, we already have those capabilities.

‍

How did AI fix a broken onboarding process?

‍

By turning a nine-page mess into a ranked plan. Greg fed 128 manual onboarding steps into Claude and asked for 50. It got him toward 70 and flagged the redundant steps Paylocity already handled. Paired with a 60-day survey, it rebuilt trust between operations and onboarding: a weekly hour-long gripe session shrank to 30 minutes, and the last one wasn't about onboarding at all.

‍

Help me get this 128 manual processes down to 50. Claude was able to say, I can get you to about 70 or 90.

‍

When is the best time to start?

‍

Now, and Greg is blunt about it. Waiting for the perfect use case is its own cost, and the PE firm pushing you rarely says what to build. You will make mistakes. That is the work, not a reason to hold off.

‍

A year ago is the absolute best time to do it. The second best time is today.

‍

The pattern is always the same: the model does the heavy lifting, and a person owns the answer. That is the version of AI a board can trust.

Logan (00:00): The HR pushback often is, it's unsecure, kind of a thing. Which is...

‍

Greg Doss (00:05): We have those fears, and I think some of those fears have been alleviated. I think if you get a good IT department that can come in and set up the rules for you, that helps out. We just recently had to sign our AI policy document, where we know, don't include PHI, try to protect HIPAA and everything like that. So that also makes us secure. But then, like you said, there's a little bit of wild, wild west out there. And you've got people who do utilize it, and we know obviously there's a reason there's a policy, is because there's gonna be somebody who's gonna play that bad actor. But we'll get all that figured out. I think the good definitely outweighs the bad on it.

‍

Logan (00:43): Yeah, and it's super new. And for everybody that's gonna start tuning in, that fear is valid, like in tech. We've seen tech over the last, you know, 15 years, the data hacks and data breaches, they happen even in the most secure companies. So that fear is valid. So we don't wanna downplay any of that today as we talk. But for those joining us, I'm with Greg Doss, who's with CMS, their healthcare provider, and he's on the forefront of using AI in his work and throughout CMS. Greg, is there anything you would like the audience to know about you?

‍

Greg Doss (01:20): Well, I've just been in HR and recruiting for about 20 years, again, trying to be in the forefront. And it's funny you said that. I can remember when I first started, people were like, this LinkedIn thing's not gonna work, it's a fad. And you know, you go in 20 years later. So you want to catch those things when they're still hot. And no, I've been doing recruiting for 20 years, very, very passionate about it, the employee experience and so forth. Started with Community Medical Services in December of last year. So eight months into it, and I've enjoyed the ride so far.

‍

Logan (01:51): Awesome. Awesome. Well, today's episode we're gonna talk all about AI. And everybody talks about AI, I think, ad nauseam. But Greg and I, I think, are both a little bit of Claude nerds. I know I am, so it's gonna be around Claude AI. So for those that don't know, Anthropic, they have their model Claude. You also have OpenAI and ChatGPT. Both models are great. We just happen to both use this one. And so what I wanted to kind of kick this off, Greg, and what really kicked our whole thing off, was you had a conversation with Carissa. Carissa's our VP of customer success here at HRBench. Told you about our MCP connector. So that's the open protocol to be able to connect software with AI tools, for those that aren't that familiar with it, and HRBench has one. And Greg, you said that once she mentioned that to you, you're like, get me in on this, let's see what we can do. Can you walk, let's start there, like your journey for making that connection.

‍

Greg Doss (02:51): Sure. I'll start off with the beginning when I started. I think I started December first, and I think on December 3rd, I was already into HRBench. And what I like about HRBench, it gives you a true dashboard. Love the dashboard look of HRBench. It gave me the details and it gave me the information that I needed. I think one of the biggest drawbacks to it was, well, you have to have an HRBench license in order to utilize HRBench. And we had about 20 licenses. So we have 1,100 employees, and we have probably one to 200 supervisors who all want that data. So it's like, well, we're gonna have limitations with the amount of licensures that we have. And so giving some of the feedback was, I love the dashboard. However, getting the dashboard information into a PowerPoint slide, or getting it into any kind of presentation, was a bit of a challenge. And we kept giving that feedback. I need to be able to just not copy and paste, cut and paste, et cetera. I want to get this on here. And you know, Carissa kept saying, wait, wait, wait. I thought she was gonna say wait, wait, wait because we're gonna get it to PowerPoint. And all of a sudden she lets us know, I think in April, she goes, I think we've got it to PowerPoint, but we also have this Claude connector where you can connect your information from HRBench into Claude and start utilizing the data. I think she told me about it on April. The very next day, I was like, show me how to use this connector. By the end of that day, I had already utilized, I was a max Claude user, already maxed out the information. I got the warning, like, you're cut off until 2 p.m. the next day. But what we were able to do with the data was then take it, create reports, and very specialized, very utilized reports. So if I wanted to know, what is the turnover rate in Ohio, well, how can I compare that to our clinics in Colorado? How can I divide that up by, and it would just, it just kept spewing out data. It was giving us, you know, taking into Claude and already utilizing the reports that I put into Claude. And that's one of the reasons why I do like Claude. I think it really speaks as an HR person. Claude does a really good job of taking my voice, figuring out the prompts that I'm using, and then it kind of mirrors the voice. So you know, it'll go up and say, well, based on your prior reports, do you want it to look like it did before? Absolutely. So we were able to create very customized reports very quickly, and really support our operations teams. Because at the end of the day, that's what we're trying to do. We're a medically assisted treatment for opioid abuse. We want to help people get better. How do we do that? Well, it's the longer tenured counselors, the longer tenured nurses that we have, the better. And so we were able to get really, really quick reporting structures going on.

‍

Logan (05:34): Yeah, and that's, I like how you talked about setting it to your preferences. One of the things that is interesting, you can go into the back, not into the back end, but into your Claude account. You know, ChatGPT has it as well, and give it preferences of how you want stuff. I've done things where I have like context documents that live outside of Claude itself, and it goes and references, and it has to pick up all of the brand assets, like here's how we write things, here's how we do this. And then it gets it right every time. And I think one of the big things that I've liked about Anthropic is how they've done accuracy. Accuracy was like their number one piece. Not to make this a commercial for Anthropic, but I mean, I think that's how they've exploded this year, because, you know, 2025, they were kind of an obscure company that was dominated by ChatGPT.

‍

Greg Doss (06:26): Yeah, and I would use ChatGPT, I think I started using ChatGPT last year, and I liked it, but there were times where I would get stuck on a problem. And to your point, Claude, if you've ever looked at Claude and started a prompt, it'll even correct itself. Like while you're watching it work, it'll say, this data doesn't seem right for what you're looking for, let me keep thinking. And that's what I was thinking when I'm looking at the data, like, yeah, this doesn't look right. And it's saying this doesn't look right, and then it keeps working itself out. And it says it in a, you know, I'm not an IT person, I'm HR, I'm a recruiter. But it takes your verbiage and it puts it into that perspective. And I think that's something that, to me, Claude speaks like me. It doesn't speak like an IT person. And I like ChatGPT, I'll use ChatGPT, it just, to me, didn't have the same functionality from an HR perspective. I also like that, you know, I was in operations as well. Our chief operating officer uses Claude too. So the reports that he was sending from an operational standpoint would match the reports that I was sending on an HR standpoint, which kind of looks like it's all the same thing, same thinking.

‍

Logan (07:31): Yeah. And you guys are able to take the reports that you're doing from HRBench, managing them through Claude, and then you put them in front of executives and the board, right?

‍

Greg Doss (07:41): We do, yes. Good example of that. It was funny. Nick Stavros, who is our CEO, he is a wonderful CEO, is probably one of my favorite CEOs I've ever worked for, wants to be at the forefront of technology. He wants to push things forward. And he wants to push it not just from an operational standpoint, but from an HR perspective. How can we keep moving our AI forward? But about a month after we integrated with the Claude HR connector, he had sent myself and our chief information officer, he saw an AI HR function, I don't know what the company was or whatever, but it was like, we can take your data and we can do this with it. I reported back to him 20 minutes later and I had a report. He's like, where'd you get this from? And I said, we already have those capabilities. I said, now that we have HRBench through Claude, we already have that. It was like, you know, thumbs up, good job. We're good with that, keep giving us that data. But yeah, he's very big on that. I thought that was great, is that we've already got this function built in. And you're gonna see a lot of actors putting out stuff, and putting out sales, and hey, this is as good as it gets right now, and it does the job of what we're looking for.

‍

Logan (08:52): Yeah. It's interesting that you say that, because I saw somewhere, and it was a lot of software companies right now, or not a lot of them, but people are creating software companies, and they're like, well, it actually might just be able to be a Claude skill or a ChatGPT skill, based on what it's doing. Once you kind of tweak it, it's almost kind of like, that's what you guys had.

‍

Greg Doss (09:13): Yeah. There's, I follow, you know, I'll put a plug in for them, but there's this email I get every day. I don't know if you're familiar with it. It's Zain Kahn, Superhuman. Have you heard of them before? Yeah, Zain, Z-A-I-N K-A-H-N, Superhuman. I think they're supported by Slack, or I think he's got a connection with Slack. What he does is he sends out an email every single day. And I read it every day. But what's really cool about this is he doesn't do it from a technical standpoint. He does it from, it's kind of got a mix. It's got your intermediate user all the way up to your experienced user. But he'll talk about things that are going on in AI. Like today he talked about the Claude Coworker. Claude Coworker now has the ability that if you turn on the recording, it's gonna look at your, yeah, it talks about that. So it doesn't need the prompts. Now it's gonna watch how you work, and then it's gonna create the work that you just did, and it's gonna emulate that moving forward. He also does a wonderful job. He'll put out like top prompts to ask something for. And there was one time where he did like the top 50 HR prompts for different systems. And I would try those out. You look at it and go, okay, yeah, this works. Because, you know, sometimes we're playing around with ChatGPT and you don't get what you want. This kind of tells you, well, send it out as this exact prompt and it's gonna give you exactly what you're looking for. And then he's got some sales, there's some things in there that are sales-worthy or whatever, that they can kind of sponsored by, et cetera. But it's a great, great email that I get every single day. Highly recommend people look at it, do what you want with that. I am not associated with them at all, but I've been doing it for the last year, and I think my AI knowledge went from I don't know what I'm doing, to I'm just good enough to get in trouble with it.

‍

Logan (11:02): Yeah, and that's the big thing. Everybody's learning right now. And what's interesting is you have everybody going out and talking about, I mean, not necessarily this newsletter that you referenced, but you have some people saying, like, I know all this stuff about AI, here it is. And then you get reports of people who are actually working on models and they're surprised at what happens. And so you're like, how does anybody actually know what's going on, if the people who are actually developing the frontier models get surprised?

‍

Greg Doss (11:32): Yeah, there's still a little bit of wild west aspect to it. You know, it is one of those things, well, hey, I didn't know it could do that, but now that it's here, here's what I'm using it for.

‍

Logan (11:40): Yeah. So I do want to address one of the concerns that I hear a lot for HR folks. We kind of touched on it as we were shooting the breeze before we got started here, was around just security of data and everything. And you being in a regulated industry is a perfect example of how you can make this work and be secure, with HIPAA compliance as well. Like Claude for healthcare, plus Claude for HR. What is Community Medical Services doing to ensure that you're on the forefront of AI, but also taking the most progressive security measures you can?

‍

Greg Doss (12:20): Sure. The first thing we did, our IT, probably about six months ago, our IT department and our HR department, let's come up with guidelines for using AI. Here's what you can use it for, here's what you can't use it for, here's what you shouldn't put on your own computer, here's what you should, these should be the only prompts or whatever that you're gonna be using. Another function that they gave for several of us, I think 50 of us got Claude for Healthcare. We got the Claude for Healthcare, like you just talked about. And what that does is, if we were to try to put in a spreadsheet that had patient information or anything else in it, it's going to reject it out. You know, so no PHI, nothing that can be used to identify an individual. We want to be able to identify groups, right? This is our 25 to 30 year old employee base, this is our 35 to 40, this is our male-female. However, you don't want to disseminate that information out to the individual level. So we've got access to that. I think also just managing as well. You know, I think you can sometimes tell when your employees are utilizing AI, and you wanna just remind them, this isn't what you want to use it for, nothing that's HIPAA protective and so forth. So we do a good job. Our HRIS system is Paylocity, our ATS system is Paylocity. So still trying to contain as much information as we can in there, utilizing some of the AI functions that even they have in there, because you know you're going to be covered with that. So I think those are some of the biggest things that we're doing to protect our patients' information, because that is very key. You know, in the industry that we're in, people want to remain anonymous, and we want to respect that as well. So we want to make sure that we're at the forefront of not giving out other people's information.

‍

Logan (14:03): Yeah, that's a good disclaimer. So as we continue to talk about AI for everybody, we're talking to a company that has taken some pretty strict precautions, and healthcare is one of the industries that needs a lot of protection for a lot of good reasons. But I want to transition a little bit, Greg, and talk about how, when you're getting ready to put numbers in front of your board and your PE firm, so CMS is private equity backed, and how do you, I guess, I think you had mentioned in kind of our discovery that your CEO came to you and said, how do I know this is right? To give you confidence that you're putting the right information in front of folks, because everybody is worried about AI hallucinations and everything like that. Like, how are you taking that approach?

‍

Greg Doss (14:49): Yeah, that's what Nick will say. I think, you know, we, he was talking one time, he goes, I love AI, we have to ensure that the data is correct. We're hearing all about AI, and I think people will say AI is gonna take my job, or AI is gonna replace this. And I'll kind of divvy it up a little bit better. I don't think AI is gonna take somebody's job. I do think AI is gonna replace people who don't utilize AI. So there still has to be a human aspect, a human check. Any reports that I receive, what I'll do is, I wanna look at the report first of all. You know, your first inclination is to say, wow, this took three minutes, it used to take eight hours to create, I'm gonna send it as quickly as I can to our bosses and show it. But no, what I'll do is we'll look at the report. And you just want to, okay, this doesn't make sense. Why isn't this, this number just doesn't make sense to me. And so we're actually using it as a checks and balances to our HRIS system. So I think, for example, we had an employee headcount that had risen a lot on the HRBench data. And I'm going, this isn't making sense, this shouldn't be right. Cross-check that to our HRIS, and we noticed there was a delta of about 35, 40 employees. So we're like, what's going on here? And what we ended up finding out is that, you know, obviously in HRIS systems, we make little tweaks here and there and don't necessarily look at what the downstream effect is. And we had coded something where it had, you know, Paylocity caught it, HRBench didn't catch it necessarily. And we were able to just go up to HRBench and say, hey, here were the differences. No harm, no foul. This wasn't something you should have seen, but it came up. Also look at it when it may say this location versus this location. You know, we've had some where we put in one location, but we gave it two different names. So we're able to go, nope, here's what it really needs to be. And then even positions. We have nurses that are LPNs, PRNs, and making sure we have the right, making sure that it flows correctly to our HRIS system.

‍

Logan (16:50): Yeah, that's a good call. I like the location data stuff. I, there, I mean, the adage is bad data in, bad data out. And I have come to the conclusion that I like to challenge that a little bit with AI, because while that is, and can be, true, AI can also help you clean up bad data in certain areas. And so if you prompt it and say, hey, this data is a mess, I need you to help me clean it up, you can actually flip the bad data in, bad data out on its head, and actually use it to clean your bad data. I don't know if you've done that, but...

‍

Greg Doss (17:10): Absolutely. No, we, there's been some times where I go, this isn't making sense. And what I like about Claude too is it doesn't just fix it, it'll give you the options. I know I was doing something just about an hour and a half ago, and it said, based on this, here's something I don't, you know, it gave me three options. Here's option A, here's option B, here's option C of why it thought the data was wrong. And I'm looking at it and I can pinpoint A and go, okay, I don't think it's A, I think it might be B. So then I click B, and then it shoots it out. Go, yep, this is what it's looking like. So no, some of the bad data in, bad data out is human. You know, how many times have we looked at something, and you got a report and say, yeah, we've had the wrong information for six months. I think now, when you've got HRBench, you've got your ATS or HRIS, when you start seeing those deltas in the numbers, it just gives you a quicker turnaround time, so that you can say, you know, bad data in, good data out, as a result of seeing the bad data.

‍

Logan (18:25): Yeah. Have you tested Fable at all with any of this stuff?

‍

Greg Doss (18:30): I have not. I'm getting a prompt on it, it is one of the things that I want to do, is start looking at it. Where I was like, okay, this is next on my list of things to do. You know, I'll probably tell you, I'm not always at the forefront. I like to be the second user. So I always joke, I didn't get an iPad one, I got an iPad two. But I'm almost to the point where, like, okay, I see people are using it, now I want to get in there and try it out.

‍

Logan (18:56): Yeah, for those that are listening in and aren't as big of nerds as Greg and I are, Fable's the frontier model. It was like flagged as being super good. The US government shut it down for a little bit, like all this stuff. I tested it on some sports data, relatively benign, and I did it between Opus and Fable. And I ran the assessment. So I had Claude Code run two different things on a scoring system I was building, and then brought it back into a chat and compared, like, wow, it really did better on all these different things. And it's pretty impressive, especially with large swaths of data.

‍

Greg Doss (19:34): I gotta ask, what sports was this?

‍

Logan (19:36): So, fantasy sports. So, to help my fantasy sports team, because I'm terrible at fantasy. I'm in dynasty fantasy leagues. And I'm terrible at it. I'm in like last place. And I'm like, you know what, I need some help. I need a lot of help. So...

‍

Greg Doss (19:51): That's it. Has it helped you out since?

‍

Logan (19:53): Yes. I've actually got like a whole system that I run. You know, I do like an hour or two at night. And my big breakthrough, I think, is matching what the market says a player should be worth versus what their actual data is worth. Because, so we're not taking a tangent for everybody, but I'm a big Moneyball guy, 'cause I'm an Oakland A's fan. And so everything I think about in sports is in that frame of mind, because when that was coming up, I was in high school. You know, we were going to 15, 20 A's games a year during that period. So it's a huge piece for me to find, like, where's the edge? And so that's kind of what it's trying to do for my fantasy stuff.

‍

Greg Doss (20:33): Okay. Yeah, at my last company, I would have, in all of my PowerPoint presentations, at the very end it would say adapt or die. Which, for those of y'all that don't know, that's a Moneyball quote, where Brad Pitt's going, guys, we got to adapt or die. You know, we're not gonna out-Yankee the Yankees, adapt or die. Going back to your fantasy and how it interacts with us, when we're running our HRBench, one of the things that we did was a 60-day survey. And we wanted to see how our onboarding was performing. So, similar to what you're saying, how am I doing? And the first data points that we got, it said, you need to improve the hospitality of your onboarders. People are like, hey, I feel like it's not a good experience for us. We fixed that. Shortly after that, I was in charge of onboarding. I said, guys, we just want to focus on hospitality, hospitality, hospitality. And the data points have seen, you see it go from not a strength to it was a strength. And so what it's also evolved out to is that, for us right now, we've got a couple of our frontline workers who, now it's no longer the hospitality, it's like, we want more hands-on training. So because we've already been able to show, here it said we needed hospitality, we got hospitality, people are happy, we were able to pivot with our training team and say, all right, we've already revamped our nursing training and our counselor training so that they're getting a full structured 90-day training, just based off this data. So we're gonna be able to see how it goes. So it's kind of similar to your fantasy football, your fantasy baseball team. They're not gonna help you out now, but it will help you out. What you're looking for is that gradual...

‍

Logan (22:08): Yeah, exactly. And that's a really good segue to the last point that I wanted to chat through with you. Taking a look at all of your manual onboarding processes, process, processes, see if I can speak, and throwing it in the cloud and seeing what you could optimize. I think it's a very tactical thing for folks. So walk us through what you did.

‍

Greg Doss (22:30): Sure. So when I started to oversee onboarding, and like I said, I started in December with recruiting, and then by February, they're like, well, we want you to oversee recruiting and onboarding. I went up to one of our onboarders and I said, what are your frustrations? He goes, I've actually, giving you a spread. He had a nine-page document that was all of the manual processes. So I'm looking at this going, we have 128 manual processes. And so then, you know, I get to my Moneyball analytics, and I'm talking to, because our financial controller was like, you guys, your onboarding team makes too many mistakes. So I went up to him, I go, we have 128 manual processes. At a 90 clip, that's 13 mistakes, 11 mistakes. And it kind of got him going, wow. So what we did is, I took the document, put it into Claude, and I said, help me get this 128 manual processes down to 50. Claude was able to say, I can get you to about 70 or 90. And it kind of showed and documented which ones were in there. And so we immediately went, partnered with Paylocity, and said, hey, these are the manual processes that are coming in that are giving us the most grief. And a lot of it, you know, when you've got a technical person at Paylocity, they're like, most of this you don't even need to be filling out right now. So, okay, can you black that out, blot that out? He was like, it already catches it here. So a lot of them weren't just manual processes, they were redundant processes. So, you know, we've got that down now to about 50, well, how do we, or about 70? How do we keep working with Paylocity? Because, you know, we're partners in this. We're gonna be with you guys, you guys are gonna be with us. How do we make this effective for everybody? But Claude was great about saying, here's a roadmap to how you can eliminate some of those manual processes.

‍

Logan (24:18): And this is where AI can do things that humans can't. I think of it as, humans need less information and AI needs more information. And so you dump as much in and be like, hey, help me consolidate this, and kind of use it a little bit as a sparring partner.

‍

Greg Doss (24:35): Absolutely. Absolutely. No, it just gave you a workflow where you're just looking at it and going, okay, these are all palpable pieces that you can digest, right? If you handed me this 128, okay, I can get you down to 127, 120, you know, but this was like, here's the order in which you should go, here's how you should attack it, here's how you need to move forward. And that allows us to do other things in HR that are a better skill set for what we have, right? We know getting the data right is such a key function of our job, but we also know there's other fires to put out. There's recruiting, there's talent acquisition, there's training, there's employee relations. So this kind of allows an HR person more time to get in the field and do things that they really need to be doing to impact your employees.

‍

Logan (25:19): And what was, I guess, what's the outcome of the onboarding process now that you've simplified it a little bit? What would you say is...

‍

Greg Doss (25:26): Less errors, better hospitality. The best example that we can say, we had gotten to the point on the onboarding where there was probably a little bit of a distrust between our operations team and our onboarding team, right? There were errors, there were things going in. Our chief operating officer at the time, he had set up, he goes, moving forward, we're gonna have a weekly one-hour meeting where we're just gonna let it all out. So for the first three or four weeks, it was more than half of our regional directors, and they were letting us know everything that was wrong. Within about three weeks, there's about two or three regional directors, and it was no longer an hour, it was 30 minutes. And so this last meeting we had on Monday wasn't even about onboarding. It flipped back to the recruiting side. There was a couple of issues on recruiting. So it was like, okay, we saw the problem, we solved for the problem, we're going to continue to manage the problem. But yeah, this was a game changer, right? Where you've got that distrust, you've got that, hey, what's gonna happen, what's gonna go wrong, to, all right, let's move on to the next thing, which is what we wanna do. Let's move on to the next fire, or the next thing that's broken out. So, like I said, just by changing those 128 manual processes and instilling a hospitality in our team, we were able to get a very good working relationship between our operations team and our onboarding team. And that allows for a better training experience, a better employee experience. And again, this is, that was a kudos to Claude for that.

‍

Logan (26:55): There we go. Well, good, Greg. Is there anything else you would like to address that we didn't address already through this?

‍

Greg Doss (27:02): I think, you know, again, I think sometimes people would say, when's the best time? I mean, when's the best time to start AI, right? I want to do AI, when's the best time to do it? You know, a year ago is the absolute best time to do it. The second best time is today. So if you're on the fence about it, this is the time to get in there. You're gonna make mistakes, right? And I think that's gonna be part of it. We've done some things with AI and, okay, I don't like the results, let's move on. But I think that would be the biggest suggestion I would have to somebody in HR, particularly on the PE side, right? We know that the information in is so key critical. So that would be the biggest suggestion I would have.

‍

Logan (27:40): Yeah, that's a good suggestion. And if you're a PE-backed company and listening to this, your PE firm is likely wanting you to do something with AI. They may not be telling you what to do with it, but just challenging you to do something. And to your point, Greg, about iterating, and maybe you don't like the output at first, there's some things that I've done, and it's taken me like three months to get it to where I want it to be, just iterating. And I'm just like, you know, I'll do it for now, but it's just not quite right yet. And I think that that's okay.

‍

Greg Doss (28:10): Yeah. And it keeps your history. That's what I like about, you've got your history, your Claude history. And I've done that. I had something the other day where I was like, you know what, I think I remember working on that in February and I didn't like the results. And then I looked at it, and it's that aha moment of, nope, this is what I should have said, or changed it. And then we got some better information out of it, we're able to use it. But no, the biggest example I would say right now is that we're having a board meeting tomorrow. You know, I think in the past, when I was having a board meeting, I knew I had to wait, had all my information, all my Excel spreadsheets, everything lined up, because your CHRO is gonna text you a question, hey, they just asked this. And then you would look, redo all your pivot tables, sometimes it would take you an hour or whatever to get the answer. And we utilized it this last one, and my COO had a question. He goes, hey, they want, I know you sent me this, here's what they want. And five minutes later, we were able to get the answer. So it just gives you a more narrow approach, which is so key critical. You don't know what questions are going to be asked necessarily in a board meeting, and what the focus may be, or what the focus may pivot to. But with Claude in there, you just go in there, and with HRBench, all right, based on this data that you sent me, adjust it to this. And two minutes, three minutes later, you get your answer.

‍

Logan (29:25): Yeah, that's a good call out. Well, awesome, Greg. I appreciate you taking the time to chat with us today and nerd out on all things Claude, HRBench MCP. Where would you like people to connect with you?

‍

Greg Doss (29:37): You can connect on LinkedIn. So, Greg Doss on LinkedIn, with Community Medical Services. I don't do a lot on my Twitter and all that, that's mostly sports related, so I try to keep it on with LinkedIn. But no, I appreciate you having me on here. And if anyone ever has any questions, I know HRBench has a couple of different companies they've sent over, where we've talked and kind of geeked out, and some people who have a lot better budget than I do, they're like, oh my God, this is what I'm gonna do, and go for it. But no, Logan, I appreciate you all having me on there, and appreciate HRBench and everything they've done.

‍

Logan (30:08): Awesome. Well, thank you very much, Greg, and enjoy the rest of your day, everybody.

‍

Greg Doss (30:12): Well, sounds good. Thank you.