You’re Not Behind (Yet): How to Build an AI-Powered Business Before 2027
Phil Risher shows home-service businesses how to pair employees with AI agents that act as co-pilots, potentially doubling output. He walks through three working examples: Trace matches leads to CRM data and calculates ad spend ROI, Turner pulls questions from call transcripts to build SEO content, and Victor connects to Google Docs, Sheets, and QuickBooks to handle tasks across the business. The key is documenting your processes first, then launching agents with human approval loops.
Chapters
We have a team of 24 people, and today every single one of my team members have an agent that works alongside them as a copilot. And over the last 12 months, we've built and deployed AI agents inside of home service businesses to do real work, from tracking return on ad spend to creating content to rank in AI search. And by the end of this video, you'll know exactly how AI agents fit inside of your home service business, how they're different than basic automations, and how to deploy them inside of your home service business without trying to replace your entire team. I'll show you the 3 agents that we're actually using today. And then I'll walk you through the framework that we use to decide what should become an agent. And finally, I'll walk you through a real process from manual to automated to agentic. So that way you can see how to do this inside of your business. So what's the difference between an automation and an agent? So automation is when something happens, a predetermined thing happens next. So let's say someone fills out a form on your website, then an instant text message and an email goes out to them. That's an automation. You should be doing that, but that's an automation. An agent is something that happens and then there's actually a decision process that comes from it and then action is taken. So for example, let's say a call comes into your business. An agent analyzes the call transcripts from the CSR based off of the SOP that you provided to it. Then it gives your CSR feedback and then goes to the office manager and says, hey, here's a weekly report from your CSR with things to work on for them in their next one-to-one. This is a fully agentic process because there's decisions that happen. So the way you can think about this is a trigger happens, intelligence layer happens, and then an action happens based off of the intelligence that comes from it. Different than an automation, but now it's fully agentic. So why is this so important? What we found is that you can actually double, if not even triple, the output of, of your employees when you have agents working with them, because every function in your business runs on a trigger, intelligence, and action. A CSR gets a call. They use their brain and intelligence to try to book that call or answer the customer's question, and then an action is taken. It's either booked or there's something that needs to be followed up with. But every single thing in your business, marketing, sales, bookkeeping, accounts receivables, everything in your business has this trigger, intelligence, action. And what we're looking at when we're building a truly AI agentic home service business, we're looking at this. Employees alone become employees plus agents. It's not about replacing your employees. It's about enabling them to do much better and more work. Using the agents that they have and deploying them inside of your business. So the way that we think about this is, let's just give you an example, accounts receivable agent inside of your business. So a manual process, this was at my home service business, was someone watches the 30, 60, 90-day aging report inside of QuickBooks and reaches out by hand once a client falls behind. At the 30-day mark, this person reaches out. At the 60-day mark, this person reaches out. At 90 days, this person reaches out. This is the exact process that we use at my home service business. Now, when we're thinking about an automated thing, it's like, okay, on day 30, it hits a trigger and then an email goes out that says, hey, you're 30 days past due. It's time to pay your bill. That's an automation. Now, what does an agent actually look like? An AR agent doesn't just email everyone on day 30 like an automation. It learns how each customer pays. It decides when and how to follow up. It handles the follow-up and then it tells you which account actually needs your attention. So imagine there's a, you know, an HVAC business that owes you money for a subcontract that you did or something. And then it reaches out to the office manager on day 29 in the morning via email, because it knows that that person usually responds in the morning when you email it. And then it gives you a synopsis every single week of where you're at and what you should be doing based off of, did they open the email? Did they respond? Like, that is an agentic process for ARs that then gives the information to the office manager, but they, there's no automated or manual process that's needed because it's making decisions on its own. So the way that we think about this inside of our business and the way that you should think about this inside of your business is where do these people, where do these agents fit in? And there's this framework, it's from Mike Michalowicz, from his book Clockwork. But this is how I think about business in general is one, attract. So marketing, can you build agents to help you with marketing? Yes. And I'm going to talk to you about some of the agents that we built for this, but think about like content creation agents, return on ad spend agents, B2B partnership agents, right? There's a bunch of different agents that happen from manual processes in your business that then you can build agents for. Conversions. Imagine you have sales calls that happen. Well, do you listen to the sales calls manually, or do you have an agent that knows the SOP and then is analyzing the actual sales calls and then giving feedback, right? We have this agent inside of our business that listens to the sales calls, gives feedback based off the SOP, and then every single sales call gets graded. And now we have talking points on how to get better. Delivery, doing the work. So inside of my home service business, we actually had one person that was designated for dispatch. So every, we had 13 routes every single day. One person in the CSR team would go in and look at the next day's routes and then try to optimize them. They would take the addresses, right? They'd optimize the addresses to make sure that there was not that much time missing, right? This is a delivery function. That's a manual process, but this could be an agent. And then collections, we just talked about the AR agent, but there's a million different types of collection agency agents that would happen based off of what's manually happening in your business. An AI-powered home service business layers agents on top of employees across all 4 of these things. So before I show you different agents and how we built them and all that stuff, let me just walk you through a basic framework for how you can build your first AI agent if you never built one, or how we think about building agents inside of our business. So the first thing is, what's a manual job that happens? What does someone on my team do over and over again? I mentioned about the dispatch thing. Let's just run with that one. So every single day, someone would go in and look at the dispatch board the next day. And they would say, okay, this address is here. This address is here. Let's move these. They would go to Google Maps and just try to figure out how long things took. And it was a very manual process. Then you want to define done. What outcomes are they responsible for? So my CSR knew that the outcome was we want all the routes, the route density to be within a 15-minute stop from each other. So then it wasn't driving all over God's green earth to try to get the jobs. It was like, hey, The goal is we have 3 jobs a day, 15-minute increments between each job. This was an efficient route. Now here's the most important part of this equation. And I would say, if you're thinking about building agents, you should do this, which is explain the SOP. Here's exactly how we do it today. What we do is we record a Loom video, L-O-O-M, and we walk through the entire process. This allows the agent to see how we are doing this manually, so then it can build how to do this itself. And the, what I can tell you about agents is that they're getting so smart, they could even like get on your browser and do all the stuff, you know, that you would normally be doing if you just show them how you did it and you can just walk them through every single step of the way. Now, I'm not saying you have to go that far, but at least you can have a transcript of exactly how you do it that you can give to it or a video of how you do it and then give it to it. So explaining the entire SOP, uh, SOP stands for standard operating procedure, and this is the exact steps that you take. Now here's where it gets fun. You go to an AI platform, let's say Claude, ChatGPT, whatever you like to use, and you say, hey, what would you need to see, decide, and do this yourself? Here's my loan video. Here's our SOP. What do you need access to? How do you do this? Now, when I talk to people about this, they say, okay, well, I don't— you want it inside of my CRM, inside of my dispatch board, moving my jobs around? Like, no thanks. I don't want an agent for that. I'm not saying that you have to do that. Because the next step is the most important piece, which is version 1 has human approval. So if this was my example, I would literally just screenshot my dispatch board. I would build an agent that knew exactly the goal and on my dispatch board, I would have the addresses and I would screenshot it. And then I would say, hey, take this screenshot and re-optimize them and put them into a Google Sheet every single day. And now I don't have to have my CSR sitting there going through for an hour or 2, looking at all these addresses and trying to move things around. My agent can do that in, in 2 minutes, and then my human approval can go to that spreadsheet and say, okay, cool, this all looks good. Let me go through and just move these around. And now we have a quick process. This is how you effectively layer agents inside of your business. When people hear agents, they're like, oh, it's going to go into my CSR, my CRM and move all these things around. I'm not saying you have to do that. I'm saying that you can use tools like this to then simply deploy them in version 1 with human approval. And eventually maybe you work up to the point of it going into your CRM. But in that example I just gave you, that could cut down, you know, 2 hours of your time. Could that CSR be answering phones? Could they be doing something else inside of the business to help it grow? Yes, of course. All right. So here's some of the agents that we actually use just to give you some ideas of how we do this inside of home service businesses. Um, and then we're going to also talk about how to actually build these things and what you should look for whenever you're implementing them. So the first one that we built was called, his name's Trace. I really like naming agents because it allows you and your team to feel like they're part of the team. When you just have some process or something that's floating out there, it doesn't feel like it's really a part of the team and it doesn't feel like they have a buy-in or a way to use it. So when we have names to it, it allows us to be able to say, hey, okay, Trace is going to do this. Turner's another one. They're going to do this. So for us, we do marketing for home service businesses and probably the number one question that everyone asks is, How do I know what's working? How do I know my customer acquisition costs? How do I know my return on ad spend numbers? And we said, okay, well, we did a manual process for this, so let's build an agent that does this. So how we built Trace, I'm just going to walk you through the exact steps as well. So manually, what we used to do was we would go and find all the leads from a different source. Let's say it's Google, and we copy and paste all those leads into a spreadsheet. And then someone would log into a CRM and they would go and say, okay, this phone number, let's go to the CRM and see what happened to it. Email, let's go and see what happened to it. And they would piece together what happened to every opportunity by hand. Obviously, this is very manual, but someone has to do this if you're going to do good marketing and know what's actually making you money to help you grow your business. Now, when we look at automating it, it's like, okay, can we have those reports with the leads dropped into a spreadsheet so then we don't have to copy and paste them? And then can we have the reports over here that we need dropped into a spreadsheet or even sent to us by email? So now we have 2 reports that can, we can look at back and forth and clean things up. And leads started to flow into spreadsheets. This was an automated process. Now, when we think about an agent, it's like, okay, an agent sits between those spreadsheets and also connects with the ad platform. And now it can look at all the stuff, say, okay, this phone number and this phone number and this email matched up here. This person doesn't have an estimate, but this person has an invoice. So our potential return on ad spend was this, and our actual return on ad spend was this. And from our Google ad spend and the campaigns we're doing, here's the breakdown. And now it sits in the middle and can use an intelligence layer. And again, with us, we still have humans in the loop of all these things, but it allows us to work much more efficiently and allow our team to focus on what they do best, which is consulting. And then it provides the data to them and the analytics to be like, hey, okay, here's what I think we should do. Here's what we should do. This is exactly how we built Trace. This is how you should think about it for your business. So what Trace does, pulls the leads in from your ad platform, matches those leads. It used to be manual, like I mentioned. And now Trace does this matching everything. So you never have to guess what your return on ad spend is. Another agent that we built was Turner. So when you think about AI search content, it's like, okay, how do I create non-commodity unique content? As a marketing company, we had to ask this question was, how are we going to create real content when we talk to our clients on a call once a month and we talk to them via text and email throughout the month, but like, We're not sitting inside of their business. And my home service business, I had 28 technicians. I could barely get them to send me pictures, let alone tell me stories of the things that happened out on the job site or on the phone calls. So I was like, okay, well, how do we do this? So we built a manual process, which we call the call to content system, which is the marketing coordinator on our team would go and find call transcripts that were over 3 minutes, take those transcripts to AI, and then create unique content. This is what we were doing a year, a year and a half ago. This was a manual process, but it allows us to create unique content from real customer questions. Then it was like, okay, can we automate the transcripts that are over 3 minutes to get added to a spreadsheet and then filtered by the most asked questions? Yes, we can. And this is what we had. We had a spreadsheet with all the questions and then, and then the most asked questions. So now we knew which transcripts we should use to create content. And now we built an agent. The agent pulls the call transcripts. It pulls in client meeting notes. It pulls in social media content, reviews, and it builds a knowledge base. And then it analyzes the call for the top questions and it creates content using the knowledge base that we built. And it checks on the website to say, has this ever been talked about before? And if not, it adds it to the website. So this is exactly how we built Turner to help clients rank in AI search. And then the last one is Vic, Victor. And this is not one that we built. This is one that we actually buy. It's an agent. It's more of a generalist agent. And it goes across like attract, convert, deliver, and collect. And the reason that we like Victor is because it can connect to Google Docs, Sheets, QuickBooks, Slack, Asana, Fathom, and other systems that we use inside of our business. And it has access to all of these things and it allows us to talk to it via Slack, and then it can do things on our behalf. It's basically like an assistant that runs through all of your systems. So it's more of a generalist AI agent that can help you with things. And that's kind of what it comes down to inside of your businesses, these specialists versus generalists. And when we think about specialists, it's like in my business, I have a very unique process for doing this thing. And that could be dispatch, that could be sales, that could be collections, that could be marketing. And that system is very unique to my business. And so this is why we built Trace and Turner, because these are very unique proprietary systems that we've built inside of our business that we have to deploy in a unique way. Now, generalists on the other side, like Vic, You call them Vic for short. This is more of a generalist tool. Like think about like Claude or something, agents. And you basically just build a generalist that has access to your emails, to your drives, to other tools that are relatively easy to connect. And now you can use it as an agent to help you as an assistant. This is the difference between a specialist versus a generalist. Now, usually when I talk about this, people are asking this question is, you're listening to this, you probably own a business. Is, oh wow, this sounds like a lot of work. And what I've found by working with home service business owners is that they see a video and then it leads to another one, and then they start vibe coding things out, like what I just said on Claude. And then they started to use Zapier, these connections, and then they have agents. And now someone who's running a $5 million business has 15 half-built AI projects and they're tweaking with this one, they're tweaking with that one and troubleshooting a broken connection at 10:30. This is a real example. And we didn't hand-build Trace and Turner ourselves. Like, I didn't build this myself. I had the manual process that we built, then we automated it with some Vibe coding and Zapier integrations, but then we brought in developers and AI people to help build them to the final product. And this is what I think about, like, for me is the highest value use of the CEO's time. And so the way that I think about this and the way that I'm making this video is because One, you gotta understand what's possible, and that's why I try to lay out like the AC/DC framework and how these technologies and agents can be used inside of your business. So you have to know the possibilities. More importantly, you have to identify the bottleneck that's costing you. The agents that I just explained to you, those are real bottlenecks in our business, right? They were manual processes, they were automated processes, but they were slowing down our team and our efficiency to do the best work for our clients. So we had to identify where's the bottleneck here? And then we had to figure out how can we deploy the right technology or partner to do this thing. And this is fundamentally a business owner's main job is to understand what's possible, identify the bottleneck, and then fix the bottleneck so that way the business can continue to grow. And this is kind of a bigger question of buy versus build. So what I would recommend, like with Victor, we buy Victor. It's a tool that's out there that you could go buy today. And what it needs is deep infrastructure integrations, accuracy, or specialized knowledge. So like Turner, for example, Turner has all of our proprietary processes and systems, and it has all the perfect integrations for home service businesses. So that way, if someone was going to go try to build this on their own, they're going to be spending a bunch of time and money to try to build this thing out when they could just buy it and plug it into their business. Same exact thing when we think about Trace. It's like, okay, return on ad spend tracking. There's a lot of proprietary things that have to happen in order for that to all work. And that's why we build it into our workflows. Now, when we think about building, this is the simple agent that I mentioned, which is screenshot your dispatch board, talk to it about your SOP, and then every day just have it run the same system. That saved you 2 hours, right? Same thing with the AR agent. You can build simple, unique internal processes that cost of failure is low. And you don't have to build all these custom API integrations. You can literally just train an agent on an SOP and have it deployed. And this is the fundamental change between buy versus build. The last thing that I'll say on this is that you should give your team the ability to create their own agents. So you could even have them watch this video or other video, other videos about how you actually use agents or build these processes inside of their workflow. It's not as big, as scary as it is, right? They can go through that exact framework that I just gave and build a simple agent on their own to help them be more efficient inside of their position. Agents do the work and humans own the judgment. This is a fundamental part that I think a lot of people miss online is that they just say, oh, just build an agent. Claude will be my marketing team and it'll just get rid of everyone. That's not how I think of agents. I think of them as tools. And today it's like your human on your team has QuickBooks, your CRM, your Google Drive with SOPs, right? They have all these tools. Maybe they even have ChatGPT or Claude and they have all these tools around. And the future is there's humans, That then have an agent or an SOP agent process, and they have access to all these different tools, and the agent can do things, and then the human can look over it and say, okay, this is right or it's not. Your people are still there. The leverage that they have is now different because they can use their top-tier skills. It literally makes your best A-players way, way better and more efficient because they can focus on what they do best, and then they can have agents run the admin tasks that are slowing them down. So here's the agent audit that I would recommend for your business. One, what happens in your business over and over again? You probably have things and ideas from what I just said. And like, what are your people doing manually over and over again that could be standard, could be put into an agent or a process that could be agentic and use some type of an agent on its behalf? Where do you have to interpret information? So where in your business is information interpretation happening? So for example, this is what I used to do. Like I would go and I would listen to the CSR calls and then I would go to the CSRs on our one-to-ones and say, hey, here's what I noticed over here. Here's what I noticed over there. Now you can literally just take all the transcripts and plug it into an agent that's built on your SOPs and it's going to give you a scorecard for every single team member. And then on the one-to-ones, you literally have, hey, this objection came up 20 times and you missed it 10 times. And so here's, let's role play this. It gets you to the information much, much faster. What happens after that decision? So essentially when something happens in your business, that trigger happens, what happens after the decision is made? Is there something that an agent could support? What's this problem costing you? When you think about your bottlenecks in your business, where is processes slowing down and where could you become more efficient? What is that costing you? Is it costing you $500 an hour? This person makes $25 an hour and they're spending 20 hours a week on it. Well, does that make sense to pay money for that? In marketing, it's pretty simple. Like you hire a marketing company to do SEO. This is how I think about it. You hire a marketing company to do SEO and let's say you pay $1,500. How are they going to market your business? Is there a better way to do that that's way less expensive? I'm sure there is. Should you buy it or build it? This is another question that you should be asking is like, is this something that I should try to build and vibe code this out, or should it something that I should go and try to find an agent that's already built that can do it? I can't hammer this point home enough. A great employee with great agents will outperform a great employee without them. It's basically a tool. And if you think about, you have an A player and you give them all the best tools, they're going to be better than an A player that has all the old tools that's broken and trying to use a knife that's not sharp. Your job isn't to become the CTO. It's to know enough to put these tools to work. So here's the checklist for you to finish this out. Before I go through this, if you own a home service business and you're interested in deploying agents and Uh, working with a marketing company that actually cares about using AI and technology to help you grow. That's what we do. I'll put a link down below. I'd love the opportunity to talk with you and see if we might be able to help you grow your business. So agent checklist, know the difference between automations follows instructions, agents interpret them. Number 2, map your business across attract, convert, deliver, and collect. And I would say identify 1 or 2, your admin bottlenecks inside of these things. I would probably say in delivery and collect. There's a lot of admin in there. There's a lot of CSRs. Like, one that just comes out to mind now is like our CSRs used to print out, right? We, we, we, our technicians had apps, but our CSRs still used to print out their job schedules the day before to give it to them. Like, what if there was an agent that could send them an email? And then you get what I'm saying. There's a lot of things inside of this that you could automate and use agents to make decisions for. Look at Turner, Trace Turner, and Victor as real working examples. These are real examples of real things that we're deploying inside of businesses to be way more efficient. Pick a manual process and move it. Manual, think about the manual thing. How could I automate it? And then how could I build an agent with an intelligence base? Don't try to build everything yourself. What I can tell you is maybe you build one thing just to, you know, figure it out, but work with companies out there are going to have agents that are already built that you can plug into your business and do things for you so that way you don't have to troubleshoot, build, you know, mess with the API connections, try to figure all this stuff out yourself. And then decide buy versus build based on the problem, not the hype. And finally, keep humans owning the judgment calls. This is not about agents replacing humans. This is about agents making humans much more productive and using them to make decisions much, much faster. All right, guys. Well, I make content about AI search, growing home service businesses, creating content, Um, and I appreciate you hanging out with me. If you like content like this, be sure to hit like and subscribe. Drop a comment if you have questions or thoughts, and I'll see you in the next video.
