Your crew is finishing a furnace swap at 5:30 PM. A homeowner three suburbs over calls about a water heater leaking into the laundry. The phone rings four times and rolls to voicemail. She hangs up. She calls the next HVAC company on Google. They answer. You just lost a $2,400 job because nobody picked up.

That scenario plays out in trades businesses dozens of times a week. The owner is dispatching, the admin left at five, and the techs are on the tools. Calls go dark. Estimates sit in inboxes. Happy customers never get asked for a review. The work that doesn’t require a licensed tradesperson still costs real money when it doesn’t happen.

AI agents are changing that math. Not the chatbot on your website that nobody clicks. Not a dashboard that “automates workflows” but still needs someone to check it twice a day. I’m talking about agents that answer the phone, route the job, follow up on the estimate, and ask for the review without a single ping to your phone.

This isn’t speculative. Trades businesses running Omni voice and ops agents are capturing calls they used to miss, converting estimates that used to go stale, and collecting reviews without chasing customers. The technology is quiet, the ROI is loud, and if you’re still running dispatch and follow-up manually, you’re leaving five figures on the table every quarter.

The Work That Costs You When It Doesn’t Get Done

Walk through a typical week in a plumbing or HVAC business doing $3M a year. You’ve got six trucks, maybe eight techs, and one person in the office who handles dispatch, invoicing, and customer calls. When she’s on the phone booking a job, three other calls roll to voicemail. When she’s out sick, the owner is glued to his phone between job sites. When a big install runs late and everyone’s scrambling, follow-up on last week’s estimates doesn’t happen.

None of that is a crisis. It’s just how it works. But add it up over a quarter and you’re looking at 40 to 60 missed inbound calls, 30 estimates that never got a second touch, and 80 completed jobs where nobody asked for a review or mentioned the maintenance plan. Each missed call is worth $500 to $3,000 depending on the job type. Each stale estimate had a 15 to 25 percent chance of closing if someone had followed up on day two. Each review you didn’t collect is a Google ranking point you gave to the guy down the street.

The math is uncomfortable. A business at this scale is typically leaking $50,000 to $200,000 a year on work that should have happened but didn’t because a human wasn’t available at the exact right moment. That’s not a people problem. It’s a capacity problem. And capacity problems don’t get solved by hiring another admin or telling the owner to work longer hours.

They get solved by handing repeatable, high-volume work to an agent that runs 24/7 and costs a fraction of a full-time hire.

What an AI Agent Actually Does in a Trades Business

An AI agent isn’t a tool you log into. It’s a worker that handles a specific job end-to-end. You configure it once, connect it to your dispatch system or CRM, and it runs. When a task comes in, the agent executes. When it’s done, it logs the outcome and moves to the next one.

Let’s take the 24/7 Dispatch Voice Agent we build in Omni Voice. A homeowner calls your main line at 7 PM on a Tuesday. Your office is closed. The agent picks up on the second ring. It greets the caller by name if the number is in your system. It asks what’s going on. The homeowner says the AC stopped blowing cold air and it’s 90 degrees inside.

The agent qualifies the job. Is this an emergency or can it wait until tomorrow? The homeowner says tonight if possible. The agent checks your dispatch board, sees you have an on-call tech available, and offers a two-hour arrival window. The homeowner agrees. The agent books the job directly into your scheduling tool, assigns the tech, sends a confirmation text to the customer with the tech’s name and ETA, and logs the call. The tech gets a ping with the address, the issue, and the time window.

Total human involvement: zero. The call that would have gone to voicemail and never come back is now a booked job with a customer who feels taken care of. That’s $800 to $1,500 in revenue you didn’t have to chase.

Now take the Estimate Follow-Up Agent in Omni Ops. You send an estimate for a $4,200 bathroom remodel on Monday morning. The agent waits 48 hours. On Wednesday, it sends a text: “Hi Sarah, just checking if you had any questions about the estimate we sent for the bathroom. Happy to walk through the timeline or options. Reply here or call us anytime.” If Sarah doesn’t respond, the agent waits three days and tries again with a slightly different message. If she still doesn’t respond, it waits another week and sends a final nudge with a soft close: “We’re booking jobs for next month. If you’d like to lock in your spot, let me know by Friday.”

That sequence converts 15 to 25 percent of estimates that would have died in the inbox. The agent runs it for every estimate you send. It doesn’t forget. It doesn’t get busy. It doesn’t assume the customer wasn’t interested. It just follows up, logs the response, and hands you the ones that are ready to close.

The Review and Reactivation Agent works the same way. Every time you close a job, the agent waits 24 hours and sends a review request. “Hi Tom, glad we could get your water heater sorted yesterday. If you’ve got 60 seconds, a quick review helps us a ton.” If Tom leaves a review, the agent thanks him. If he doesn’t, it tries once more a week later. For customers who haven’t called in six months, the agent sends a reactivation message tuned to the service interval: “Hi Lisa, it’s been about six months since we serviced your furnace. We’re booking fall tune-ups now. Want to grab a spot before the cold hits?”

These aren’t sophisticated AI moves. They’re just consistent execution of the work a great admin would do if she had infinite time and never took a day off. The difference is the agent actually does it, every time, for every customer.

Why This Matters More in Trades Than Most Industries

Trades businesses live and die on speed. When a homeowner has an emergency, they call three companies. The first one to answer and offer a time window usually gets the job. When an estimate goes out, the window to close it is about 72 hours. After that, the urgency fades, the customer gets another quote, or they just decide to wait.

You can’t win that game with a five-person office team working 9 to 5. You need someone picking up at 6 AM when the restaurant owner realizes his walk-in cooler died overnight. You need someone following up on estimates at 8 PM when the homeowner finally sits down to review the numbers. You need someone asking for reviews on Saturday morning when the job finished Friday afternoon and the customer is still happy.

A human can’t do that without burning out or doubling your overhead. An agent can. And because trades work is high-ticket and repeat-driven, the ROI shows up fast. One extra job a week from after-hours calls is $40,000 to $80,000 a year. Three extra closes a month from estimate follow-up is another $60,000 to $120,000. Ten more reviews over six months moves your Google ranking enough to shift your inbound mix.

We see this in the numbers when we run the AI audit for trades businesses. Most owners assume they’re missing 10 or 15 calls a month. When we pull the phone logs and map them against booked jobs, it’s closer to 40 or 50. They assume their close rate on estimates is 50 percent. When we track follow-up, it’s closer to 30 percent because half the estimates never get a second touch. The gap between what you think is happening and what’s actually happening is where the leakage lives.

The Three Agents Every Trades Business Should Evaluate First

If you’re running a plumbing, HVAC, electrical, or roofing business doing $1M or more, you don’t need a dozen agents. You need three, and you need them running well.

The first is the 24/7 Dispatch Voice Agent. This is the one that answers every call, qualifies the job, books the slot, and confirms with the customer. It plugs into whatever dispatch tool you’re already using—ServiceTitan, Housecall Pro, Jobber, or even a Google Calendar if that’s where you live. The agent doesn’t replace your dispatcher. It handles the inbound volume your dispatcher can’t cover, especially after hours, during lunch, and when the office is slammed.

The second is the Estimate Follow-Up Agent. This one tracks every estimate you send and runs a follow-up sequence tuned to your close cycle. It doesn’t spam. It doesn’t sound like a bot. It just nudges at the right intervals with the right message and hands you the warm leads when they respond. If you’re sending 20 estimates a month and closing 10, this agent will get you to 13 or 14 without changing anything else about your sales process.

The third is the Review and Reactivation Agent. This one asks every customer for a review within 24 hours of job completion and reactivates past customers at the right service interval. Reviews drive your Google ranking, which drives your inbound call volume, which drives your revenue. Reactivation turns one-time customers into repeat customers, which is the cheapest revenue you’ll ever generate. Most trades businesses do neither consistently because it’s boring work that falls off the list when things get busy. The agent just does it.

Those three agents cover the highest-value, highest-volume work that doesn’t require a human. They don’t replace your team. They free your team to do the work that actually needs a person—closing the big jobs, handling the complex customer situations, and running the business.

What It Looks Like to Actually Implement This

You don’t rip out your systems and rebuild everything. You start with one agent, connect it to your existing tools, and let it run for 30 days. We usually start with the voice agent because the ROI is immediate and the risk is zero. If the agent can’t handle a call, it transfers to a human. If it books a job wrong, you catch it in dispatch and fix it. But after two weeks, you’ll see 20 to 30 calls that got answered and booked without anyone on your team touching them.

Once that’s running clean, you layer in the follow-up agent. Pull your last 90 days of estimates, load them into the system, and let the agent start working the list. You’ll get responses within 48 hours. Some will be “not interested,” which is fine, you stop wasting time on them. Some will be “we have questions,” which your sales guy closes. Some will be “we’re ready to book,” which goes straight to the calendar.

The review agent comes last because it’s the lowest urgency, but it compounds. Every review you collect this month makes next month’s inbound calls easier to convert. Every reactivation message you send this quarter turns into revenue next quarter. It’s not flashy. It’s just math.

The whole process from first call to all three agents running takes 60 to 90 days if you’re moving at a normal pace. If you want to move faster, you can do it in 30. The constraint isn’t the technology. It’s how fast you can make decisions and give the agents access to your systems.

The Overhead Question Everyone Asks

The question we hear most often is: what does this cost compared to hiring someone? A full-time admin in a trades business runs $40,000 to $55,000 a year when you include benefits and taxes. That person works 40 hours a week, takes vacations, gets sick, and can handle one task at a time.

An AI agent runs 24/7, doesn’t take breaks, and costs a fraction of that. The three agents I just described—voice, follow-up, and review—typically run $1,200 to $2,400 a month depending on call volume and complexity. That’s $14,400 to $28,800 a year for coverage that would require two or three full-time people to replicate.

But the real comparison isn’t cost. It’s capacity. Your admin can’t answer the phone at 10 PM. She can’t follow up on 40 estimates in parallel. She can’t send 60 reactivation messages in an afternoon without dropping everything else. The agent can. So you’re not replacing a person. You’re adding capacity that didn’t exist before, which means you’re capturing revenue you couldn’t capture before.

The break-even on these agents is usually four to eight weeks. One extra job a week pays for the whole system. Everything after that is pure margin.

What Happens If You Wait

The trades businesses that adopt this stuff early will have a compounding advantage over the next 18 months. They’ll answer more calls, close more estimates, collect more reviews, and reactivate more customers. That turns into more inbound volume, which turns into more revenue, which turns into more capacity to invest in the next layer of automation.

The businesses that wait will keep losing calls, leaving estimates on the table, and wondering why their Google ranking isn’t moving. They’ll hire another admin to cover the gap, which helps for six months until call volume grows again and they’re back in the same spot.

This isn’t a technology risk. The agents work. The ROI is clear. The implementation is straightforward. The risk is waiting until your competitors have been running this for a year and you’re trying to catch up while they’re compounding.

If you’re not sure where to start, grab the After-Hours Call Recovery Plan for Trades. It’s a one-page worksheet that walks you through how many calls you’re actually missing, what those calls are worth, and which agent to build first. It takes 15 minutes to fill out and it’ll give you the numbers you need to make a decision.

Or skip the worksheet and just book your Omni Audit. We’ll pull the data, map the leakage, and hand you the plan. Sixty minutes, three outputs, no fluff. You’ll know exactly what you’re losing and exactly how to fix it.

The Boring Truth About AI in Trades

AI agents aren’t going to run your business. They’re not going to replace your techs or your sales team. They’re not going to make strategic decisions or handle the weird edge cases that require judgment.

What they will do is handle the repetitive, high-volume work that costs you money when it doesn’t get done. Answering calls. Following up on estimates. Asking for reviews. Reactivating past customers. The work that a great admin would do if she had infinite time and never got tired.

That’s the opportunity. Not robots running the company. Just consistent execution of the boring stuff that drives revenue. The businesses that figure this out in 2026 will be printing money by 2027. The ones that don’t will still be wondering why they can’t break through the next revenue ceiling.

The practical next step is the free Working With Claude field guide. Thirty-two pages covering the ecosystem, Claude Code, and how to govern a rollout properly. Get your copy.