The Plumbing Company That Gave Every Tech Their Lunch Break Back

The radio crackles. "Carlos, can you take the water heater replacement in Fort Worth?"

Carlos glances at his GPS. He just finished a job in Plano - 45 minutes north. Fort Worth is an hour south, straight through afternoon traffic.

"Copy that," he says, because what else can he say?

He pulls onto 635, merging into the crawl. His phone buzzes. It's Marcus, another tech from the company. They're in a group chat.

Marcus: "Where you headed?"
Carlos: "Fort Worth. Water heater."
Marcus: "I'm literally 10 minutes from there. Just wrapped in Arlington."

Carlos doesn't respond. What would he say? The dispatchers are doing their best. They can't possibly track where 22 technicians are at every moment, cross-reference their certifications, check real-time traffic, and make optimal decisions in seconds.

So Carlos sits in traffic. Again. And his lunch - still in the cooler from this morning - stays there. Again.

That's field service inefficiency. Not dramatic. Not obvious. Just expensive. Every single day.

Redline Plumbing isn't some massive corporate operation. It's a family-owned residential plumbing company serving the Dallas-Fort Worth metroplex. Third generation. The kind of place where the owner knows every tech's kids’name.

Thirty-four employees total. Twenty-two technicians. Service area spanning 60+ miles from Plano to Fort Worth, Frisco to Duncanville.

The math in field services is brutal: efficiency directly equals profitability. A tech sitting in traffic isn't generating revenue. A truck burning fuel to zigzag across the metroplex is burning cash. Time wasted is money lost, and the losses compound.

Two dispatchers managed the schedule. Good people. Smart people. But they were playing an impossible game.

Here's what they were juggling:

They did their best. Their best wasn't enough - not because they weren't talented, but because the problem was mathematically impossible for humans to solve optimally at scale.

Let's talk about what "doing their best" actually looked like in practice.

The dispatchers used a system I've seen in hundreds of field service companies: they looked at who was available and assigned the next job. Proximity? Sometimes factored in. Skills match? When they remembered to check. Real-time traffic? Maybe they'd glance at Google Maps. Customer preferences? If there was time.

The result was what I call "availability-based dispatch." Not terrible. Just suboptimal in ways that add up fast.

The pattern looked like this:

A tech finishes a job in Plano. The dispatcher sees a new water heater installation request in Fort Worth. The tech is marked "available" in the system. Assignment made. Done.

Meanwhile, another tech who's been with the company for eight years (the one who can do water heaters in his sleep) just finished a simple faucet repair in Arlington, fifteen minutes from that Fort Worth job. But the dispatcher didn't see him as available yet because his ServiceTitan status hadn't updated.

So the less experienced tech gets a 90-minute round trip. The expert tech gets sent to a complex job he's less suited for. Both jobs take longer than they should.

Multiply that by 22 techs and 40-60 jobs per day.

Here's what it cost them monthly:

Over 90 minutes per tech per day in unnecessary drive time. That's windshield time, not wrench time. For a crew of 22, we're talking about 660 hours of lost labor monthly.

About 25% of their dispatches were suboptimal. Wrong tech for the job. Wrong route. Wrong sequence. Not catastrophically wrong, just... not right.

$4,200 in excess fuel costs. That's the literal gasoline being burned to drive technicians past each other on the highway.

Fifteen percent fewer jobs were completed per day than the schedule could theoretically support. The techs weren't working less hard - the routing was just inefficient enough to kill their capacity.

The owner, Mike, told me something I won't forget: "I kept thinking we needed more trucks. More techs. Turns out we just needed them in the right places."

This wasn't their first attempt at solving the problem.

They'd tried ServiceTitan's built-in dispatch board. Helpful for visualization, but it couldn't make optimization decisions. Just showed them where everyone was and let them drag jobs around manually.

They'd tried a standalone route optimization tool. Cost $300/month. Sounded perfect. Except it wanted them to plan the entire day's routes in the morning, then stick to that plan. Anyone who's run a field service business knows that's fantasy. Jobs run long. Emergency calls come in. Techs call in sick. The perfectly optimized 8 AM plan is worthless by 10 AM.

The routing software couldn't account for skill matching. It would route a journeyman to a simple faucet repair and send an apprentice to a complex water heater installation. Technically efficient driving. Operationally disastrous.

Mike's words: "The software was built by people who'd never actually dispatched a single truck."

So they went back to human dispatchers doing their best because imperfect humans who understood the business were better than perfect algorithms that didn't.

That's where things stood when we started talking.

Here's what we built: an AI agent that sits between incoming jobs and dispatch decisions, making optimal assignments based on location, skills, traffic, and job requirements.

I need to be honest about something. This wasn't plug-and-play. This wasn't "buy software and turn it on." This took real work.

Here's how it works, conceptually:

A new job enters ServiceTitan, either from a phone call or online booking. Or a tech marks their current job complete in the field. Those are the triggers.

The moment that happens, the agent wakes up.

It pulls the real-time location of every available tech from their mobile devices. It calculates the ETA from each tech's current position to the new job location, factoring in current traffic conditions via Google Maps API.

Then it gets smart. It looks at the job requirements: what skills are needed, what equipment, how complex the work is. It matches those requirements against each tech's certifications, experience level, and track record with similar jobs.

It checks customer history. Has this customer worked with anyone on the team before? Did they request someone specific? Are there notes about access issues or specific preferences?

It runs all of this through a decision matrix we spent two weeks calibrating with Mike and his senior techs. Not just "who's closest" but "who's closest with the right skills and availability and customer fit."

The output? A recommendation in ServiceTitan with a Slack notification to both the dispatcher and the selected tech.

Here's the critical part: the dispatcher can approve it with one click, or override with a reason. The agent doesn't eliminate human judgment. It augments it. The dispatcher went from making 60 educated guesses per day to reviewing 60 optimized recommendations and focusing on the weird edge cases.

The technical stack (for those who care):

The complexity nobody sees:

We had to map out every certification and skill combination. A "water heater installation" isn't just one skill. It's gas line work, electrical connections, permit knowledge, and load calculations. Some techs can handle tankless. Others shouldn't touch them.

We had to build rules for how to weight different factors. Is it better to send a tech who's 15 minutes farther but has done this specific job type 50 times? What if traffic delays the closer tech by 20 minutes? What if the customer has complained about rushed work in the past?

We tested it in parallel for two weeks before Mike trusted it enough to use it for real decisions. The dispatchers would see the agent's recommendation, make their own decision, and we'd track the outcomes. The agent was right 86% of the time. The humans were right 71% of the time.

That's when Mike flipped the switch.

If you want to understand more about how AI agents actually work in business operations, we wrote about the fundamentals here: Your Competitors Just Automated That Thing You Hate Doing. This dispatch agent is a perfect example of the "surgical fixer" approach—one narrow problem, solved really well.

Five weeks from our first conversation to the agent running live. Another two weeks of calibration and tweaking. Then the numbers started coming in.

Average drive time per tech per day: Dropped by 28 minutes. That's 28 minutes per tech they got back. Not 28 minutes total. Per technician, per day.

For the techs, that meant they could actually finish their routes without hitting overtime. It meant they could take a real lunch break instead of eating while driving. It meant they got home when they thought they would.

Suboptimal dispatch rate: Went from 25% down to 6%. That's the percentage of assignments where hindsight revealed there was a significantly better option.

The remaining 6%? Those were edge cases where customer relationships trumped pure efficiency, or emergency situations where speed mattered more than perfect optimization. The kind of decisions the dispatchers still made. Now they were making six exceptions per day instead of fifteen.

Fuel costs: Down $3,100 per month. That's just gasoline. Doesn't count reduced vehicle wear, maintenance, or the environmental impact nobody asked about, but Mike mentioned anyway.

Daily job completion: Up 18%. Same techs. Same trucks. Same service area. They just got to the right jobs faster, with the right skills, and fewer wasted miles between stops.

Time from customer call to tech dispatch: Dropped from 15 minutes to 3 minutes. Because the agent could evaluate options in seconds, not the time it took a dispatcher to check availability, call a tech, wait for callback, then assign.

The surprise metric nobody predicted: customer satisfaction scores went up. Not because the work quality changed (it was always good). But because wait times dropped, techs showed up with the right equipment and expertise, and jobs finished on schedule.

Mike told me something six weeks in: "Our senior tech (been with us eleven years) called me on a Tuesday. Said, 'I don't know what you did, but I actually had time for lunch today.' That's when I knew this was working."

I asked him what he meant.

"For three years, that guy ate lunch while driving between jobs. Every day. Never complained. That's just how it was. Now he parks. Eats. Breathes for twenty minutes. Then gets back to it."

That's not in the ROI calculation. But maybe it should be.

Let me be clear about something: this wasn't a technology project. It was a capacity unlock.

Mike didn't add a single tech. Didn't buy a new truck. Didn't expand the service area. The only thing that changed was how dispatch decisions got made—and that freed up capacity that was already there, just trapped behind inefficient routing.

Eighteen percent more jobs per day. That's real growth without a corresponding increase in overhead.

The dispatchers didn't lose their jobs. They got upgraded. Instead of playing calendar Tetris all day, they now handle customer relationships, manage complex situations, and focus on the 6% of cases that need human judgment. The agent handles the repetitive optimization. They handle the relationship work.

Here's what I want you to understand: the agent does one job. Dispatch optimization. That's it.

It doesn't manage payroll. It doesn't handle customer complaints. It doesn't schedule vacations or process invoices. It makes optimal dispatch recommendations based on real-time data and business rules.

One job. Done a hundred times per day. Done consistently. Done well.

That's the model. Narrow problems. Reliable solutions. Measurable results.

Maybe your field service business isn't plumbing. Maybe it's HVAC, electrical, cleaning, pest control, landscaping. Doesn't matter.

If your technicians are spending more time driving than working, that's a bottleneck that's costing you real money every single day.

If you're routing based on availability instead of optimization, you're leaving capacity on the table.

If your dispatchers are overwhelmed trying to make perfect decisions in impossible timeframes, you're losing efficiency and burning out good people.

Here's the question that matters: What would your business look like if every tech got an extra 28 minutes of billable time per day?

For Redline Plumbing, it meant 18% more revenue without adding headcount. For your business? Run your own numbers. [Number of techs] × 28 minutes × [average hourly rate] × 22 working days per month.

That's what's sitting in traffic right now.

We don't build general-purpose software. We build specific agents for specific bottlenecks in specific businesses. Sometimes that's dispatch optimization. Sometimes it's scheduling, proposal generation, or inventory management.

But it always starts the same way: identifying the one bottleneck that's strangling your capacity.

Want to find yours? Book a Bottleneck Audit - 30 minutes on a call where we map your operation and identify where capacity is getting lost. No cost. No pitch. Just a clear view of where you're hemorrhaging efficiency.

[Book a Bottleneck Audit →]

Want to see more stories like this? Download "Unstuck: 25 AI Agent Blueprints That Freed Real Businesses from Their Most Expensive Bottlenecks." Real companies. Real problems. Real results.

[Download Unstuck →]

One bottleneck. One agent. That's how companies are finding capacity they didn't know they had.

by WB
for the AdAI Ed. Team