The AI Answering Service Case Nobody Makes, Missed Calls Are Wasted Ad Spend

A plumbing company came to us spending $14,000 a month on Google Ads with call assets on every ad. Cost per call was about $38. When we turned on call reporting and looked at connection status, roughly one paid call in four rang out. That's around 90 calls a month, or $3,400 in ad spend, going to voicemail. Nobody had checked because the ads dashboard reports calls, not answered calls.

That's the whole argument for an AI answering service, and it has nothing to do with the technology being impressive. It's a leak in a bucket you're paying to fill. This post sizes the leak, explains what ElevenAgents does about it, and gives you the break-even math so you can decide in ten minutes. We're ElevenLabs affiliates, and we run paid media for businesses whose phones are the whole funnel, so read the numbers with both facts in mind. For where this fits in the bigger picture, our B2B performance marketing guide covers the full funnel.

The Leak You Already Paid For

Every call from an ad has a price. Google charges the same CPC for a tap on a call asset as it does for a headline click, per the Google Ads call assets documentation. So a missed call isn't a missed opportunity in the vague sense. It's a specific invoice line with no lead attached.

The 26% figure comes from Invoca's aggregate customer data, and it's an average. Home services ran 27%. Some professional services categories ran far higher. If you'd rather not trust a call-tracking vendor's dataset, you don't have to. Your own Google Ads account has the number, which is the next section.

Speed matters even for the calls you do answer. The Harvard Business Review audit of 2,241 US companies found average lead response time was 42 hours, and companies that responded within an hour were 7 times more likely to have a meaningful conversation with a decision maker than those that waited even one more hour. A returned voicemail the next morning is already on the wrong side of that curve.

Size It From Google Ads Before You Buy Anything

Don't take a vendor's word for your miss rate. Pull it. Google's call reporting assigns a forwarding number to your ads and logs call duration, start time, and whether the call connected. Here's the five-minute version.

1. In Google Ads, turn on call reporting at the account level if it isn't on already.
2. Open the call details report for the last 90 days and export it.
3. Count calls that didn't connect, plus calls under your minimum useful duration. That's your miss count.
4. Divide total spend on call-bearing campaigns by total calls to get cost per call.
5. Multiply miss count by cost per call. That's your monthly leak in dollars.

For the plumber above, the math was 90 missed calls times $38, or about $3,400 a month. On a $14,000 budget, that's 24% of spend producing nothing. Most accounts we audit land between 15% and 30% on this test, and the owner has never seen the number because "calls" in the campaign view counts the ring, not the answer.

What an AI Answering Service Actually Does

An AI answering service picks up the phone with a conversational voice agent, not a phone tree. The caller talks, the agent talks back, and the agent can do things. ElevenAgents is the ElevenLabs version, and per the ElevenAgents documentation it runs four pieces together. A speech to text model to hear the caller, a language model to decide what to say, a text to speech model across 70+ languages to say it, and a turn-taking model that handles the pauses so it doesn't talk over people.

On top of that you get a knowledge base you upload your own documents into, tools that let the agent call your booking system or CRM, a visual workflow builder for multi-step calls, and automated tests so you can run your common call scenarios before a customer ever hears it. ElevenLabs' own claim is that you can ship voice agents in minutes, and in our setup the first working agent took an afternoon, most of it writing the knowledge base.

• Answers on the first ring, every time. ElevenLabs' line is "never miss a call with ElevenAgents," and for paid traffic that's the entire value.
• Qualifies before it books. ElevenLabs states the product can increase lead qualification by up to 30%. Configure the questions, the agent asks them.
• Works on your existing number. Native Twilio integration plus SIP trunking for Vonage, RingCentral, Telnyx, Bandwidth, and most standard providers.
• Scales without headcount. ElevenLabs' phrasing, and it's accurate. Concurrent call capacity is a plan setting, not a hiring decision.

If you want the full product family explained, including where Agents sits next to the Creative and API products, read our ElevenLabs products guide. For what makes a voice agent different from the chatbot on your website, we wrote up the AI agent vs chatbot distinction separately.

What ElevenAgents Costs

Every number below is from the ElevenLabs Agents pricing page as of September 3, 2026. Calls bill at $0.08 a minute on every plan. Go over your concurrent call limit and the overflow calls bill at $0.16 a minute. Text messages are $0.003 each.

Two things the table doesn't show. First, the language model cost is billed separately and varies by which model you pick, and telephony from Twilio or your SIP provider is passed through at cost. Budget another few cents a minute for both. Second, there's no cap on the number of agents you can build. You're constrained by minutes and concurrency, not by how many phone lines or departments you want to cover.

The Break-Even Math

Take the plumber. Creator at $22 a month gives 275 minutes. Average service call to book a job runs about 4 minutes, so that's roughly 68 answered calls before overage. Add $0.08 a minute past that. If 90 calls a month were going to voicemail, the full month costs about $22 plus 85 overage minutes at $0.08, call it $30 total, plus a few dollars of LLM and telephony.

The 30% booking rate is our conservative assumption, not an ElevenLabs claim. Even at 10% it's 9 jobs and $3,150 against $50 in cost. The point isn't the exact number. It's that the answering service costs less than one missed lead in nearly every service category we've run ads for, so the decision is about whether you're missing calls at all, and the Google Ads export already answered that.

When an AI Answering Service Doesn't Pay

• Your miss rate is under 5%. If a human answers almost everything, buy the Free plan for after hours and stop there.
• Your calls are complex from the first sentence. Legal intake with conflict checks, medical triage, anything where a wrong answer creates liability. Use the agent to capture details and book a callback, not to advise.
• Nobody will maintain the knowledge base. An agent that quotes last year's prices costs you trust. Someone has to own the document.
• You haven't tested it. ElevenAgents ships simulation and automated tests. Skipping them and going live on your main line is how you end up with a bad review that mentions a robot.

How to Roll It Out Without Breaking Anything

1. Run the Google Ads call export and confirm the leak is real and where it clusters.
2. Start on the Free or Starter plan. Build one agent with your top ten caller questions in the knowledge base.
3. Write those ten scenarios as automated tests and run them until the agent passes every one.
4. Route after-hours calls only. The alternative was voicemail, so the downside is near zero.
5. After two weeks, compare booked appointments from after-hours calls against the prior month.
6. Add overflow routing during business hours once after-hours is clean. Upgrade the plan when you hit the minute cap.

Verdict

If you'd rather have someone find the leak and fix the routing for you, that's the work we do. Our flat-fee Google Ads model covers call reporting setup and conversion imports, because a campaign that makes phones ring nobody answers isn't performing, whatever the dashboard says.