If you run a service business, you already know the math. Every minute a lead waits, your close rate drops. Sixty minutes in, you've lost half of them. Twelve hours in, the lead is gone — they booked someone else.
The fix everyone tells you is "hire someone." But you've tried that. Good ones cost $60k+ a year. Bad ones lose more leads than they close. And neither one answers a text at 11pm.
So I built Sarah.
Who Sarah is
Sarah is the AI sales agent for both of our Merry Maids franchise locations — Bellingham and Burlington, Washington. She works inside Google Local Services Ads, where about half our leads come in.
She's not a chatbot. She doesn't open conversations with "Hi, I'm a virtual assistant." She doesn't say "I'm sorry to hear that." She doesn't reroute customers to a phone tree. She quotes prices, answers questions, books visits, and sends pay links.
She follows the same playbook the best human sales rep we ever had used to follow. Same warmth, same pacing, same questions. Except she replies in 90 seconds instead of 90 minutes, she works at 6am on Sunday, and she doesn't quit.
The before
Until last week, our Burlington location had a real problem: leads were sitting overnight.
A customer reached out one evening for a teacher-appreciation clean. She got our agent on the first message. Conversation went smoothly. We closed it the same night.
But the next morning, a different lead came in at 6:30am. Customer was packing for a move, wanted a quote before her morning meeting at 10. Our agent didn't see it until I personally pinged her at 9:11.
That lead almost died.
We dug into the system and found the bug — our agent was checking for new messages against the wrong timezone, so anything that landed before 10am Pacific looked like it had already been answered. We fixed it that morning.
What that looked like, specifically
That $1,043 number is the part most case studies wave at. Here's what it actually was:
- A 3-bedroom 2-bath rental move-out clean in Burlington, $551
- A standard recurring biweekly in Bellingham, $189 first clean
- A detailed deep clean in Burlington, $303
Three customers. Three conversations Sarah handled end-to-end. Three pay links she generated and sent inline, no human in the loop. All three completed payment within 24 hours of the booking.
I watched it happen from my phone while I was at dinner with my family.
What Sarah cannot do
Every case study you've ever read promises "AI that does everything." That's a lie. Here's what I tell every prospective client up front about what Sarah does not do.
- She does not handle commercial accounts over $2,000. Those route to Andrea, our office manager, the same way they did before.
- She does not handle complaints or refund requests. Those get flagged and Andrea calls the customer within an hour.
- She does not negotiate below tier pricing. If a customer pushes, she holds, then escalates.
- She does not work outside of our service area. Anything outside Whatcom and Skagit County gets refused with a referral.
- She does not call customers. Phone is a different problem; we're solving it later.
The point is not "AI replaces your team." The point is "AI replaces the part of your team that's stuck answering low-complexity messages at 11pm so the rest of your team can focus on the work that actually moves margin."
What changed at our office
The honest part. Three things.
One: Andrea stopped getting pinged after hours. Used to be five or six late-night customer texts a week routing to her phone. Now it's the genuine emergencies, maybe one a month.
Two: our LSA channel finally started compounding. When you respond in 90 seconds instead of 90 minutes, Google rewards you with more leads. Our LSA position has been climbing for three weeks straight.
Three: I sleep through the night. That's the whole reason I started GEGP. To buy and run service businesses without my phone being the bottleneck. Sarah is the first agent that proved it can actually work.
What this would look like in your business
I built Sarah for Merry Maids because Merry Maids is one of the businesses my holding company operates. We also run Cedar & Stone Remodeling and we're closing on Fast Patch Drywall Company. Same agent stack, different per-business configuration.
The pattern that translates:
- Find the single channel where your leads die fastest. For most service businesses, it's after-hours text or LSA chat.
- Build one agent there first. Not "AI for the whole business." One agent, one channel.
- Measure response time before and after. If you cut it from hours to seconds, the leads start compounding.
- Add the next agent only after the first one has been running clean for 30 days.
This is the playbook. It's not magic. It's specific work, built specifically for one business at a time.
Want one in your business?
I run a 30-day pilot for a small number of service-business owners per quarter. One agent, one channel, real numbers.
Book a 30-min call →How the pilot works:
- Step 1. We diagnose where your business is losing leads today. A free 30-minute call, by phone or Zoom. You decide if you want to keep going.
- Step 2. If we both think there's a fit, we build one agent for one channel. Pilot runs 30 days, you see the numbers.
- Step 3. If the agent is closing more than it's costing at day 30, you decide whether to scale it. If not, you walk away with the diagnosis and the test data.
No long contracts. No "AI transformation" packages. One agent, one channel, 30 days, real numbers.
If your service business is doing $300k–$5M in annual revenue, you're owner-operating it, and you're tired of being the bottleneck, this is for you.
If you're at a different stage, this probably is not the fit, and I'd rather tell you that up front.