The problem
The owner runs a commercial insurance brokerage in Chicagoland. Most of its customers are contractors, and they call because a job is waiting on a certificate or a quote. Nearly all of that business comes in by phone. Before a dollar went into ads, I wanted to know how many of the people already calling reached anyone. The phone system's own call log, read one call at a time, looked a lot worse than its summary reports. In the full weeks of June and July, 2,136 calls hit the main line during weekday business hours. Half (1,077) reached a person. Nearly 4 in 10 (817, or 38%) reached nobody and left nothing. No answer, no voicemail, no message.
What I found
None of the causes showed up in a summary report. Every one of them was sitting in the call log or the account settings. - A lunch rule. A company rule sent calls between noon and 1 p.m. straight to a voicemail box. Not one of the 384 lunch-hour calls in June and July got answered live. - A clock an hour off. The owner's extension was set to a time zone that ignores daylight saving. A 9-to-6 schedule ran an hour late, and calls before 10 a.m. went to voicemail without ringing. The same setting was on 18 of 28 users. - A number that looked like spam. On outgoing calls, 10 of the firm's 32 numbers showed the placeholder caller ID "NO CALLED ID". That's the kind of label that gets a call back treated as spam.
What we did
The brokerage's people still answer first. I changed what happens when they can't. The front desk rings first. What it misses goes to an AI assistant that answers in English or Spanish. It puts callers through to the person they ask for by name, rings the whole team at once when a caller wants a quote, and otherwise takes a structured message. The support teams ring all their members at once for 35 seconds before voicemail. Before real callers ever reached the assistant, it had to pass 14 scripted test calls, including attempts to get pricing and coverage advice it has to refuse. 12 passed. The two failures got fixed and tested again. A drill took the assistant's line down on purpose and found a dead end, so we built a backup route the same day and proved it in a second drill. No customer was affected. All 28 users run on the correct time zone now. All 32 numbers show one caller identity, and 29 are registered with the major caller-reputation providers. A 911 call from an office line now alerts the front desk by text and email. Before, it alerted nobody.
The numbers
First, how I measured. A call counts as reaching a person only when one of the brokerage's staff picks up. An answer by the assistant never counts as a person. The formula and the pass bar (15% or fewer callers reaching nobody) were written down and approved before the first after-number existed. The owner's weekly call report uses the same math, over the same line and hours, on both sides. In the four full weeks from August 17, 936 calls hit the main line in business hours. 7 in 10 (651) reached a person, and those were the brokerage's own staff. About 1 in 8 (118, or 13%) reached nobody and left nothing. The rest left a voicemail or a message with the assistant. Week by week the share reaching nobody read 13%, 14%, 11% and 13%. Under the bar every week. Lunch didn't go away. It changed shape. In the week of September 7 the assistant picked up all 33 lunch-hour calls, but only 4 of those callers reached a person, and that hour still held 29 of the week's 61 calls that didn't reach a person. The support teams were confirmed on real calls in September and they're still holding. From October 12 to November 10, all 65 calls that rang a team for its full hold stopped on time and went to that team's voicemail, the way it was set up. This is the phone work behind the AI phone receptionist for contractors and home services in Chicago.
What I can't show yet
Call volume was lower after the rebuild, about 234 business-hours calls a week against about 310 in June, and I haven't pinned down why. Several changes landed in the same weeks, so these figures show what they did together. Four weeks is a start, not a full season. I also haven't measured whether the caller-ID registrations change how carriers label the numbers.
You can start this yourself. Export one month of your call log and count, hour by hour, the callers who reached a person, a voicemail or nothing. Or send me the month and I'll send back the same picture this owner started with. [Send your call log to No Brainer Media](https://nobrainersmma.com/contact/).