Talkify / Integrations / Coach

Your agent listens to its own calls.

Once a week Coach reads the calls your agent actually took, works out what is annoying callers and costing you jobs, and proposes a fix. It rehearses that fix against the calls that went wrong before it goes anywhere near your phone, and a week later it tells you whether it worked.

Start free → See a week of it

$39 a month to have it suggest, $69 to let it make the safe changes itself. Your first month is free and suggests only.

Runs: Weekly
First month: Free
Changes a week: 2 max
Tested first: Always

CoachWeekly review Monday

What it found

Bookings asked five questions before it offered anybody a time. Seven of twenty-three booking calls last week. Two callers hung up during the questions.

Questions before a time was offered5.2 The week before4.8

What it wants to change

Offer the next two free times first, then ask for the name and the number. Bookings › what to say

Rehearsed against the seven calls that went wrong. 6 of 7 reached a booking, against 3 of 7 on your build today.

Approve Decline

Declining is remembered. It will not ask again.

The Monday email. On Improve, a safe change like this one is already made and the email says so.

Nobody listens back to two hundred calls.

Your agent is answering every call, and mostly it is going fine. The trouble is the mostly. Somewhere in last week's calls there are four people who hung up during the questions, two who were asked for a number they had already given, and one who was told "let me put you through" to a phone that rang out. None of that shows up as a complaint. It shows up as a quiet month.

It reads your callers.

Not a general idea of a good phone call. Last week's real calls on your own line, against the week before them, with your own build open beside them so a finding can name the part that was talking.

It rehearses before it touches anything.

Test calls built out of the calls that went wrong, run against your build and the proposed one. The proposal has to do at least as well or it is thrown away without ever reaching you. Your live agent is untouched while that happens.

Two a week, and no more.

One per section of your instructions. A report with eleven proposals is one nobody reads, and eleven changes at once means you can never tell which one made the difference.

It marks its own work.

A week after a change goes in, the same numbers get looked at again and the next report says whether it did what it promised. If it did not, there is a button to put it back exactly as it was.

No means no.

Decline something and it does not come back next week in different words. It also leaves alone anything you have edited yourself in the last week, because you have just made a decision there.

Read. Propose. Rehearse. Mark.

Monday, before you are up. It reads the week's calls and the week before them as a baseline. A quiet line gets a fortnight, so a slow week is not read as a change.
It writes up what it found. Each finding tagged, so when the same thing turns up three weeks running you are told that rather than being told it once, three times.
It rehearses the fix. Against the calls that failed, on a copy. A proposal that does worse than your build today never reaches your inbox.
You approve it, or it makes it. Depending on which plan you are on, and on whether the change is a wording one or a structural one.
A week later it marks itself. Same numbers, honest verdict, and a button to put it back if the verdict is bad.

What it listens for

Too many questions before the caller gets what they rang for
Asking again for something already said
Asking for what the caller ID already knew
An email or a number taken and never read back
Talking over the caller
Silence after it looks something up
"Let me put you through", twice
A call that ends with no next step
Callers asking for something it cannot do yet
Hours, location or price answered wrongly

Each one is a named pattern, so recurrence is countable.

Tell you, or just do it.

Same reading, same rehearsal, same weekly report. The only difference is whether a safe wording change waits for you to press a button. Structural changes wait either way.

Suggest $39/mo

Nothing changes until you say so. The weekly email carries each proposal with the evidence behind it, and Approve or Decline on every one.

Weekly findings and proposals
Approve or decline by email
Every change rehearsed first
A verdict a week later

This is what the free first month runs.

Improve $69/mo

Safe wording changes go in on their own and the email tells you what changed and why. You are reading a report rather than working a queue.

Everything in Suggest
Safe wording changes applied for you
Structural changes still come to you
One button to put any change back

Anything with a price, a number, a transfer or an integration on it always asks.

Both are on top of your agent, with no setup fee and no lock in. Your first month is free, in Suggest mode, and starts from your first payment rather than the day you sign up, so it lands when your agent has actually taken some calls. Business agents are $129/month at the founding rate for your first 3 months, then $199, including 200 minutes of calls. Coach needs an agent that answers calls: it has nothing to listen to on its own.

Frequently asked.

The questions we hear most. If yours isn't here, ring our live demo agent on 03 242 1262, it'll either answer or queue a callback from a human.

What is Coach actually reading?

The calls your own agent took last week, and nothing else. It reads the summaries and the transcripts of that week's calls, compares them against the week before, and looks at your build to see which part of it was talking. It is not reading a general idea of what a good call sounds like and applying it to you. It is reading your callers.

What sort of things does it find?

The things that quietly cost you jobs. Asking four questions before the caller gets what they rang for. Asking for something the caller already said, or something the caller ID and your address book already knew. Taking an email address and never reading it back. Talking over people. Going silent after it looks something up. Saying "let me put you through" twice. Ending a call without the caller knowing what happens next. Callers repeatedly asking for something your build cannot do yet. It tags each finding so it can tell you when the same thing keeps happening.

Does it change my agent without asking?

That is the difference between the two modes, and it is your choice. On Suggest it changes nothing: you get the weekly email with Approve and Decline on each proposal and nothing moves until you press one. On Improve, safe wording changes apply themselves and the email tells you what changed. Even on Improve, anything structural always waits for you: adding or removing part of your build, rewiring it, merging agents, and anything touching a price, a phone number, a transfer destination or an integration.

How do I know a change will not make it worse?

Because it is rehearsed before it is applied. Coach builds test calls out of the very calls that went wrong, then runs them against your current build and against the proposed one. The proposal has to do at least as well as what you already have. One that does worse is thrown away and never reaches your inbox, let alone your phone. Your live agent is untouched the whole time this happens.

What if a change goes in and does not help?

It gets marked. A week later Coach looks at the same numbers again and tells you whether the change did what it said it would, in the next report. If it did not, you get a button to put it back. Coach keeps the original wording from before it touched anything, so putting it back is exact rather than an attempt to remember.

Can I tell it to leave something alone?

Yes, and it stays told. Declining a proposal is remembered, and it will not come back next week with the same idea in different words. It also leaves alone any part of your build a person has edited in the last week, on the reasoning that you have just made a decision there and it should not quietly undo it.

How many changes will it want to make?

At most two a week, and only one per section of your agent's instructions. That cap is deliberate. A weekly email with eleven proposals on it is one nobody reads, and eleven changes at once means you can never tell which one made the difference.

What does it cost?

Suggest is $39 a month and Improve is $69 a month, on top of your agent, with no lock in. Your first month is free and starts from your first payment, in Suggest mode, so you see what it finds on your own calls before you decide whether to pay for it. If you do nothing at the end of it, it stops.

Do I need anything else for it to work?

An agent that answers calls, because Coach has nothing to listen to without one. It is switched on from the Coach card in the agent builder, and it needs your agent to have been live long enough to have taken some calls. A brand new agent is the best possible time to run it, which is why the free month is at the start.

Who gets the weekly email?

Whoever you say. It defaults to the main address on your account, and you can point it somewhere else, so the person who actually looks after the agent is the one getting the report rather than whoever set the account up.

What happens to my call recordings in all this?

Nothing changes about where they live. Call audio, transcripts and summaries stay in New Zealand data centres, encrypted at rest with AES-256, in line with the Privacy Act. Australian numbers are handled in Australian data centres. Coach reads what is already there rather than sending anything new anywhere.