voice bots · outbound calls and call handling

Half your calls never make it to a conversation: an answering machine, and the caller hangs up

A voice bot is not an answering machine, and it's not "press 1." It recognizes live speech, holds a conversation with a scripted flow that can flex, calmly pauses when interrupted, and hands the call to an operator along with the context. It calls a contact database and takes inbound calls. A minute of conversation runs about ₽3–4, against the payroll of a human operator. Automation roadmap — ₽0 and 14 days.

Get a free consultation

Case: consumer bankruptcy: lead loss from 60% to 15%
Minute of AI conversation: about ₽3–4 against operator payroll
Case: call monitoring: from 10 hours to 30 minutes a week

voice bots · outbound calls and call handling — AI implementation diagram

01Sound familiar?

The phone works eight hours a day, but customers call around the clock

Someone dials your number at 8:40 PM — right after work, when they finally got a free minute. They hear: "Our office is open nine to six." They hang up and dial the next number on the search results page. In the morning you see a missed call, call back at 11:15, and are told politely the matter's already been resolved. None of this shows up in a report anywhere: there's no line item in the CRM for "lost due to being closed."

During the day the picture's different, but the outcome is the same. A manager sits on outbound calls: confirming tomorrow's appointments, reminding people about visits, checking with someone from March's database whether their request is still active. The call takes forty seconds, dialing takes three minutes, and four out of ten numbers pick up. By lunch the manager is burned out on the monotony, by Friday the database call campaign gets pushed to next week, and it happens again every week.

Meanwhile the same person is also handling the inbound line, repeating the business hours, address, and standard service price for the twentieth time that day. This isn't sales and it isn't support — it's a front desk. You pay salary, taxes, and vacation time for it, and every one of these calls eats into time that should go toward the conversation where a deal actually gets decided.

02The cost of inaction

How much are you losing while you read this page

Don't calculate "how much a robot costs" — calculate how much it costs to leave a call unanswered, or an appointment without a reminder. Plug in your own numbers: how many inquiries come in per day, what share never reaches a live conversation, and what one customer is worth to you.

Inquiries per day 30

Lost before a reply 50%

Average order value 70,000 ₽

Lost revenue per month — per year — —

A rough estimate based on your inputs — we calculate exact numbers on a free consultation.

Get this money back

03How it will look

Your Tuesday, once the phone stops being a bottleneck

You open the telephony summary in the morning instead of a list of missed calls. Yesterday the bot made 340 calls: confirmed today's appointments, rescheduled four of them and set new slots in the calendar itself, checked with eleven people from the old database whether their request was still active, and handed them off to managers with a transcript of the conversation. There were 62 inbound calls, and the bot closed 41 of them on its own — business hours, order status, address, standard service price. The other 21 went to live people, each already with a filled-in card: who's calling, why, what they've already said. The manager picks up and continues the conversation instead of starting it over. You never once heard the phrase "we didn't get to that call" all day.

Now

Customer calls at 8:10 PManswering machine: "our office is open 9 to 6"

Remind about tomorrow's appointmentmanager calls manually, gets through half the list

Database of 4,000 old contactssits dead weight, never gets around to it

Inbound call with an off-topic questioneats up 6 minutes of live time

With AI

Customer calls at 8:10 PMbot answers on the first ring, logs it, puts it in the CRM

Remind about tomorrow's appointmentcall campaign runs itself, reschedules land in the calendar

Database of 4,000 old contactsgets called through in a week, warm leads go to the manager

Inbound call with an off-topic questionclosed by the bot, manager isn't even interrupted

04How it works

Three steps, and you don't write a single line of code

01 · Step 1 · 14 days, ₽0
Automation roadmap
We listen to recordings of your actual calls and break the telephony down into flows: how many calls follow the same pattern, how many happen outside business hours, how many minutes a week the database call campaign eats up. We flag which conversations a voice bot will take over completely, which it'll only start, and which shouldn't be touched. Separately, we check the legal side: where your numbers came from and what the consent form says. From there, the decision is entirely yours — no obligations.
02 · Step 2 · pilot
One scenario on live traffic
We take your highest-value area — usually appointment confirmations or handling inbound calls outside business hours. We write the scenario based on your actual conversations: lines, fallback branches, escalation rules, how it responds to being interrupted. We connect telephony and the CRM so the bot sees the customer's card before the call and writes the summary in after. We launch on a share of the flow, you listen to the recordings and fine-tune the wording. The metric is locked in before launch so there's something to compare the result against.
03 · Step 3 · full loop
Several scenarios and ongoing management
We add adjacent tasks: reactivating the old database, a post-deal survey, initial qualification of inbound calls. We set up handoff to a live operator along with the transcript, so the customer doesn't have to repeat everything from scratch. We review call recordings — both the bot's and the humans' — and flag where the scenario falls short. From there we keep monitoring, adjust the lines as your business changes, and stay responsible for the system's uptime.

05Proof

Inquiries came in around the clock, but got handled Monday to Friday, nine to six

Edemium case study · consumer bankruptcy

A law firm handling personal bankruptcy cases. The flow of ad-driven inquiries was steady, cost per lead was known and tolerable, and roughly one in seven inquiries turned into a signed contract.

The problem wasn't sales — it was the schedule. Someone in debt looks for a solution in the evening and at night — exactly when nobody's in the office. By morning they'd already reached out to three other places, and whoever answered first got the conversation. From the outside it looked like "poor sales conversion." In reality, sixty percent of leads were lost before the first live contact.

The owner was skeptical of AI, and for good reason: the topic is sensitive, the person is in crisis, and a wrong answer costs reputation and complaints. The condition he set was strict — the assistant answers only from an approved knowledge base, any legal question outside it goes straight to a human, and he personally reads every conversation for the first two weeks.

We built an assistant that ran the entire chain, not just individual replies: it took inquiries at any hour, gave a substantive answer within thirty seconds, gathered the details of the person's situation — debt amount, assets, number of creditors — created the CRM record with fields filled in, assigned a task to the lawyer, and booked the person into an open consultation slot. Sensitive or emotionally difficult inquiries it handed off to a live specialist along with a ready-made summary of the conversation.

After two weeks of reading the conversations, the owner dropped the daily oversight on his own. Lead loss before first contact fell from sixty percent to fifteen: what remained were people who wouldn't have continued with a human either. Contract conversion rose from fifteen to thirty-five percent — not because the machine sells better, but because the lawyer was now reaching people who'd already had a substantive conversation. Additional revenue came out to about three million rubles a month, on the same ad budget.

"I was afraid a robot would be talking to people in a hard situation. Turns out the worse problem was that nobody was talking to them at all until morning."

This is a real case; the company's name is withheld under NDA. The first contact in it was text-based — the mechanics are the same as a voice bot: round-the-clock intake, answers from an approved knowledge base, escalation to a human with context. Your numbers depend on your current response speed, traffic quality, and how many conversations follow the same pattern.

I want the same

06While you're thinking it over

What happens while the voice bot stays "for later"

Nothing dramatic — and that's the problem. A missed evening call doesn't send you an angry email — it just becomes a customer of whoever picked up. A customer who didn't show up because nobody reminded them doesn't count as a loss — it counts as a "no-show." A database of four thousand contacts doesn't fall apart loudly, it just goes cold: in a year, half the numbers won't remember who you are. Meanwhile acquisition cost climbs every season, and the edge shifts from "who spends more on traffic" to "who processes the traffic they've already paid for." A year from now your competitor's inbound line is covered round the clock, and their database campaign runs in the background without a single person — and what you'll have to catch up on isn't the technology, which is available to both of you, but the accumulated gap in conversations that actually happened.

07I know what you're thinking

Honest answers to the biggest doubts

Robots annoy everyone, my customers won't forgive it

Specific things annoy people: a robot pretending to be human; a robot that won't let itself be interrupted; a robot that reads out the main menu at the first live question. All of that gets fixed with the scenario. Our bot introduces itself as an assistant, goes quiet the moment a person starts talking, and transfers to an operator the moment the topic goes outside its scope. And there's a flip side: customers are far more irritated by not being called back at all. A voice bot isn't being compared to a perfect manager — it's being compared to ringing into the void after six in the evening.

Our sales are complex, a bot can't handle them

And it shouldn't have to. A voice bot works well where the conversation is short and predictable: confirming an appointment, reminding about a visit, reactivating an old database, initial qualification of an inbound call, a post-deal survey. It works poorly where you need to hear an objection and talk someone through it: complex consultations, emotional or conflict-heavy topics, selling something expensive. We separate the two streams: the bot handles volume and hands a live person a conversation that's already underway. We're not replacing the salesperson with a bot — we're freeing the salesperson from conversations that don't need a salesperson.

Cold-calling with a robot — isn't that illegal?

With a cold, purchased database — yes, and we won't take on work like that. Advertising law requires the subscriber's prior consent for marketing calls, and 152-FZ requires consent to process personal data; fines get issued for missing paperwork, not for the fact that a robot called. It's different with your own customers: confirming an appointment, a reminder, clarifying an order — that's contract fulfillment, not advertising. We check at the start where your numbers came from and what your consent form actually says, and tell you plainly which scenarios you can run and which you can't.

We already tried IVR, it didn't help

IVR and a voice bot solve different problems, even though both sit on the same telephony. IVR is a menu: it doesn't understand speech, it waits for a keypress and walks a tree you drew up in advance. A voice bot listens and understands an arbitrary sentence, asks clarifying questions, and changes course based on the answer. The difference shows up on the first unusual reply: IVR says "please repeat your input," a bot asks again like a person would, or hands off to an operator. If your past experience was a three-level tree, you haven't tried a voice bot.

If the bot makes a mistake on a call, I'm the one who answers for it

That's why the bot has narrow boundaries. It only says what's built into the scenario and the knowledge base; it isn't authorized to promise deadlines, discounts, or terms — that stays with people. Any complaint, any mention of a refund or a lawyer, an irritated tone of voice — the escalation rule sends it to an operator. Every conversation is recorded and transcribed, and for the first few weeks you listen to a sample yourself and fix the wording anywhere you don't like the answer. During the pilot the bot runs on a share of the flow, not all of it.

We don't have much call volume, will it even pay off

It might not — and we'll tell you that during the review, before any money changes hands. A rough benchmark: fifty or more similar conversations a day, or a regular database call campaign. Below that threshold, the savings on minutes won't cover building the scenario, and it's more honest to start somewhere else — reviewing call recordings, or a text-based assistant in messengers. At the automation-roadmap stage we calculate ROI on your own numbers and show you the math, rather than selling a bot to everyone regardless of fit.

✓

A risk-free pilot for you

We launch a pilot in 2 weeks. You set the success metric yourself. If you don't see results on it, we refund the pilot. The risk is on us, not you.

08Cost

What a voice bot costs, and when it pays off

A call center operator: salary and taxes every month, eight hours a day, 60–80 conversations a day on a good shift, vacation, sick leave, and turnover every six months. A voice bot: about ₽3–4 per minute of conversation, a hundred lines at once, the exact same script on call one and call three thousand, working weekends and at 10:40 PM.

Consultation

0 ₽Process review + ROI estimate

Funnel review and loss points
Lost revenue calculated in rubles
An honest read on where AI will pay off

Get it free

StartScenario pilot

from 90,000 ₽Payback in 2–4 months

from ₽90,000 — pilot: one scenario, integration with your telephony and CRM, launch on live traffic under your oversight. Plus conversation minutes — about ₽3–4 per minute

Launch a pilot

Turnkey

Custom quoteFull stack + ongoing support

from ₽250,000 — full implementation: several scenarios across inbound and outbound, tied into your CRM and telephony, handoff to an operator with context, recording review, retraining on accumulated conversations, and ongoing management

Discuss your project

09FAQ

FAQ on voice bots

How does a voice bot differ from an answering machine or IVR?

An answering machine plays a recording and understands nothing. IVR asks you to press a number and walks you through a rigid menu — the customer picks from options you set up in advance, and by the third level they usually hang up. A voice bot recognizes live speech: a person says "I want to move my appointment to Friday" — the bot understands the intent, checks an open slot, reschedules it, and confirms out loud. You can interrupt it mid-sentence, and it stops, like a person would. If a question is outside its competence, it doesn't dump you back into the main menu — it transfers you to an operator along with a transcript of what the customer has already said.

Will the customer realize they're talking to a robot?

Usually yes, within the first few seconds — the even intonation gives it away. We don't hide it either: the bot introduces itself as the company's voice assistant. What actually irritates people is something else — when a robot is passed off as a human and then hits a dead end on the third line. The practical rule is simple: if the conversation is short and to the point (confirm an appointment, remind about a visit, clarify an address), the format doesn't bother anyone. If the bot tries to run a half-hour consultation, irritation is guaranteed — and we don't take on scenarios like that.

How much does a minute of conversation cost, and what does the price consist of?

Our voice AI runs at roughly ₽3–4 per minute of conversation: that covers speech recognition, the language model, and voice synthesis. Telephony is separate — the carrier and the number each have their own rates. The total is simple to calculate: average call length multiplied by the number of calls. Calling 1,000 contacts at a minute and a half each runs about ₽5,000–6,000 in conversation minutes. Scenario development and integration are a one-time pilot from ₽90,000 — a single payment, not a subscription.

Is it legal to call customers with a voice bot?

Calling your own customers about operational matters — confirming an appointment, reminding about a visit, clarifying order details — is legal: it's contract fulfillment, and no separate marketing consent is required. The moment a call includes an offer to buy, it becomes advertising, and under Russia's advertising law you need the subscriber's prior consent specifically for marketing calls, plus consent to process personal data under 152-FZ. Purchased contact databases can't be used. We check at the start where your numbers came from and what boxes were checked on the consent form, and help fix the wording if needed.

Can the bot connect to our CRM and telephony?

Yes, that's a standard part of implementation. We work with amoCRM and Bitrix24 through open APIs: the bot reads the deal card before the call, writes the call summary into it, changes the status, and assigns a task to the manager. Telephony — Mango, Telfin, Beeline, and other SIP-based carriers. Recordings and transcripts of every conversation land in one place, and you can listen to a sample or run them all through a separate call-analytics scenario.

What happens if the bot doesn't understand the person?

First, it asks again — once, phrased differently. Second, if it still doesn't understand, or the topic is outside the scenario, it transfers to an operator and passes along everything it already learned. The escalation rule is set up in advance: a list of topics, trigger words like a complaint or a mention of a lawyer, and a response to irritation in the voice. The bot doesn't silently hang up or loop a customer around — that gets specifically checked during the pilot.

How many calls a day do we need for this to pay off?

A rough benchmark: fifty or more similar conversations a day, or a regular database call campaign. At that volume, the bot takes over work equivalent to one or two operators, and the gap between ₽3–4 per minute and a full payroll becomes noticeable within the first month. If you get 15 calls a day and every one is unique, you don't need a voice bot — it's more honest to set up proper call recording and have AI review it. We'll tell you this on the free consultation.

How fast does the first scenario launch?

A pilot for one scenario is a matter of weeks, not six months. First we listen to your actual calls and write the scenario the way your people actually talk, not from an internet template. Then we build the bot, run it against you and your team, and fine-tune the wording and pauses. Next comes launch on a small share of live traffic — you listen to recordings and tell us what you don't like. Only after that does the volume increase.

Be the first in your niche to answer a call on the first ring

Voice bots have stopped being an experiment — they've become a question of discipline. While most companies in your niche are still running on an answering machine after six and a call campaign "whenever there's time," the difference between you and your competitor gets decided in the first twenty seconds of the call. Not in the quality of the synthesized voice — everyone's voices sound about the same. It's in how many calls go unanswered, and how many appointments fall through without a reminder. This won't give you a dramatic revenue spike — it removes losses you can't currently see in your reports: a missed call leaves no trace anywhere except in the numbers of whoever picked up.

I want to be first in my niche

10Connection point

Let's break down your calls for free

14 days, ₽0, no obligations. We listen to your actual calls, calculate how many minutes go toward confirmations and reminders, and show you on paper: which conversations a voice bot will take over completely, which it'll only start, and which shouldn't be touched. If you get 20 calls a day and a bot won't pay off, we'll say so directly during the review — not after you've paid.

Message us on Telegram