October 3, 2026 · Ken Gaffga, Founder
A Plumbing Company Hit Zero Missed Calls — and Lost $346K a Month. Here's the Lesson
Our last post made the case for fixing missed calls. This one is the necessary counterweight, because the loudest stories in trades AI right now are the success stories, and a business owner only hears the full picture by seeing both sides.
The Story
One plumbing company owner posted a cautionary account this fall: after switching to an AI receptionist, the dashboard showed zero missed calls — the system was answering everything. Monthly revenue told a different story, falling from roughly $1.16 million to $814,000, a $346,000 drop. Digging in, the booking rate had fallen 17 points. The AI was answering every call. It just wasn't converting them into booked jobs as well as the humans it replaced.
This is a single, self-reported account, not an audited case study, and we don't have visibility into everything else that changed in that business during the same window. Take the exact numbers with appropriate caution. But the shape of the story is worth taking seriously regardless of the precise figures, because it points at a real and common failure mode.
Why "Zero Missed Calls" Isn't the Same as "More Revenue"
This is the trap in a lot of AI-receptionist marketing: the pitch metric (calls answered) isn't the business metric (jobs booked). An AI that answers every call but can't handle a homeowner who mentions a locked gate, a warranty question, and an address in the same breath — or that sounds robotic enough to make an anxious caller hang up and try the next result — can technically "fix" the missed-call number while making the actual problem worse.
A plumbing company owner made this same point directly on LinkedIn this fall, pushing back on the "customers will hate talking to a robot" fear from the other direction: "Mine do not, and I own the plumbing company... here is what actually makes callers hang up. Not that it is AI. That it is bad." That's the real variable — not whether it's AI, but whether it's good. A well-configured AI receptionist and a poorly-configured one produce opposite outcomes on the exact same technology.
What This Means for How You Evaluate Any AI Tool
Don't let a vendor's dashboard metric stand in for your actual business metric. "Calls answered" is not the same as "jobs booked." "Leads captured" is not the same as "revenue closed." Before rolling out any AI tool that touches customers, agree on the real number you're trying to move, measure your baseline, and check it again 30 and 60 days after go-live — not just the vanity metric the tool itself reports back to you.
This is exactly the gap a diagnosis-first approach is built to catch: not "does this business need AI," but "does this specific implementation actually move the number that matters, or does it just move the number that's easy to measure."
Sources: Revenue-drop account via X/@freeconlon (single self-reported account, unverified — treat as illustrative, not audited data). Practitioner quote via LinkedIn/Lance Spriggs.