A parent calls your coaching academy at 8:40pm. Nobody picks up. She rings the next academy on her list and books a demo class there.
You see one missed call in the morning and no idea what she wanted.
This is the usual way a lead is lost on the phone. And this month the tools for fixing it got better.
What changed
On 28 September 2026, ElevenLabs released Eleven v4 and Eleven v4 Turbo, two text-to-speech models. The company says Turbo has a median time to first speech of about 150 ms and is built for real-time voice agents and live calls. v4 supports more than 90 languages. (source: VKTR)
Those are the company's own figures. But the direction is clear. The long, robotic pause that made AI calls feel fake is shrinking.
So an AI voice agent for business in India can now answer at 8:40pm, during a consult, or on a Sunday. Speed is no longer the main question.
The risk
A fast voice will say anything you let it say.
The model does not know your fees, your batch dates or which doctor sees which case. If it guesses, it guesses in a calm, confident voice. The caller believes it.
A wrong answer on a call is worse than a missed call. A missed call costs you one lead. A wrong fee costs you the lead and an argument later. We wrote about the same problem on chat: don't invent the price.
Three things to decide before voice goes on your number
1. What is it allowed to answer?
Timings, location, courses, how booking works. Only things you gave it. Nothing else.
2. What happens when it doesn't know?
It should say a person will call back today, then pass the whole conversation across. No guessing. See when they ask for a person.
3. Where does the call land afterwards?
In the same thread as the WhatsApp chat. If the call lives in one tool and WhatsApp in another, the lead gets asked the same questions twice. That is why we keep one memory across every channel.
The call should end with a next step
A call that ends with "thanks, we'll get back to you" is a missed call with extra steps.
It should end with a booked slot, or a clear next step and a note you can read. For a clinic: patient name, the problem in one line, the slot. For an academy: class, board, parent's preferred time. For a real estate desk: budget, location, visit day.
This is the same lesson as the missed call that is still a lead. The call is the start of the thread, not the end of it.
What this looks like on a normal day
A clinic's front desk is with a patient when the phone rings. Today that call goes to voicemail, or it rings out. With a voice agent behind it, the caller gets asked what they need, is offered two slots, and the desk sees a written note when they are free.
A real estate desk gets calls from people who saw a hoarding or an ad. They want to know the price range and whether a visit is possible on Saturday. The agent can take the budget and location, offer a visit slot, and flag the lead for a person.
An academy gets most calls between 7pm and 10pm, after school and after work. That is exactly when the counsellor has left.
None of this needs a clever voice. It needs a short list of things the agent may say, and a clear rule for everything else.
One honest limit
Tell callers they are speaking to an assistant. People forgive a short, plain line at the start of a call. They do not forgive finding out later. Also test the agent in the language your callers actually use before you put it live.
Where PROXe fits
PROXe answers on WhatsApp, Instagram, your website, email and voice. It qualifies the lead, books the call and follows up. It answers only from what you gave it. When it doesn't know, it hands over to you with the context attached. All channels share one thread. PROXe Core includes 250 voice minutes a month. You can read the full picture on what PROXe is.
Try this tonight
Call your own clinic or academy number at 8:30pm. See who picks up.
Then ask yourself one thing: by tomorrow morning, what would you know about that call?
If the answer is "a missed call and nothing else", book a demo and we will show you what the same call looks like with a desk behind it.
Questions people ask
Can AI answer phone calls for a small business?
Voice models are now fast enough for live conversation. Quality depends on what the AI is allowed to say and where the call is recorded.
What if it doesn't know the answer?
PROXe says a person will follow up and hands over the full conversation.
How much does it cost?
PROXe Core is Rs 9,999 a month + GST and includes 250 voice minutes. See pricing.
Related
Do not invent the price
A fee you made up is not a booking.
February 2, 2026 · 3 min read
When they ask for a person
"Can I talk to someone" is not a dump. The desk still asks, then hands the thread over.
June 12, 2026 · 3 min read
One lead, four channels, one memory
WhatsApp Monday. Instagram Thursday. Site Saturday. Call later. Same person. They should never repeat themselves.
May 5, 2026 · 3 min read
Recent
Only 13% see AI returns. Fix the first reply
Gartner found only 13% of companies see major returns from AI. For a clinic, academy or property desk, one job pays back first: the first WhatsApp reply.
October 23, 2026 · 5 min read
CureBay raised ₹100 Cr. Is your front desk ready?
CureBay raised ₹100 crore to grow rural clinics and add AI screening. Here is what a small clinic can fix first: the reply to the first message.
October 22, 2026 · 5 min read
A reminder is not a reply
You set a phone reminder to reply at 6. The lead needed a slot at 2. A reminder is a promise to yourself. The desk is a next line in their thread now.
September 22, 2026 · 5 min read
