Insight · AI & automation

From call to ticket: how we connected Hey Pocket to our own CRM

I bought Hey Pocket because I wanted an easier way to capture my own calls. A summary sitting in a separate app does not fix a business process. So we wired the recordings into our own system. Together with Vincent, I show in The Ship Sh!t Show how a call automatically becomes a summary, tasks, a ticket and time logged against the customer.

Mitchell van Honk Mitchell van Honk · founder of Mantella
Published Read time7 min VideoThe Ship Sh!t Show · 12:17

Watch how we build the first version

Mitchell van Honk holding a blue Hey Pocket recorder, next to a 45 min build and the flow from recording to ticket
The Ship Sh!t Show · episode on this automation Video in English

Video in English · the write-up on this page is in English too. This episode of The Ship Sh!t Show is by Mitchell van Honk and Vincent.

  1. In this video
01

It started as something I needed myself

Hey Pocket is a small device that records conversations and turns them into a transcript and summary with AI. I bought one myself, with my own money, because I wanted to capture my calls more easily. No partnership, no sponsorship: Pocket is in this story because I use it, not the other way around.

The device does what it promises. The recordings and summaries still lived in a separate app, while our work happens in our own CRM and ticket system. Vincent and I built the connection that fixes that live, during a Ship Sh!t Show stream. We now use the automation ourselves.

Bought it myself, not gifted Video is not sponsored Built live during the stream
02

A summary is not automation yet

A recorder or AI app will give you a transcript and a tidy summary. That looks finished. In practice the work is only starting. That summary still has to go somewhere:

decide which customer the call belongs to
copy notes into your own system
create or update a ticket
log tasks and follow-ups
log the time spent on the call
send the information to colleagues
write a follow-up for the customer

The real time saved is not in recording the call, but in what happens automatically afterwards.

03

From conversation to processed work

This is the workflow we use today: from the moment a call is recorded until everything is processed in the CRM.

Input
1 Record the call with Pocket, or another source
2 Process audio and transcript summary and to-dos arrive automatically
Mantella automation
3Create an AI summary
4Recognise or select the customer
5Create or update the ticket
6Log tasks and follow-ups
7Log time against the customer
8Prepare a customer follow-up
04

What Pocket does and what Mantella adds

The split is simple: Pocket supplies the input, the automation does the work.

InputHey Pocket
Records the conversation
Reports every step via a webhook: recording, transcript, summary
Delivers the transcript, with speakers and timestamps
Delivers the summary: bullets, action points and mind map
That is the input for the process
The solutionMantella automation
Processes the information against your own business rules
Links the call to the right customer
Turns the content into the format you need
Creates or updates a ticket
Turns action points into real tasks
Logs time against the customer
Writes everything into your own CRM
Can prepare a customer follow-up
Fits your existing systems and way of working

Pocket captures the call. Mantella makes sure something actually happens with it.

05

What that looks like in our own system

This is the same CRM as in our Zendesk case: the system where tickets, tasks and hours already meet. The call automation writes straight into it. The screens below are mock-ups based on test data, not live customer files.

Incoming calls as tickets. Every Pocket recording arrives automatically as a “Call:” ticket in our own support system, ready to pick up.

Detail of a processed call. The AI summary with key points, call duration and ticket details. If recognition is missed, you link the customer in one click.

06

The first working version was up in about 45 minutes

In the video we do not start by generating code. First the AI coding agent is asked to inspect the existing Laravel project. It reads how our models, services, API connections and webhooks are built, then writes an implementation plan. That happens in read-only mode. Only after the plan is checked is the integration allowed to be built.

  • Webhook for immediate processing
  • API sync as a safety net
  • Webhook signature for security
  • Unique recording ID to prevent duplicates
  • Automated tests
  • Review by a second AI model

After about 45 minutes the first test call arrives as a ticket. The ticket holds the summary, the call duration and the action points that were found. That was the first working version, not the finished product. We refined the integration after the recording. We now use this workflow ourselves.

07

Where businesses can use this workflow

Support calls

Automatically capture a phone call, summarise it and attach it to a support ticket.

Sales calls

Put agreements, objections and follow-ups with the right prospect or customer automatically.

Client intakes

Turn wishes, decisions and action points into structured project information.

Internal meetings

Capture decisions and tasks without one person having to write up the whole meeting afterwards.

08

It does not have to be Hey Pocket

We do not sell a fixed Hey Pocket package. Our work is analysing, connecting and automating the full process. Where the conversation data comes from hardly matters:

Hey Pocket Microsoft Teams Zoom Telephony Another recorder Email Forms Existing business software

Your systems and way of working come first. Pocket is the entry point in this example. The automation behind it can be wired to almost any suitable system.

Case · custom software

Also see how we replaced Zendesk with our own CRM

Tickets, tasks, AI support and time tracking come together in the system that also receives this call automation.

Read the case
Get in touch