Your bot promised a customer something you never approved.

You didn't hear it. That call isn't in your dashboard, because nobody made it.

One call · we send you the transcript · your public number is all we need

The audit — $450. One payment. What's included →

Where the hole is

Your dashboard holds the calls that happened. The hole is where they didn't.

You have recordings, and that's true — most platforms include a transcript of every call, and opening them costs nothing. Open them today. That's the right first step, and we don't charge for it.

But those are conversations with people who got through to you. There is no call at 9:40 on a Friday night if nobody called you at 9:40 on a Friday night. And there's no question about a service you don't offer — which is exactly where a bot breaks.

A free one-call report: Thursday 7:42 PM, three questions and three answers, with the bot agreeing to a discount that was never offered.

You can't audit your own work. Not because you listen badly, but because you know what's supposed to happen.

Check my line

Free

One call. No cost, no commitment.

You give

An email, one number — the one already printed on your website — and a tick box saying we may call it.

We call

Once, a live person, as an ordinary buyer. The call opens with a recording notice.

You get

One page in your inbox. What we asked, what it answered, at what time, how many seconds to pick up. Word for word.

One call is not a sample, and we say so on the report itself. A bot can answer correctly once and break on the third try. The free call tells you whether there's anything worth talking about. The audit is thirty calls across a grid of hours.

After the audit

Five things change.

1. What you know about your own line

Now

It answers. Past that, it's a feeling. Nothing to check it against, nobody to ask.

After

Thirty conversations with quotes and timestamps. You can see which hour and which question it breaks on.

2. What «correct» even means

Now

There's no ruler. Nobody wrote down in advance what the right answer is, so every argument ends in opinion.

After

A twelve-field fact sheet, signed by you. Prices, what you do and don't do, hours, service area, phrases that are off limits. It stays yours.

3. The conversation with your vendor

Now

«I think it sometimes says the wrong thing.» — «Send us an example.» You send one, and you're told it's an isolated case.

After

A document signed by someone who didn't build that bot. Quote, the line it broke, what to change. Thirty calls, not one.

4. After the fix

Now

One thing gets fixed, something next to it breaks, and you find out a month later from a customer.

After

A re-run with its own column: what broke that wasn't broken before. Included in the price.

5. Who stands behind the finding

Now

A dashboard. The metric is calculated by the company that sold you the bot.

After

Every finding confirmed by a person and signed with a name. We didn't install this bot and we're not going to.

Start with the free call

Price

The audit — $450. One payment.

Nothing on the right is our number. Those are other people's published prices for the same work.

The work	Elsewhere
Thirty mystery calls across a grid of hoursFive windows: before opening, morning, midday, evening, late evening, weekend	$750
Setup: fact sheet, 120–200 scenarios, call script	$300
Review of thirty conversations against 34 checks	$150
A signed written protocol	$500
A second run after you make the fixes	$750
Included on top
Line PassportTwelve fields, yours forever. Anyone can audit you against it, including you	$150
Translated for your vendorThe fixes in his language, one page, so you don't have to argue in ours	$100
What broke that wasn't brokenIts own column on the re-run. Nobody else gives you this	$200
The night shiftPart of the calls land in the hours you never call yourself — Saturday, eleven at night, seven in the morning	$150
The 34-check listOpen before you pay. Audit us before we audit you	$0
If bought separately	$3,050
The audit	$450

What isn't on that list, and what we don't charge for. Recordings and transcripts of your own calls. You most likely already have them, included in your platform's plan. We sell the thirty calls that aren't in your dashboard, because nobody made them.

Start with the free call

Proof

This is what the protocol looks like. We ran it on our own agent.

You're paying for a document, and you're entitled to see what the document looks like. We can't show you someone else's — that's their business, and publishing findings about them isn't ours to do. We can show you ours.

We have a voice agent too. Her name is Vera, and we ran the audit on her. Three calls, thirteen minutes, five places where she broke the brief we had written for her ourselves. We fixed those.

What we don't have yet, and won't pretend to. The full thirty-call run on her, and the re-run after the fixes. Until both are done there is no «what broke that wasn't broken before» column to show you, and we're not going to show you half a document and call it proof. When it's finished it goes on this page whole, including the parts that don't flatter us.

A before-and-after protocol on our own voice agent: five violations found in three live calls, four of them shown, each with the fix applied.

Why this is here. It's a product sample, not a confession. And you can call her yourself and compare her to what the protocol says: +1 424 781 1913 (the call runs up to fifteen minutes, and she says goodbye herself when the time is up).

If what you need is to install an agent rather than audit one, that's a different page and a different price. We don't sell you a bot from here — the company selling the bot can't be the one auditing it.

Ten working days

How the audit runs.

Day 0

One free call and a one-page report with quotes.

Day 1

Twenty minutes on the phone. One question: what counts as the right answer here. Out of your answer comes the twelve-field fact sheet.

Days 2–6

Thirty calls across five windows. No more than two a day to one number. We give you the numbers we call from in advance, so you can filter them.

Days 7–9

Review. Every finding is first flagged by software, then confirmed by a person.

Day 10

The protocol. Quote, the line it broke, what to change. Plus a line saying what we didn't check — that one is always there.

Then

You fix, we run it again. One re-run included, sixty days.

Two things most people hide, and we say before you pay. The audit eats your minutes — thirty calls of two to three minutes is a noticeable share of a cheap plan's monthly allowance, and we reimburse them. And we check a sample, not every conversation that could ever happen.

Guarantees

We promise the work, not the verdict.

No protocol in ten working days — you don't pay

The clock starts when your fact sheet is signed, not when you pay. We return the full $450 within five working days. No form, no explanation needed. We don't hand over a half-finished protocol: either the document or the money, not both.

A clean report is the same result and the same price

If your line passes all thirty-four checks, you get a signed clean report with transcripts, dates and the numbers we called from. If it cost less, we'd have a reason to find something.

If your line blocks our numbers, that's a finding

It goes into the protocol as its own line, the protocol counts as delivered, and we don't ask for more.

Why we don't say «nothing found, nothing owed»

It sounds stronger. But we'd be the ones judging the condition, and a fee that depends on finding something hands your vendor a free argument: «they get paid to find something.» We sell independence. You can't take money for the finding and call yourself independent.

Second door

If you install bots for other people.

Then what you need isn't a check, it's a third-party signature. Your own dashboard won't convince a client — they know who's counting.

First agent

Audit, prompt fixes worked through with you, a re-run, and a before-and-after document you hand to your client. $1,200

From three agents

$300 per agent

Regression

A run after every prompt change. $250 per agent per month

We don't install voice agents and we're not going to. Your client stays yours, the prompt stays yours, and there isn't a line of our own pitch anywhere in the report.

Fair questions

Four you've already asked yourself.

«I have all the recordings in my dashboard»

You do, and that's true — most platforms include a transcript of every call. Open them today, it's free, and we don't charge for them. But those are conversations with people who got through. There's no call at 9:40 on a Friday night if nobody called you at 9:40 on a Friday night.

«Cekura and Hamming already test voice agents»

They do, and they can dial your public number — I won't pretend otherwise. Three things differ. They run your agent through scenarios you wrote, which tests your imagination, and the branch nobody anticipated never gets tested. Connecting them needs access to a system someone else installed for you — and if you had that access, you'd already be in there. And their log is your own log. You can't put it in front of a third party.

«You're doing this to sell me your own bot»

There isn't a line of our offer in the protocol. Easy to check: we ran the audit on our own agent and wrote down what it found — three calls, thirteen minutes, five violations of our own written brief. We fixed them. The thirty-call run on her and the re-run after the fixes aren't done yet, and we say so on this page rather than rounding it up into a finished case.

"I'm not sure this is actually my problem"

Then don't buy this one. We have a separate step for exactly that case — when everything hurts at once or you can't tell where: the deep diagnostic. A study of your business from open sources and your own answers, $500 once, credited against your first implementation payment. If we don't see where it pays off for you, we'll say so. What's in it →

Your bot promised a customer something you never approved.

Your dashboard holds the calls that happened. The hole is where they didn't.

One call. No cost, no commitment.

Five things change.

1. What you know about your own line

2. What «correct» even means

3. The conversation with your vendor

4. After the fix

5. Who stands behind the finding

The audit — $450. One payment.

This is what the protocol looks like. We ran it on our own agent.

How the audit runs.

We promise the work, not the verdict.

Why we don't say «nothing found, nothing owed»

If you install bots for other people.

Four you've already asked yourself.

Listen to one call first. Then decide.