AI employees aren’t the future; they’re already doing real work inside Alarm Masters.
In this episode of Entry & Exit, Stephen Olmon and Collin Trimble break down the digital employees they’ve built to handle estimating, prospecting, sales opportunities, operations, collections, and customer service. They explain how these AI agents work alongside their team, communicate with each other, and automate tasks that used to require multiple handoffs.
They also share where AI still needs human oversight and why voice-based agents could be the next major step for service businesses.
In this episode:
→ Building AI employees that do real work
→ Automating estimates and prospecting with Eddie and Stu
→ Using Kevin to uncover missed customer demand
→ How AI agents communicate and complete workflows together
→ Automating operations, case management, and collections
→ Where humans still need to stay in the loop
→ The future of voice-based AI employees
Connect:
Stephen Olmon — https://x.com/stephenolmon
Collin Trimble — https://x.com/TXAlarmGuy
Owned and Operated
New Episodes Every Wednesday!
Subscribe For More
Got into the AI world and I started seeing people say, Oh yeah, I've got a whole team of digital voice. Like, yeah, what do you have like five Claude threads named sales, marketing? Like what what is what are we talking about here? That's funny. That's funny because it's true. Whoa, whoa, whoa. Stop. Did you know that you can like and subscribe this podcast, Steven? I did not. Okay, well, I'm glad we're talking about it.
But I actually saw the future 10 years ago. At times we are going to want to personify and put skin on digital employees and throw them on a screen, but you can further integrate them into your business. It'll be interesting to see how far we take that. Welcome
to Entry and Exit. My name is Steven Ullman, and I have my co-host and business partner with me, Colin Trimble. And we run this podcast to help people out, help people understand how to grow and scale their business. And one of the ways that we have done that is by building digital employees, like AI-based team members. And we're going to talk about how we engage with them and how we have our team members engage with them in our business.
Whoa, whoa, whoa. Stop. Did you know that you can like and subscribe this podcast, Steven? Like you could do it right now on every single platform. You can click the subscribe button and share it with all your friends and family members. And like it super helps us. Did you know that? I did not. Okay. Well, I'm glad we're talking about it. And for all those listeners that don't know, it would mean the world to us if you would just click that button or even engage with us. Just like leave us a con uh comment. I don't know why that was.
I wanted to say comet, but I really wanted to send hey comment the gif of a comment. Is it gif or gif? I don't know. The world may never be. Yeah, I don't know. Well, we'll know. Find out in heaven. Jumping back in. Jumping back in.
Yeah, yeah. Here we
go. Um, so digital employees. This is a fun concept. I love talking about this because when I first got into the AI world and I started seeing people say, Oh, yeah, I've got uh a whole team of digital employees. It's like, yeah, okay. What do you have like five Claude threads named sales, marketing? Like, what is what are we talking about here? That's funny. That's funny because it's true. Yeah, it is. Well, this whole like, I'll give you 10 prompts, like, I got. I'm like, please stop. Um there's a lot of hyperbole out there. And if you are stressed about why the fact that I don't have digital employees, you should know that that is edge case. Like what we're doing is very far along, and and there's a lot of value that can be had way on the other side of the spectrum just using Claude and connecting all your systems. That would be the first thing I would say is like connect all your systems into a single uh AI uh tool, and you're gonna start seeing the most amount of impact there. Uh, but we do have digital employees. So let's go through a list, and then we can kind of talk about kind of how they're communicating on what surface they're communicating on and what that looks like. So the first one we built was Eddie, and Eddie is the estimation agent. Um Eddie's whole job is to help us do proposals. Uh, we came back to Eddie later, trained him on NFPA 72 fire alarm planning and design. So now Eddie can do full fire alarm um estimation and takeoff and all that for us. Um just put in the drawings, answer some questions, it'll give you the assumptions, but um it's pretty slick. So that's Eddie. Um from there, I think we did Stu, which is our SDR agent that finds new logo, leads and prospects, enriches them, gives us a point of view, puts it on a map, and then also does route planning for our team. So it'll literally you say, Hey, I want to visit 10 prospects in Sugarland, and it'll find, it'll use a map to geolocate those locations and optimize for drive time, and then send a meeting invite to the reps.
In between Eddie and Stu, we actually spent time with Casey, but Casey is kind of parked, you know.
So I and I need to pick it back up because some technology's improved in a big way.
Yeah. Um that was probably 90 days ago when we last touched Casey metaphorically. Uh, and that was uh it was uh metaphorically.
I'm glad you I'm glad you added that display.
To clarify, you know, it's just because we're talking about our relationship with our digital employees. Yeah, that's right. And so um that was kind of like a um call handling, you know, inbound call handling agent. Uh connected it to Salesforce, gave it access to data, tested it quite a bit, pretty far along. Just we have to give Casey some more love. And then we went to Stu.
Yeah, then we went to Stu. Uh Stu is out live production, doing great. Stu is also our only employee today. We'll get back to this more, is that's running on local inference, meaning we've got a Mac Studio uh that because it's the most token heavy of our agents, and so it's running all that uh itself. So that's super cool. Very excited
about that. Um from there, I think we went to Kevin, and Kevin was sort of out of this need for like, hey, it's really about our current set of customers. Help us find high propensity to buy coffee drops. This is where it started. Help us find white space, meaning, like, if I've got a customer that has uh video and berg, then they would be a good fit for video verification, for example. Um, helping them like clean up their pipeline, our reps would be like, hey, here's some things you need to clean up in your pipeline, you're missing, you know, amounts or whatever. So um that that was Kevin. Then he morphed into what I think is his his uh like Super Saiyan form, which um great reference. Yeah, and it really spawned out of a situation where a customer called in to our service department and referenced that they needed new, they needed help with their keypad, but they also had an issue with their cameras. And when the technician went out there, they found there was other people out there installing cameras. And we're like, what the heck? And they're like, Oh, we didn't know you did cameras. And so it was basically a way to listen to all of our recorded phone calls, text messages, and emails that are coming in through the general channels, not individual employees, uh private emails or anything like that, but just general community customer comms and trying to find literally just demand that was either explicitly stated, like somebody said, Hey, I need a quote for this, and somehow it got dropped before got to sales, or implicitly stated, like they said, hey, do y'all do cameras or whatever? Like, and so Kevin listens. He's at what we call our 360 agent, he's listening to all of our communication and he's surfacing leads to our uh to our team to say, Hey, you should you should go and call on that. So that's Kevin sends us a morning debrief, a digest. Another super cool thing about him. We actually tied him into uh there's a uh Google Maps is owned by Google, obviously, and so you're restricted for what you can do with Google Maps, yeah. But there's a free open source version of it through the US Agricultural Department, I think.
I think that's right.
That yeah, and so it's I think basically just as good from what I can tell and just as up to date. And basically our team can say, hey, f help me find the locations that um are large sprawling campuses for our existing customers, and it will go search for large parcels, and it will do three pictures a really far out one, a closer in one, or a really close in, and it will look at the pictures and say, Oh, this has got five buildings across what looks like you know 600,000 square feet, and then it'll actually go to uh our appraisal district and it'll see if it can pull the actual square footage to say, Hey, you should go call on these people.
Yeah,
you've touched a couple times on um cadences or interactions. So I think also as we're going, let's kind of define how we're interacting. So like and and how our team is like some of them are ad hoc. Like when you need help from Eddie, you go, you know, Slack Eddie, submit something to Eddie, just like you would someone else. You'd say, Hey, I need your help with this, essentially. Um, then there are automated or um scheduled jobs, essentially, right? And so that are, you know, we're getting something every morning from Kevin as an example, or or our reps are. Um and then there are some other things that we've built that we don't necessarily interact with as much, but are still happening in the background. Um, so there's there's kind of some different ways that we're interacting in a different cadences or um different differing levels of regular engagement.
Yeah, so we have some agents, uh both Stu and Kevin. Stu is almost entirely trigger-based. So Stu is searching the internet, looking for changes in public data, flagging that and producing a morning digest to our team. Yeah, Kevin is both proactive, listening and scanning calls, finding demand and surfacing that in a morning digest, and also uh reactive, meaning he's a chat agent, and there's a chat service where he could say, Hey, uh rep. The rep says, Hey, Kevin, find me leads in this area or whatever, and he'll go do that. And so um they both kind of interact uh with the individual, but they also can interact with each other. So, like Stu can send a message to uh Kevin uh under certain circumstances as it relates to like, or actually it's really more the other way. Kevin will send a message to Stu for route planning. So it'll be like, hey, can you put a route together for this? And Sue's like, yeah, bam, and it'll send the meeting invites. And then Stu sometimes can reference to Kevin if they want to look at the parcels of land, but we're not doing that a ton, but it has happened where they'll be like, go look this up. And Sue is like, hey, at Kevin in Slack, like, can you do this? One really cool example of one that I did not build this agent. It's um, I don't know why he named it. It's uh Albert built OLI. Oh, OLI, the operations agent. And um, Ollie is trigger-based on when we close when a deal, it pushes a bunch of workflows out, creates a work order, sends the deposit invoice, sends the opening of the project email to the internal team and Slack messages, creates a folder. Um, but then it also will uh do it, it's really cool because it also serves the service team. So if a technician puts notes and says, hey, we need a return trip meeting because we need a quote, it'll actually take the notes, copy and paste it into Eddie's channel. Eddie will do all of the estimation and push the parts back to Ollie, and Ollie will then put them into Salesforce and generate the service quote. So they are like having some conversations back and forth. Obviously, we can see all of that in Slack that's happening real time. And it's really just saved us a bunch of time. I mean, think about how often this happens in people's businesses where it happened in ours for years. Technician says, I need a quote for this customer, and their brain turns off, like, okay, I've told the office there needs to be a quote. And like I put my notes in there. But then the office folks are like, oh yeah, I gotta do that, and I gotta get it to the person that's gonna do the quote. And it's like only a couple people on the scheduling team know how to do that, and then they're like coordinating, and then you gotta get a thing, and it's like just doesn't happen, or it happened takes too long. Now it's just like bam, it's done.
Like within an hour.
So where I think that this I actually saw, like, and and we haven't really talked about this in depth, and you know, I'm I'm gonna go a little more future forward here, but I actually saw the future 10 years ago. Um, I used to work for a software company, uh, and we had a kind of thing like a Teledoc sort of thing. If you imagine like a iPad rolling around the office on like a stick, and it had like a single wheel, almost like a unicycle. And so our CTO is like super nerdy and awesome. And so we had a couple employees in a different city, and so they would roll around our office and jump on the screen, and you could like have a face-to-face with them, and this is like 2016, and that was really cool. Well, I think where this is gonna go is at times we are going to want to personify and put skin on digital employees and throw them on a screen and have voice to voice. Yeah, and today that's already possible, like with Hagen or Levin Labs, and basically like working with Eddie through a proposal, voice to voice, uh, is actually not far away. We couldn't force it to happen today. It'll probably be six to twelve months from now before it's like really feasible. But you can further integrate them into your business if you want, and so it'll be interesting to see how far we take that.
Yeah, yeah, super true about that. Two more agents I want to cover. Number one is fully ramped up and out, and the other one is about to be ramped up and pushed out. It's in it's in dev kind of beta right now. But the next one is Chase the case agent. Chase is on the case, Chase is on the case. Shout out Paw Patrol.
Paw Patrol.
Yeah, exactly. Uh, and so Chase the case agent is basically doing what Kevin is doing for sales, but for for case handling. So he's listening and reviewing all the case stuff, and I'll say, Hey, this customer sent you an email uh case handler, and you haven't responded to the customer in 24 hours. You should do something about that. And by the way, what we think you should say to the customer is this, and it drafts the full email for you, and then you can edit it right there in Slack, and it literally copies it into the Slack dedicated Slack thread for every case. So all the documentation is there, and it's self-learning. We don't have agents today, trying to make sure to keep myself honest here, that are autonomously communicating with customers in a dynamic way. Like it's not saying email send, customer replies, email replies back without us uh having control over them.
Kind of right now.
Yeah, it's like, hey, it'll send one email, but then if there's a reply back, then it'll suggest some stuff, and then the person has to approve. There's like some guardrails around that because we want to make sure we deliver a good experience. Um, so that's Chase the Case Agent, super cool. Um, Cali the Collections Agent. This is a working title, by the way. Cali the collections agent. I don't know how we feel about that. We can workshop it. Hey, if you by the way, this is a good stopping point. If you like the name Cali, the collections agent, put a comment in. If you think it's if it's better, if you think there's a better one, we want to know. So help us name Cali.
The first four letters of the word collections are the first four letters of your name. I know.
I thought about that.
All in the collections. I really don't want to do that though. That would confuse people. I don't either. So I'll pull out it.
The problem is we don't have any females, and I was really feeling like it's all masculine names. And then I wanted anyone, I don't have a daughter.
What about what about collette?
Yeah, that kind of sounds like a little bit of a stripper name.
Oh goodness, from what decade?
I mean, I didn't mean it that way. I just mean like it has that connotation.
It's not like Jasmine. I'm sorry, if we've if we've got Noah Fins, by the way. Yeah, we've got a Colette. If we have a Colette listening, Colin is so sorry.
I'm so sorry. I didn't mean it. Leave me a comment and tell me how pissed you are. Yeah, it's a problem. So Callie, the collections agent, will stick with Callie. So so Callie, the collections agent, um, is basically just doing all the manual tasks, sending emails, creating tasks for the reps to make phone calls on. It's actually looked uh looped into a direct mail provider so I can send out letters. And it all that's wired up, I haven't hit go yet because I'm just like, I need to test the like letter thing. And anyway, so all that to be said, that's sort of our stack as it pertains today. There is limited workflows where the agents are talking to each other, but there are some. Um, and it's gonna increase in uh the amount that they are kind of interacting with each other. But it, you know, we're kind of moving at the pace the technology is moving. Like some of it's a stretch, like Casey, the call agent, like nine months ago just wasn't at the quality level we wanted it. It was fully stood up, it worked, just yeah, wasn't at the quality we wanted. But now there's like a whole new version of these agents. Uh apparently Grock 4.6, the voice mode on it is like next level. So we'll probably experiment with that when we've got some time. Oh, last one. This is another one that was a voice agent that uh we didn't roll out because it has to be dialed, is Ty, the Technician Tier One helper. Oh basically. That's right. Yeah, he has all the manuals and product sheets.
When that looking back, like a year or two from now, I actually think that that outside of Eddie might be the single most impactful AI we have.
Where this came out of was my rep saying, Man, I tried to call an XYZ manufacturer and I was on hold just to figure out how do I do this thing. And then they sent me an article online that's like here are the steps. I was like, oh shoot. So, like, okay, then let me use an agent that'll scan all that. Uh and like we could roll it out today as a chat-based agent, but I really their text would only really use it if it's voice-based. So, um, so yeah, we've got a bunch of different agents. We've got several that are in dev right now that we're still working on and messing around with. Um so yeah, a lot, a lot to come. Anything else you want to share, Steven?
No, if uh if you found this to be helpful, we would find it helpful if you would like and comment and subscribe and forward, and you know, maybe uh print out a picture of us, put it on your desk, you know, just give it a little inspiration. Small ask. We appreciate it. Thanks everybody for listening. Have a great week.
.webp)
