AI Employees for HVAC Businesses to Automate Booking & Increase Revenue (Part 1)

Who is James Christian and what is Part 1 about?

Gary McCreadie: What's up, guys, welcome back. I am remote again doing this introduction. It's been crazy out in the field the last two or three weeks, just call after call, so I'm recording this intro from my phone in my van. AI is a big topic right now, so we have a guest on to talk about AI and how it applies to the technician, the business owner, all of that wrapped into one. His name is James Christian, and he's from a company called Podium. We'll talk about what Podium does later on in the podcast. It's another two-parter. If you're interested in what AI can do to help you with your day-to-day, whether you run a business or you're a tech in the field, there are things out there that can make you more efficient at your job. So let's get into this. This is the HVAC Know It All podcast, and I'm your host, Gary McCreadie.

Gary: Welcome to the HVAC Know It All podcast, recorded from a basement somewhere in Toronto, Canada. Your host and HVAC tech, Gary McCreadie, will take you on a deep dive into the industry, discussing all things HVAC, from storytelling to technical discussion. Enjoy the show.

How fast is AI evolving in the trades?

Gary: Jim, so here's the thing: AI has come a long way, and every single day it seems like there's a new advancement, or something that was cutting-edge yesterday has improved today. It happens really, really quickly. I want to have this conversation with you about how an HVAC technician, HVAC business owner, or any professional in the HVAC and refrigeration skilled trades can utilize AI to their benefit, to be better at what they do. There are all kinds of tools out there. On the technician side, Copeland Mobile has an app that uses AI to sift through years and years of compressor data, so you can put in a model number and pull up all the specs on a compressor that's twenty years old. Things like that are really helping technicians in the field, and they can help business owners too, but from a business owner standpoint, I think that's more along the lines of what we're here to talk about with you today. Let's talk in general — I just want to get your opinion on how AI has evolved over the last couple of years. Give me your general take on it.

Jim Christian: Yeah, for sure. Well, Gary, first of all, thanks so much for having me here today. Really appreciate the opportunity. Nice to meet you.

What is a large language model?

Jim: My general opinion on AI — I think you're totally right that it's advancing wildly fast, probably the fastest of any software technology in the past few decades. The arc I've observed with AI is that we started with these traditional LLMs, these frontier models that can communicate via text.

Gary: You might want to define LLM, because there are a lot of people listening who might not know what that is.

Jim: LLM is a large language model. You hear about ChatGPT's models, or Claude's models — those are all LLMs. They're large language models that have taken in a bunch of information from the internet and other sources and put it together. The actual core of an LLM is pretty simple: what it does is predict, for every word, what the next word is. That's a relatively simple view of what's going on, but what ends up happening is the first LLMs were able to piece together and make responses based on questions, like you still do in ChatGPT, where you say something to it and it responds back with a reasonable explanation. Then we started adding the ability for these LLMs to reason and make sense of what's going on, go search for the right thing, understand the intent of the question, parse out what's happening there, and reason through it. So in the initial days of AI, the very first applications were things like an AI CSR — being able to receive a message from an inbound lead and respond back with something reasonable, generating natural human language. The bigger advancements lately have been in AI models that can actually do real work for you — not just the LLM layer that's the communication interface, but models that can reason, understand what to do, and then go do real work. We call those tools. A concrete example of that: the very first AI introduced to the HVAC and home services industry was AI that could text-respond to a lead. That's where we started. Over time, that's evolved into AI CSRs that can do more than just text and respond to a lead — they can actually go into the software and take an action. After negotiating a time with a lead that makes sense, they can go into the software and book that job for the customer.

Can AI function as a virtual employee?

Gary: So that makes sense. Can we define that as a virtual employee, almost?

Jim: A hundred percent. Yeah, that's exactly how we think about it. If you think about where the value is, oftentimes these AI employees, these virtual employees that contractors are buying, are a replacement or an addition on top of the employees they would have paid for.

Gary: I see what you're saying, because I've been in the industry for a very long time, mainly with the same company for the majority of my career. I saw different dispatchers come through that company. When you say the AI virtual employee is there to assist the actual person, I can see that being so beneficial, because I've literally seen dispatchers answer a landline and a cell phone at the same time — tell one caller to hold on, put them down, talk to the other person, then get on the computer, trying to multitask three or four things at once. That leads to mistakes — sending somebody to the wrong place because they weren't looking at the GPS map properly, and that technician getting pissed off because he had to drive two hours across the city when it could have been someone closer. It leads to mistakes and burnout. So I can see how a virtual employee could really help assist an actual person. Let's speak more about that, if you don't mind.

Jim: So many thoughts I want to talk about there. The first thing that comes to mind is, when I started looking into AI employees and building some of these tools with a team, the original angle was a lot about replacement of people. What I've learned in diving in really far with home services businesses is that it's not about replacement so much as it is about augmentation and helping businesses do more. Think about this as a longer arc in software: before businesses were using any kind of software, they were running pen and paper or whiteboards in the office to manage technician schedules, everyone getting together at the beginning of the day, figuring out where they're going to go, writing down notes, and then going. There was a whole era of software where we started bringing things in — technicians could have information on their phone, live updates, automations to take care of simple things like sending a review invite after a job gets completed, or following up on an invoice that needs to get paid. Those automations were there, and these AI employees are kind of the next level of that. I work with tons of customers who have one office manager or dispatcher who's totally overwhelmed, too much going on, and they spend all day just keeping the technicians enabled and going throughout the day, with no time left for the other revenue-generating things you want to do for a business — following up on unsold estimates, building your membership program, doing marketing to help drive leads, looking ahead on the schedule to see where things are going to slow down, and making a plan to keep those technicians busy and the revenue coming. AI employees really help with that.

How did Podium evolve from reviews to AI employees?

Gary: So you're with a company called Podium. We're not going to turn this into a commercial for Podium, because that's not the purpose here, but you do offer services that can help. I want to give you a couple of minutes to describe what Podium does, and then we'll get into some other topics when it comes to AI.

Jim: Let me tell the quick story of Podium, how we've evolved. Roughly fourteen years ago, our founders got started with the idea that locally run businesses mostly get found on Google Maps. When we need support from an HVAC business, we go to Google Maps, search, and start calling. The only way businesses get found on Google is by having reviews. So that's where Podium started — we were a review solicitation company helping local businesses quickly get a review, send out that review invite, and get something from the customer to help build their online reputation. Over time, we expanded into being a full communication software. We did review solicitation over text initially, then expanded into email — phone is definitely the most important channel for communication for an HVAC business. What became really evident as this communication and lead management software evolved was that the most important thing for a business's lead management is their speed to lead. In the home services world, the stats say that if you don't respond to a lead within five minutes of them reaching out, your chance of winning that lead drops by eighty to ninety percent. That's where we started building — that's where the idea of AI employees was born for us. The great thing about an AI CSR is it's twenty-four seven, always on, always working, whether that means handling overflow, like you mentioned, the dispatcher trying to pick up two phones at once, or working after hours. AI employees never have a bad day. They're always positive, they always run the best playbook, they always talk about things the right way, and there's nothing else distracting them from what they're doing. That's how we got into AI employees. Where Podium has really evolved in the last twelve months is the realization that an AI employee that can simply pick up the phone and talk to a lead is, in a way, a little bit of a glorified voicemail. If the end of that conversation is, "Hey Gary, thanks for reaching out, I'll take notes on your problem and have someone call you back," that's something — it's better than nothing. But the best AI employees can actually say, "Hey Gary, I'm so sorry to hear your furnace won't turn on. Jim is actually on call tonight, he could be out to your place in thirty minutes — do you want me to send him your way?" In achieving that, that means we have to be tied into the entire operating system. We have to know the technicians, what their skills are, what types of work they do. We have to know what our price book is. We have to know what estimates and invoices look like. We need context on that customer — have we worked with Gary before, do we know his first name, do we know if he's the homeowner. That's the journey we've been on.

How does AI match technicians to calls by skill level?

Gary: You said something interesting to me that I'd like to revisit — you said we need to know who the technicians are and their skill level. In a company, you could have a technician who's been there for thirty years and can fix anything under the sun, and you could have Jimmy, who started two weeks ago and they just threw him on call rotation because they needed to — no one else wanted to fill that gap, the ins and outs of running a business. How would you tell AI that your thirty-year tech can work on anything under the sun that comes in on a call, but the new guy, you can only send him on simple residential calls, you can't send him to big chillers or something along those lines? How do we implement that information? How do we make the AI know who we send and why?

Jim: That's a great question. There are two parts to the answer. The first part is a really simple, boring one, which is: in your system of record, in your field service management software, you define who the technicians are, what skills they have, what jobs they can do. That's just your standard database setup — go ahead and configure that Gary has thirty years of experience, Jim has two years, here are the types of jobs we'll give Gary. That's all pretty straightforward. Then what the AI employees can do is start understanding things while on that conversation — start classifying what type of work needs to be done, how much uncertainty there is. It's even better when you start having information like, in your system of record, how old is the system in this house? Do we have that information from past work we've done for them? Who did the original job? Has the customer ever expressed interest in having a certain technician? Is there a normal guy who goes out there and knows the system? From there, that becomes decision-making you can actually train your AI employee to do. The last part — the quick, high-level version — is the way we look at building AI employees is we look at them as trainable employees. We always think about how a business owner would interact with a human doing this job. There are key components to that: there's initial onboarding out of the box — does the AI have the kind of handbook for how it's supposed to work? And then there's feedback, testing, and performance cycles. Can I give natural language feedback to this AI employee — say, "Hey, I didn't like how you said that," or "You booked Jim on a job he shouldn't be doing"? The AI employee can take that information, go update its own rules and understanding of how things work, and then play it back and say, "Hey, I made updates — do you want to test me out to make sure I'll handle this scenario right the next time?"

How does AI use GPS for real-time arrival estimates?

Gary: You said something else interesting when you were defining which technician would go to the call — you threw out a time there, "such and such tech will be there in thirty minutes." Do we connect the AI to a GPS of where that truck is at, so it knows how long it's going to take that technician to get to the site?

Jim: Yeah, a hundred percent, that's exactly how you do it. I think this one, Gary, is actually pretty interesting, because in the traditional, fully human-powered office model for an HVAC business, you've got CSRs who are picking up the phones and getting people booked, and dispatchers who are managing and optimizing the route. It's this intake process, then optimization, then sending out and dispatching.

Jim: What's great about AI employees is they can handle all of that on the fly — it's a computer doing the work, so it can take those things in, and you can actually start optimizing as it goes. For example, you can set simple rules — you can train the employee with things like, when picking a tech, if there are multiple technicians, pick the one who has the shortest drive time, or never book out more than thirty minutes without escalating to a human for approval. There's one last thing that's really interesting — we've started building more of an internal agent that can help. You can kind of think of it as, dare I say, Claude Code — the way we talk about it in the tech world is think about it as Claude Code specifically for an HVAC business. The other way I think about it is an internal office assistant who isn't interacting directly with customers, but is doing behind-the-scenes work. You can prompt it and say things like, "Hey, look at my schedule next week, find any inefficiencies, and propose changes to optimize our routes to minimize drive time — make sure you don't assign techs to jobs they can't do." It'll go off and do all that analysis, which is oftentimes a pretty complicated set of things to do. In the traditional model, you have to run different scenarios and permutations to figure out the best options. The power of these AI employees is they can do all of that behind the scenes, basically for free, and then spit out options for review and approval. Then they can actually go back and do the work — shift around the schedule, communicate with the technicians, communicate with the customers about the changes, and even set it up to do on repeat.

What's Gary's reaction to how far AI has come?

Gary: This kind of stuff blows my mind. It really does. It makes me wonder if the Terminator movies were onto something with the Skynet thing, and how we're all going to get attacked by robots. But I'm just kidding — we're a little ways away from robots. I do appreciate the technology, and I appreciate how it saves people with busy lives real time. I don't know what it is, man, but I seem to be getting busier and busier daily as I get older. It seems like I have free time maybe at nine o'clock, I sit down, I'm awake for about twenty minutes, and I'm fast asleep. So anything that can make someone's life a little bit easier, more efficient, and take some stress off, I think is a really good thing. Now, I don't want to beat this phrase to death because a lot of people use it — working on the business instead of in the business. I think that comes from the book that Michael E. Gerber wrote. I believe it's called The E-Myth.