Most businesses do not lose leads because their offer is wrong. They lose them because the lead filled in a form at eleven at night, nobody called until Thursday, and by then the person had spoken to two competitors. A voice agent exists to remove that gap entirely. It is also the automation people are most sceptical about, and some of that scepticism is deserved, so this covers both.
What it does on a call
- Calls within seconds of a lead arriving, at any hour, including the ones that come in on a Sunday.
- Qualifies against the criteria you set: budget, timing, location, whatever actually decides whether this is worth a human's time.
- Handles the ordinary objections, because most objections at first contact are the same four questions.
- Books straight into your calendar, so the lead never has to be called back to arrange the call back.
- Retries the ones who do not pick up, on a schedule, which is the part humans quietly stop doing after attempt two.
The result we can actually point at
Grow Strong MMA is an MMA gym in Australia running Meta Ads. They had good lead volume and no capacity to call everyone quickly, so conversions leaked out through slow follow-up. We built a voice agent that calls every new lead in under two minutes, qualifies interest, handles objections and books trial classes, with daily retries for anyone who does not answer. The measured outcome: 100% of leads contacted, with no human involvement, around the clock. That is their result. Yours depends on your volume, your offer and how good your list is.
Where voice agents are the wrong tool
This is the part most articles on the subject leave out, and it is the part that decides whether the project is worth doing.
- Complex consultative sales. If the first conversation genuinely requires judgement about a bespoke problem, an agent qualifying it is fine but an agent selling it is not.
- Very low lead volume. If you get four leads a week, pick up the phone. The build cost will not pay back and you will learn less.
- Anything where a wrong answer is expensive. Regulated advice, medical guidance, contractual commitments. The agent should be routing those, not answering them.
- Cold lists you have no relationship with and no permission for. That is a legal question before it is a technical one, and the answer varies by jurisdiction.
Cost, including the part that is not fixed
Build is typically $4,000-6,500 depending on how many systems it has to talk to and how complex the qualification logic is. Voice is the one automation where the running cost deserves real attention: you pay per minute of call time, so the ongoing number scales with your volume in a way a chatbot's does not. Model that before you commit. At low volume it is trivial; at a few thousand calls a month it is a line item you should have forecast.
The things that decide whether it works
- Speed. Under two minutes is the target, and the difference between two minutes and twenty is most of the value.
- Knowing when to stop. An agent that will not hand over to a human is worse than no agent.
- Being straight about what it is. Callers work it out anyway, and pretending otherwise costs you the trust you were calling to build.
- Retry logic that is persistent without being a nuisance. Three attempts across a few days, then stop.