Subscribe
Share
Share
Embed
Serious People Podcast: An Operator's Manual to AI is a podcast for business leaders and operators running complex, real-world businesses. The ones selling supplements, managing caregivers, or running service crews. Not software.
Host Noah Levin brings nearly two decades of experience at Amazon, Whole Foods, and healthcare tech to weekly conversations about what AI outcomes work inside a business, and what doesn't.
Equal parts practical and irreverent, but useful above all else.
[00:00:00] Jon Harmeling: 10 hours of my week is now freed up using AI. The big question for all the managers and folks out there that are freeing up time because of AI: what do you do with that time? If you're like me, you find more stuff that AI can do for you. Maybe I could use that free capacity and I could fill that extra capacity with leadership. So instead of doing more AI for management stuff and getting tasks done and being transactional, how can I be a better leader and be transformational and influence people?
[00:00:30] Noah Levin: Welcome to Serious People. Okay, so you're running your business and you've used AI to automate a bunch of routine work. So what do you do with the time you get back? Jon Harmeling has an answer. He is a retired Army major, a former West Point instructor, and now he runs Chick-fil-A Miami Shores. Jon is using AI to automate tasks like onboarding and scheduling, but he's doing it so he can spend more time on the front lines with his team.
Jon's here speaking not as a Chick-fil-A representative, but as an experienced operator and leader of frontline teams who's putting AI to use in the real world. Here's Jon.
[00:01:07] Noah Levin: Jon, I, I think of you as like a kindred spirit a little bit, uh, and I th- which is a really self-aggrandizing thing to say to someone who has served in Afghanistan, taught at West Point, and is now running your own very, very challenging business in Miami Shores. I appreciate about you so much that you have this philosophy of how AI applies. What's the word you just used? Uh, "in a deskless environment"?
[00:01:28] Jon Harmeling: Yeah. Thank you for, for teeing that up. It's just been interesting trying to implement AI in, yes, uh, I would call it the frontline or deskless, basically non-corporate, non-cubicle, you know, type setting. So I lead a restaurant team right now. The majority of my team is not sitting behind a laptop. In fact, there's only a handful of us.
So how do you implement AI in that type of situation? And that's where this aha moment in the last couple months, where you go from using AI for management and productivity over to using AI for leadership. And then not just that, but like, how does AI help me be a better leader? There's been some fascinating revelations here in the last couple months. I'm working on a couple tools. It's just been really, really cool.
[00:02:09] Noah Levin: I know that anyone who's listening to the show who's ever experienced a Chick-fil-A as a guest, and it's a guest, right? Not customer.
[00:02:16] Jon Harmeling: That's right. That's actually very intentional. If we put it in writing, typically we'll even capitalize the G just to show honor. This is, in fact, our guest.
[00:02:23] Noah Levin: So anyone who's ever been a guest at a Chick-fil-A, anyone who's ever run an operation of any complexity, can appreciate what's going on in that box. And there's a lot to talk about in terms of the operation on its own terms and then the way you've been able to bring AI into that operation. But before we do any of that, I wanna rewind a little bit and talk about some of your origin story, because I think it clearly has shaped your values and your approach to running your business today. Can you tell us a little bit about what systems engineering is?
[00:02:53] Jon Harmeling: I joined the military when I was 18. Had the opportunity to go to West Point, was a cadet there, played football for Army. Really loved that experience. Started my military career after finishing that. Moved all over the country. I think we moved, like, 12 or 13, 14 times. Spent some time at war in Afghanistan, lived in Germany for a little while. And then one of my last assignments before retiring was going back to West Point to be an instructor, or a P-, professor. Specifically was in the Department of Systems Engineering, and the idea there is just how do you build a system of systems, and how do you measure that entire system from start to finish?
And to me, it's very much like a puzzle. And so when we start talking about Chick-fil-A, it's very interesting, 'cause there's a massive system, right? The machines have to be working. Team members have to not sleep in and actually, you know, show up to work on time, be in position, be trained to do what they do. Everything has to be flowing in order to be able to hand that sandwich to the guest that comes in through that.
So getting a chance to teach over at systems engineering was really cool. Specifically, the class was called Project Management. And so we just kind of walked through how do you build a project with time and cost and scope constraints, and then how do you measure each of those steps? How do you go through change management?
[00:03:58] Noah Levin: I want to ask you about both the operations and the values that the military and Chick-fil-A have in common, because I feel like there is some continuity there. But starting with the operations, what is it about the military and Chick-fil-A that appeal to a, a similar caliber or mindset of operator? Why is there continuity from you teaching systems engineering and doing it in the field to now running your own operation inside of a restaurant?
[00:04:26] Jon Harmeling: I think a little bit of context is, with the military, at least in my experience, I changed jobs and I moved every six months, a year, no more than two years. And so, you know, you're in a number of different jobs. In some jobs, you're a little bit more of a business-based kind of office cubicle, working on some software.
In the next job, you're leading hundreds of soldiers at war in Afghanistan, where the stress level is pretty high when you're trying to do your job and folks are shooting at you. And so when you think of that environment, then you also think of a Chick-fil-A environment, like, can Jon lead 100-plus people? I've done it before. I think it went okay. Can Jon build teams? Can he handle stress? Has he ever been in charge of, like, assets or facilities? Has he ever worked through operations? The early mornings and the late nights. Uh, I mean, there's just so many parallels that come with being a military leader and then also leading a restaurant.
Yes, I understand some people may or may not see the leap there, where it's like, so you shot guns and now you can make grilled nuggets? Like, I don't see the leap there. But when it comes to leadership, I do think there's just a lot of intangibles there, whether it's in the military or whether it's, you know, leading a restaurant. Pretty cool transition there in my career.
[00:05:36] Noah Levin: You've described your work for Chick-fil-A and the restaurant that you run as your ministry, you're in your wife's ministry. Can you talk more about what working in this restaurant is to your values and your faith?
[00:05:47] Jon Harmeling: First off, my faith in Christ is a huge, huge part of my life. I think it feeds into a lot of who I am as a person, as a father, as a husband, as a business leader as well.
I had a chance to go to an event. It was called Life Surge. Happened to be here in Fort Lauderdale. I was not yet selected for Chick-fil-A, but I had already retired from the Army. It was a very unique season of life. It was about two years of this kind of limbo, like, will I ever get selected? Am I gonna get terminated? Should I get another full-time job? Like, what are we doing for my career while we wait potentially on a Chick-fil-A opportunity that's not guaranteed? And so I went to this event called Life Surge, and it was kind of business and faith being mixed together. And Tim Tebow got up on stage. If that name sounds familiar, he won the Heisman Trophy years ago.
Tim gets up on stage and there's thousands of people there, and he looks right at me, and one of the comments he shared changed my life. And he said, "The greatest tragedy in one's life is to be successful at things that don't matter." And so I don't know if it's for you, Noah, if one of your readers, it kind of hits home where it's just like, dang, like, I've been chasing success for a long time. Nothing wrong with that. Don't mishear me there. Like, I, I wish everyone great success. But then it just gets you thinking, where it's like, man, if someone is at my funeral and they're speaking about my eulogy, are they gonna say, "Here lies Jon. One of the greatest things I remember about him is doubling catering sales year over year"? Like, who cares about any of that stuff? Like, it's good, it's fine, but if that's what I'm known for, shame on me. And I just felt like I was focused on the wrong things.
And then this past year, Noah, it's just been this huge kind of revelation of a lot of young adults that just want to be led and coached and developed, or they just want to be heard, they want to be hugged, they want to be seen, they want to be valued. And so you're right, I do view my team as a ministry. This is the human element. This is what I want to be known for. I want to be known for caring and influencing people and bringing glory to God with my life. It's people over profits all day long.
[00:07:40] Noah Levin: I have three reactions to that. One is just that it's beautiful and admirable. Two is that anyone who hears you say that would think it has very little to do with chicken. And three is that anyone who hears you say that would certainly not expect you to then launch into a conversation about the value of AI. Probably the stereotype out there would be that people who are so human-oriented as you are and so focused on caring for the people on the team would sort of be naturally, maybe not opposed to AI, but at least not that interested in the power of AI to create leverage in a business and an ability to reinforce the things that you care about in yours. I'm wondering if you could share kind of your origin story with AI in your business, and when was the first moment that you tried to bring it into your operation? Did it work or not? And when, when did you have your, your aha, your aha moment?
[00:08:34] Jon Harmeling: I'm a couple years into this thing. I'm about two and a half years into the restaurants. But if you can just kind of imagine launching a brand new restaurant for a guy who's never run a restaurant before. I've-- when you launch the restaurant, there's just so many unknowns. Chick-fil-A is phenomenal to provide, I call them master trainers. They work full-time at different Chick-fil-As. They come to your restaurant to help you launch it. And so the idea is, like, let's make sure your grand opening day is successful. If you think about it, that's the day that your team is the least experienced. It's literally their first day.
And then it's also a huge launch, and media's there and, and newspapers and cameras and the mayor. You know, it's a big deal. So you have a huge amount of people with the least amount of, you know, experienced people on your team. So Chick-fil-A sends these trainers that come out, and for the first couple days, they do a majority of the work. Your team kind of watches, looks over their shoulder, and then slowly they step back, your team steps up, and then eventually they're gone and you're on your own.
So when I launched the restaurant, I thought for sure I needed a full-time admin director, right? Just do all the paperwork and the emails and the invoices and the payroll, like, everything else. And she was gainfully employed that first week, right? Figuring everything out. And then her 40-hour week went to about 30 hours of admin tasks, and then kind of 25 and then 20. And she got really good at doing those tasks, and there was just less of them to do. And we got to a point where it's like, crud, we've hired her into a full-time, 40-hour-a-week role, and I only have 10 hours of admin work to do. I felt really bad. It was just a misjudgment on my part, honest mistake.
And so we got to a point where it's like, "Hey, I think we need to start looking at maybe some other opportunities. I want you to grow in your career." She wanted to be an HR professional. And so she moved off our team. Great win for her. We were sad to lose her. She was wonderful. But who's doing that 10 hours of admin work? Jon is. And so that now fell on my plate. Now, I don't have to do it, but my philosophy was if I could do more of the administrative work myself, that's less crud that my team has to deal with, and that allows them the time and mental capacity to focus on leading their teams instead of doing emails and paying invoices and everything else that's super important that has to happen.
It just doesn't need their big brains to do it. It just needs someone to do it, so might as well me. It started becoming a lot. It was just taking up so much of my time, and it-- none of it's hard, right? It just-- I need someone to click some buttons, move some stuff on a spreadsheet, submit reports, pay a couple invoices. And so I was right on the edge of hiring, like, a virtual assistant, executive assistant, or even taking one of our team members and saying, "Hey, for, like, just a couple hours a week, can you knock this stuff out?"
And right around that same time, I started really leaning into AI and started coming across not just AI as a thought partner, which I had been using it at that point for a long time as a thought partner, but, like, AI automations. Like, it'll actually do stuff for me. And that was just fascinating. And if anyone's listening and they still are using AI as a thought partner in research and analysis, that's great. But I just really encourage you to lean in to some of the innovation that's happened. It will do the work for you. Now, what work does it do for you? Can you trust the response? What kind of information are you feeding it? Definitely don't feed it sensitive information.
But again, I don't need any judgment to do some of these admin tasks. So for example, our onboarding process is 15 steps. Only two of them require a human, so maybe I could use AI to automate the other 13, and I did. It took a long time, right? We screwed up a whole bunch together, but we figured it out. And now it's at a point that we just hired Noah, and I click a couple buttons and say, "Hey, onboard Noah," and it does 13 of the 15 tasks. And then those last two tasks are the ones that require a human.
It's just been a really interesting transition from “Jon's doing all the admin” to, “AI is doing all of the admin that is repeatable and that has a low risk of getting messed up or has a low amount of judgment needed.” It's just click here, move that, do that. So that has been really cool, because then I've been able to free up, call it 10 hours of my week is now freed up using AI. The big question for all the managers and folks out there that are freeing up time because of AI: what do you do with that time? If you're like me, you find more stuff that AI can do for you. And I'm not suggesting that's wrong. Maybe you're in a season of your business that you need to do that.
But these last few months has been maybe I could use that free capacity and I could fill that extra capacity with leadership. So instead of doing more AI for management stuff and getting tasks done and being transactional, how can I be a better leader and be transformational and influence people? And this shift into using AI for success like a manager into using AI for significance like a leader has just been really fun, really rewarding, and, quite frankly, just really exciting.
[00:13:26] Noah Levin: I know we're gonna go into a lot of what you fill the time with and why it matters that you can compress the busy work using AI, but I wanna save that for a second from now and just help people understand what literally is happening that's allowing you to get leverage on your time.
I think the reason it's interesting to hear about your experience with this is that you are not from a technology background, right? I haven't heard you say computer science anywhere in your, in your story. I also know that Chick-fil-A is providing you with a lot of the systems that I assume that you're expected to use for parts of running your operation. And so you're not bringing your own stack, you're not building from scratch. You're, you're working within the constraints and tools of a system that has been given to you.
Maybe just before we get into use cases, can you say, like, one or two more things on what are the red flags that you put up when you are thinking about a use case that you'd say this is or is not a fit for uploading to your ChatGPT or Claude account?
[00:14:25] Jon Harmeling: I encourage folks just to be, just to be super sensitive. So if anything involves Social Security numbers, medical information, team member birthday, guest credit card information, address, like, a lot of those kind of things, trade secrets, maybe anything that could be, if it became public knowledge, it might be a big announcement, it might be a reveal, it might be a new restaurant launching, it might be a new initiative, stuff like that. Just be very, very cautious how you do that. And I just encourage folks, if they have a higher headquarters or, you know, some legal folks, maybe sit down with them.
Now, in terms of what specific tasks I think folks could upload to AI, if you could imagine kind of a quad chart a little bit, where we have a little bit of a, of a T there. Some things that are repeatable, predictable, require very little judgment by a human, I think those are phenomenal opportunities to just automate the whole thing. Not from the beginning, but just, you know, start with a little bit, maybe step one, maybe step two, step three, and just kinda work its way out.
Don't fully trust everything that AI spits out. It'll spit out the wrong answer very eloquently and very confidently. It doesn't mean it's right. So start small. Even start with non-consequential, maybe non-business stuff. Just have it analyze something that has nothing to do with anything, just so you get a little bit more familiar with the tool.
But I think there's also some great opportunities here where maybe AI does a lot of the work and the research, and then a human reviews it and makes sure that it's good to go before you take action on it. I think that's great. I think there's another part where the human does most of the work, and then maybe AI just kinda double-checks it, sees if there's any type of illogical comments or redundancies, maybe even something simple as spell check, contradictions, that kind of stuff.
And then there's gonna be some things, I would suggest almost all the leadership stuff, those moments are still gonna have to be really human. What I don't want folks to start getting into, that I think we're gonna start to see more and more, is like, "Oh, that was a mistake. Well, AI told me to do it." It's like, all right, you know, a human still needs to own the output and the results and the consequence of everything we do. We can't just point fingers and be like, "The robot told me that this was a good idea, so I just did it."
[00:16:24] Noah Levin: Today's episode is sponsored by Valet. Jon and I talk about what happens when AI makes something useful inside of a real operation, but the next problem is getting that work out of a chat window and into a place other people can actually use it. For this episode, I used AI to turn our conversation into an operator field guide: the practical lessons, source links, and a few prompts you can try with your own team.
That page lives on Valet. It gives an AI-created site a stable address and a place to be shared instead of leaving the work behind buried in a conversation or folder. Valet is the founding sponsor of Serious People. They took a bet on this show before it existed, and I'm grateful to have them along. Go to valet.dev to see what you can build today.
So you mentioned one use case in onboarding employees. What's another use case that you just have been thrilled to get leverage from AI with?
[00:17:14] Jon Harmeling: Yeah. So I mentioned onboarding. Uh, offboarding is another one. Online reviews, that's been really interesting for us. So I think there's a clear shift from SEO to AEO. So SEO, search engine optimization. Someone types in, you know, "Miami restaurant." Maybe we have a Google ad that pops up, or maybe our restaurant pops up on the first page.
You know, search engine optimization I think is really important. But I think there's also a very clear trend that we're heading toward ask engine optimization. So, you know, if someone grabs their phone and just says, "Hey, I'm heading to Miami on vacation," you know, "plan out our whole three days that we're there, including the top-ranked places to eat." Is it gonna recommend my restaurant or not? And that depends on your online reputation as a component of that. So I wanna make sure that our online reputation is, you know, accurate and that we do manage it. And so AI has been really cool to log in and... Or excuse me, I log in, but also while I'm logged in, I'll ask it to analyze all of the previous reviews on our account, whether it's Google Maps, Google Business reviews, whether it's DoorDash reviews, Uber Eats reviews.
Tell me if there's an issue here, and especially if it's something that's time-sensitive. Like, it would be really cool if, or it is really cool, is when AI sends me an alert saying, "Hey, I just reviewed your DoorDash reviews. Of the 20, 19 of them are five-star. We love your restaurant. One is not, and it says that you were missing part of their order."
If that pops in within 15 minutes of the actual event, I might still have an opportunity to recover that guest and, and kind of win them back. And so I think that's a really cool use case of using the automation to provide an alert to the human and then for the human to take action, in which case maybe their order was two bags and we gave the driver one bag and they took off and the second bag is still sitting there. I'll remake the food, I'll deliver it myself. Like, "I'm so sorry we screwed up." I think that's a really cool use case there.
[00:19:03] Noah Levin: Do you have that use case running on your, on your computer today?
[00:19:06] Jon Harmeling: We do.
[00:19:07] Noah Levin: Is this your personal computer or is it one of your team members' computers?
[00:19:10] Jon Harmeling: My personal computer. I do the majority of it for our team. Again, I wanna try to take basically any computer time off my team as possible. The only computer time I kinda want them to be doing is the schedule. Aside from that, I really don't want them doing, you know, emails and invoices and all that kind of stuff. We have such an opportunity to lead and develop and coach our team, and I want you guys out there on the floor doing those things and providing that exceptional, you know, level of service. So I've used all the tools, whether, you know, it's the ChatGPTs and the Claudes and the Claude Coworks and the Perplexities.
[00:19:41] Noah Levin: I think something that is a common misconception, or at least people get sort of turned around on, is the idea that AI in a corporate environment has to be complicated, and they hear words like agents and loops, and they don't quite know how to execute that. I just want to map, like, what you did to set this up. So I mean, where we are in the chain is, I think you said it's on your personal computer or your, your work computer, but it's yours, right? It's not a hosted system. It's not a server rack in the back room. And you, it sounds like you installed Claude Cowork or, or ChatGPT. Do you have a preference between those two? It doesn't really matter, but--
[00:20:14] Jon Harmeling: You know, I think everyone started on ChatGPT, maybe branched off into whatever was the tool of the week. And then some of these releases that they've had have just been, like, monumental improvements on language models and capabilities. So went from ChatGPT, went to Claude browser extension. That's when I actually started, like, watching it. I leaned forward, I, like, watched it on my screen click buttons and do things for me. Never submit on my behalf, but, like, draft it all for me. And I was like, "Oh my gosh, that's, that's correct. That would've taken me a while to type that out." From the browser extension, then went to Claude Cowork, and then recently ChatGPT just seems to be pushing out some great tools that is rival, rivaling-ing, uh, Chat-- whatever that word is. It's, you know, being more competitive. I think it just depends what your task is. You know, are you trying to generate, uh, a video or, you know, a flyer, a picture? Maybe they're the best tools, maybe they're not the best tools. So try a little bit of everything.
[00:21:08] Noah Levin: I've been on that exact journey of ChatGPT to Claude Code and then back to now ChatGPT. I think that it's unbelievable what that team is doing, and we can get into that if you want. But that, that would be called a harness. You've got that harness open on your computer, and so you just prompted it to do that review process for you, right?
[00:21:25] Jon Harmeling: That's right. And typically just start with, you know, step one. For onboarding, something we do, I think, is unique, but I think it's really cool, is we create what I call baseball cards on everybody. So it's got their profile picture, whatever picture they submitted. It has their name, it has a biography, they submit their own kind of bio. And then it's also got what country they represent. So if it's, especially if it's outside the United States, I think that's really cool.
They would submit a profile picture, they would submit a bio, and basically it was just all these different stovepipes of information. So it would be grab this, you know, picture here, pull up the baseball card there, add the profile picture, change the name, update the tracker, how many folks from this country, from this country, from this country, add their bio, sort it by alphabetical order.
So each of these are just kind of one step at a time. Initially, it does take a little while, uh, to get set up, but I would set it and I would just stare at the screen and just watch it and then be like, "Whoop, whoop, nope, you messed that one up. I actually meant this over here," whatever the topic might be. But you just set up the step one, and once step one is, like, locked in, let's get on step two and then step three and step four. And so now when we have someone, I use you, for example, when you join the team, I have my spreadsheet open with all 15 steps, and I'll instruct AI just, "Onboard Noah." That's it. I don't say anything else. So it knows where to go to pull that information. It creates your scheduling account. It can order your, you know, your uniform, because you already submitted your sizes. It sets up a one-on-one meeting with me and you. It shares my calendar with you. You set up the meeting. We get a chance to kind of hang out, get to know each other.
Now, again, if any of these steps get messed up, it's a very kind of low risk. The name was misspelled or in the wrong place, or we got the wrong country flag added to their bio, like, whatever that might be. I can fix it real quick. I think those are perfect candidates to try to replace with AI automation.
[00:23:09] Noah Levin: Th-this is one of the things I want people to take away from our conversation, is just the, the way to get started is to get started. And the idea of kind of going through the process step by step through the agent, as opposed to doing it yourself by clicking around to different apps, it's sort of neutral to maybe slightly harder the first time you do it, but the second time it's way easier, and the third time is easier than that. And so now you're in this virtuous cycle of the corrections you've made are steering the agent for next time, and now you can just say, "Onboard Noah," and it knows all the scenarios that it needs to weave around to get there.
[00:23:45] Jon Harmeling: So everything we've been talking about for the most part is management, right? It's task-oriented, it's getting stuff done. It's important, again, but, like, the excitement has really kind of come out in these last couple weeks using AI for leadership, and I think a very quick, easy example is just creating a goals database for everyone on our team.
You know, we have anywhere between 80 and 90 folks that work on our team. Each of the people is just like, what goal are you working on? Not necessarily in the restaurant, but maybe even outside the restaurant. For some folks, it's just getting into college. For others, it's learning English. "I wanna learn how to drive a car. I've never driven a car before." "I wanna save up an emergency fund of $1,000." Whatever these goals might be, I think it's so cool. Like, why can't we just capture that, put it on a database, and then nudge me as the leader to help coach and encourage these folks towards their goal? That's the piece that AI can come in and just help significantly.
So right now, as we're recording this, if I opened up my AI tool and it said, "Here are the three people that you need to pull aside on the floor today. Noah is one of our new hires. He just finished his first week of the job. You need to go and check in with him." And so I just think it's a really cool way to use a tool, is to nudge me as a leader and say, "Here's some suggestions, recommendations of what you should do today as a leader. Now, you human, go have those human moments."
And what was driving me nuts was my own screw-up, is when I would have these conversations on the floor with people and I'd walk by Jim and he would mention something personal he's going through, a challenge, or he would mention a goal. Unfortunately, I forget about those, because as you're walking the floor, there's so many-- not distractions-- your brain is all over the place. And so AI's been really good to help, you know, keep me on track and just nudge me in the right direction to help our team accomplish their goals.
We use Slack right now for our team for communications. We have some AI that's embedded in that. AI is able to capture that information and add it to the appropriate kind of database or whatever that might be.
Here's a fun example for you, Noah. We had a young lady that worked a double shift on a particular day, and AI, one of the tools I had set up was, if anyone is heading into overtime hours, they're on track, or if they keep working the rest of their schedule, they're gonna be over 40 hours a week, just send me an alert. It doesn't mean I'm gonna make any changes. I just wanna be aware of what's going on. We had one lady that actually got flagged as heading into overtime. And so that tool was telling me, like, "Hey, Jon, red flag alert, something potentially negative." I don't wanna say negative, but, like, something you're wanting to manage, which is overtime, is coming up. Go take some action. Now, the next week I asked AI who on my team is being overlooked.
It recommended that same lady, and here's the reason. She worked a double shift because she was covering for one of her teammates, because her teammate had to go to the hospital with her son to get some medical treatment. That comment that said, "Team member A is going to the hospital with her son, team member B is gonna cover for her," that happened in our Slack app in the attendance channel. I don't read the attendance channel. That's for our shift leaders to figure out their shift and who's here, who's late, who called out, all that kind of stuff.
So two things happened, Noah. One is it said, "Go encourage her. She just worked a double shift to support her teammate." Again, AI last week told me to flag that. That was a bad thing. AI for leadership this week is saying, "Jon, go encourage her. Give her a thumbs up and say, 'I see you. You worked 16 hours last week. Great job to help your teammate.'" And then part two, who the heck checked on our other teammate who's at the hospital with their son? Oh my goodness. I didn't-- I had no idea.
And so AI, we gave it the kind of the context of what we were looking for: find leadership opportunities. And it found this leadership opportunity, and it allowed me to go up to the first team member to give her a pat on the back, give her a hug and say, "Great job. We see you." Ironically, she does not speak English. She only speaks Haitian Creole. So we were able to pantomime some things together, and I sent her some messages in her language on Slack. And then part two was to go up to our other team member and just say, "How are you? Like, how's your son? Is everything okay?" Uh, you know, "Big medical bills? Are you, like, stressed out, strained? Like, how can we help? How can we better serve you?" And it turns out everything was okay. But it was just two great human moments that were created as a result of AI nudging me as the leader to go and have those human moments. So that was kind of the aha of, like, wait a second, I think AI can do a lot more than just productivity. I think it can help us leaders lead better.
[00:28:09] Noah Levin: Something that I use as a framework thinking about AI is, there's a human input, and then there's AI kind of doing its computation in the middle, and there's a human output. And I want to ask you about the input and output of that story. You were able to use AI to surface both this scheduling opportunity and also this very human opportunity for you to, you know, intervene as a leader and offer support to your team. How did your AI tools get the context to know about those things? And then I'm very curious, you as a leader, how do you make sure that going over and offering someone empathy and support doesn't come off as ingenuine when it's being kind of prompted by AI or nudged by AI on your side?
[00:28:53] Jon Harmeling: I'll answer your last one here first. So please don't allow AI to replace the human moments, and please don't allow AI to be the leader. It's almost like on our birthdays, I'm sure a lot of us get very random postcards in the mail saying, like, "Jon, we're so glad you're a customer. We wish you the best birthday. We value you so much. From Bob's Landscaping." Or, like, something, and it's like--
[00:29:17] Noah Levin: I don't mind when my local Chick-fil-A sends me a, a free sandwich coupon, for what it's worth.
[00:29:21] Jon Harmeling: That's true if you have the app. So I'll plug the app right there. Make sure you download the Chick-fil-A app. We do send out a heck of a lot through the app, which is a good point. That's a welcome one. But in that example, it's just like, it's such a disingenuous... I don't even know who Bob's Landscaping is, much less, like, the CEO. You're telling me I'm a valued customer, and I hear from you, like, once a year. It's just, like, throwaway spam kind of stuff. And what you don't want to do is have AI send a message and just say, "Happy birthday, Noah. We're so excited you're on our team. You're a valued member of the team," sent by Claude AI. It's like, come on. Like, did Jon actually write that? Or, like, clearly he do-- Jon doesn't even know that it's Noah's birthday, but AI made it look like it, it was a message from me to you. You know what I'm saying? So just be very cautious with some of those things.
Now, what I think you can do, and I would encourage folks to do, is when you wake up and it says, "Hey, Jon, if you want to be a good leader today, remember it's Noah's birthday." And it's like, "Oh my gosh, that's right. I have 90 people on our team. I can't remember everybody's birthday. Thanks for reminding me." And so that way, when I see you on the floor, I'm like, "Hey, man, you're working on your birthday?" You know? And just create that kinda human moment between us. So don't allow AI to replace those human moments. But I would suggest use AI to help elevate and highlight those human moments and maybe even make you better as a leader in those moments.
And then back to your first question, one of our challenges that we've had: there's so much amazing data and information out there. It's not sensitive information, it's out there, but it's on so many different tools, on different collections, repositories, websites, text message, all that kind of stuff. And so one of the challenges I noticed is we have forecasted sales, and therefore that turns into a forecasted schedule, and then we have actual sales, and then we have the actual people that showed up and clocked in and worked. All of those are moving pieces, right?
But one of the challenges is, how do we bring those together? I wanna know what we forecasted. Was that actually what happened, whether it's sales or schedule or transactions, type thing? So that was one of the first use cases for me, because one of our biggest opportunities for improvement is creating a very efficient schedule. How do you create a very efficient schedule where it factors in how much I need to start? How do I ramp it up to lunch rush, ramp it down, ramp it up for dinner rush, ramp it back down?
Again, forecasted schedule plan, you know, our plan, and then what is reality. So I used AI to try to bring these together and said, "Hey, Jon and team, you guys are actually over-forecasting on, like, Mondays and Wednesdays, and you're actually under-forecasting on Fridays and Saturdays."
I was able to pull off a very busy day with very few people, with, you know, not as many people, and we saved money on labor. That's great on a spreadsheet. That's terrible for your team that is being overworked, and now morale is going down. They're feeling burned out, and they're questioning if they want to even continue working here, because there's a good chance they could go down the road, get a different job, potentially roughly the same pay.
And so we used AI to try to pull all of these data points together. And so every morning, I have a daily brief sent out to our leaders that says, "This is what happened yesterday in comparison to what was scheduled. This is what we think is gonna happen today, and then this is what we think is gonna happen tomorrow." And so it allows us to kind of get this feel for plan versus reality. So to answer your question, all of this was very intentional data and information, and AI just brings it all together in one nice executive summary for our team. So that's how we start our day, is just reviewing what happened yesterday, confirming the plan for today, making any changes we need to make for tomorrow.
[00:32:53] Noah Levin: When you talk about scheduling, and I think ordering was the other one that's forecast-driven, you mentioned those are two things that you ask your team to do that are at the desk, not on the floor. What do you expect your team to be able to do with AI? And how do you, as their leader, empower them to use the tools that can help them do their jobs?
[00:33:15] Jon Harmeling: Yeah. The biggest opportunity is scheduling, and it's been the biggest flop for me personally. It's, it's driving me nuts. And so how do we use AI to create a schedule? And one of the things that was very difficult is when we started to try to figure out, like, "AI, you just do the schedule for us, 'cause it's taking us leaders five hours a week to try to figure out, like, okay, Noah just started classes, so he's not available anymore on Tuesday afternoon. He is avail-- Oh, no. Shoot, it's football season, so he's got football. They're off. Their family's on vacation this week." And these puzzle pieces that we're trying to put together keep changing every single week. And so it was just really hard to use AI to try to put together the schedule. But what it started to do is it started treating every single person as basically a mathematical computation and just shoved it all together.
And it just, even though we would do all these different iterations, it just lost so much of the context, 'cause I know Noah and Jon work really well together, or I know that Noah is actually one of our experienced shift leaders. Jon's a new shift leader and I'm trying to become better at it, so maybe I'm gonna shadow you today. So AI is losing that context, that human context. And it was just really tough, 'cause it would spit out a schedule, and it's like, mathematically it was flawless, but just when you then add in the human element, and it's just like, man, that's, uh, it's just not gonna work.
We also have to take the schedule and then input it. It goes into our pre-approved Chick-fil-A systems as well. So that's a little bit of a constraint. We build the schedule over here, but then the time punches and everything else go into these Chick-fil-A systems. My next project at some point I'm gonna work on is just building almost a custom schedule app that is only for us and our team, where I think we can just give it all of that kind of context.
And then here's something crazy, is, what about having the team members select their own shifts? And so here's all the shifts for the week. You know, if you want to work as many hours as you can, Noah, this week, just pick all the hours, you know, all the shifts that you can. If you want to have a chill week and maybe you have a bunch of exams coming up, got it. Just only work when you want to work. So I don't know. Uh, I--
[00:35:20] Noah Levin: What's cool hearing you talk about this is some of these things are like modules that you can go and buy from, you know, scheduling providers that make SaaS programs that have the team-member-facing widget where you can let the people opt into their pref-preferred schedules or help you f-flex wages up and down based on how hard they are to staff. But some of the things you described are probably not off the shelf in any real way. Like, when you look at a schedule and you can intuitively just say, "That guy's not gonna wanna work. It's his first day back after a long trip abroad, and, and these two people work great together. Why do we have them on split shifts?" And you can give this qualitative narrative over it that doesn't really model out in any reasonable way in, like, a, an off-the-shelf SaaS program. And for you, your mindset is you don't have to care which of these things are features you can buy and which of these things are being invented out of thin air.
You can just ask for what you need, and the tools that are available to you will bring something back, and then you can just iterate on it from there. I think what you're about to bite off sounds like an incredibly challenging project, but one that's, you know, it's within your reach for the first time in history. R- you know, instead of having to do it on paper or go and buy an expensive off-the-shelf program that only does half of what you need.
[00:36:31] Jon Harmeling: That's right. So at some point we'll get around to the schedule. What we're doing now works, and it works just fine.
[00:36:36] Noah Levin: Well, what I think is so amazing about your approach to running your business and to your use of AI in the business is that there's just this cascade of humans caring for humans, right? Y- and you've described it throughout the interview, but just to play it back to you in sort of one breath, you've got frontline team members taking care of guests, and you've got your team, your leaders taking care of your frontline, which you're enabling by making yourself more efficient so that they have more time to spend with the team, and yourself more efficiency, you can spend more time with the team. And so there's this whole cascade of people supporting each other from behind. And it's such a human-oriented story that I think it is easy to dismiss how much is happening under the surface, how fast those swans' legs are paddling to keep all that going. It's just what AI is really good at, is taking care of the mundanity so that humans can be human to each other. It's a little glimpse of, I think, like, the optimistic version of where this could all be going, right? Is that humans can get back to being present for humans.
[00:37:51] Jon Harmeling: You nailed it. But also to connect it to, you know, a business sense. So people, it's-- we're about 30 minutes away from lunch rush starting as we record this. I'm not at the restaurant right now. And so when a guest comes to the restaurant, what are the odds they're gonna run into the operator? At least today is zero. I'm not there. I might go in tonight. But what are the odds they're gonna run into one of our 16-year-old team members? But there's a higher likelihood that they're gonna engage with, uh, one of our frontline team members.
So I just think it's so important to, like, lean into that. Let's make sure they're the best versions of themselves. Let's make sure when they show up to work, they're not like, "Ah, I have to go to work." But it's more like, "Yes, I'm excited to go to work. My best friends are at work." And we have fun there. Like, yes, we work hard, but there's always moments to maybe just kinda laugh or be silly, or something crazy is gonna happen, and we're gonna figure out a way to get through it. And so-- and that also-- so that establishes a great guest experience. We don't nail it every time. We definitely screw up. We've made plenty mistakes.
But I think on average, a lot of folks dine with us because they're expecting that great service. Like, I hope the food is delicious and it's fast and it's accurate and all those things, but I hope you feel good. I think when you come and dine with us, and people leave, I think the next week they may or may not remember what they order, but, "I remember when I went to that Chick-fil-A place, I felt good. I want to go back there." And so I think a part of that is just leaning into those frontline team members. And that requires me to lean into our directors, who lean into our shift leaders, who lean into our trainers, who lean into our frontline folks. We call them team members. But here's the other piece too, Noah: retention.
This is not why we do all of this stuff, but I do think there's a very clear parallel with team members that feel heard and valued. They're developing, they're getting coached, they're encouraged. They have friends at work. They know the purpose of our organization. They see kind of the higher calling of really making a difference in people's lives. All those things lead into engaged employees. Employees that don't-- when they clock out, they're still thinking about work. I get texts from young adults all the time. They're not even on the clock getting paid, and they'll be like, "Hey, Mr. Jon, I was just thinking about something. Is there any way we can move the sauces from here to there? Because I noticed I have to walk a little bit extra to get the sauce to hand to a guest." Yeah, that's amazing. That's an engaged team member. And when you have engaged team members, they tend to want to stick around longer than the average person; therefore, retention goes up and turnover goes down.
And so I, I do think it's this element of a flywheel. Like, yes, I get it, that sounds beautiful on a PowerPoint slide, but I really do think in, like, real life, if you can lean into these principles and AI can help you be a better leader, I do think you're gonna see guest service go up. I think you're gonna see your reviews increase. I think you're gonna see your retention go up. I think you might see your turnover go down. And that's all things that we've been experiencing at our restaurant. We're so blessed and so fortunate, please don't get me wrong. But, man, that's really, really cool. And we haven't even really leaned into how our leaders can do better using AI. We're on the cusp of that right now.
[00:40:44] Noah Levin: Tell me about the book you're writing, "Frontline AI Leadership." What's the book about, and what's the lesson you want people to take away when they read it?
[00:40:51] Jon Harmeling: So first off, not trying to sell anything. If we have a chance to sell some books down the road, all the proceeds are gonna go to my actual frontline teams. This is genuinely just, like, I wanna, I wanna help. We've learned a lot of stuff over the last couple years, and so I think a lot of people are talking about AI. God, a lot of people have been and will continue to talk about leadership. But I think this combination of not just AI and leadership, but specifically in a frontline environment, meaning deskless or not corporate or not kind of a business setting or cubicles where everyone's on a laptop every day. We only have four or five people out of 90 that even use a laptop for their job.
And so how do you be a leader in this era of AI in a frontline environment? And so it's been really fascinating. A lot of the stuff we talked about today, I try to capture some of those lessons learned in the book, but it talks about just kind of my journey of using AI as a, as a thought partner and then moving into automation and then agentic AI. And then how do you create, like, a full stack where you have an AI agent that's kind of like your marketing person, you have an AI agent that is your labor cost, you know, person, and they all work together, and they're coordinated by an AI agent called chief of staff.
And so one of the reasons I wanted to write the book is just, I've had a chance to meet with a number of different business leaders, whether it's at, like, through a church group with some business folks, whether it's around the table at Chick-fil-A or just even in these kind of podcasts. As soon as I start talking about, like, using AI not just for management but for leadership, and I, I use those terms very specifically, right? Managers are tasks and transactionals and deadlines and that kind of stuff, and then leaders are influencing people. They're coaching, they're developing, they're encouraging. How do we use AI to be a better leader? And that to me is what is just super fascinating. So I talk about some of the tools I'm trying to build right now, whether it's a goals database, it's a system to help, you know, a leadership development system. How can we coach people to add more responsibility onto their plate?
And then if you think about it, this actually, this whole system, AI is gonna help create more capable people and a more capable, you know, workforce. As that flywheel spins, it's just gonna create even more capacity down the road, 'cause they're able to do more in less time. So I just dive into the book a little bit with some of those tools and just really excited. We're somewhere around three-fourths of the way finished writing it. I think in October we're gonna send it out there. Keep an eye out for it. It's gonna be called "Frontline AI Leadership." Stay connected with me on LinkedIn if you want, and I'll post some more information about it.
[00:43:14] Noah Levin: Sounds good. So folks can find you on LinkedIn, and if they want to see you during the lunch rush, where can they find you?
[00:43:20] Jon Harmeling: So our restaurant is Chick-fil-A in Miami Shores, so very close to downtown Miami. We're just about 15 minutes north of it. So if you're coming to hang out at South Beach or cheer on one of our sporting teams or just passing through Miami Airport, I'd love to host you. Ask for the bald guy when you get there. They'll know where to direct you.
[00:43:37] Noah Levin: Jon, thank you so much for coming on Serious People. I appreciate everything you're doing for your team and for the world.
[00:43:42] Jon Harmeling: Yeah, my pleasure, Noah. Good hanging out.
[00:43:44] Noah Levin: Thanks for listening to Serious People. If you want more conversations like this in your inbox every week, subscribe to Serious People on Substack at seriouspeopleai.substack.com. You'll get the episodes and practical notes in your inbox, along with something to read and something to try. And if you're trying to make AI useful inside your own business, Serious People can help you find the right work, build the system, and put it into production with your team. Find us at seriouspeople.ai