Cases / Software
Built by Mantella, for Mantella

Why we cancelled our own Zendesk account, and in a few days built a system back ourselves.

A ticket system that works for the average customer rarely works for your own way of working. We built our own replacement, with AI, in a few days.

ClientMantella itself SectorSaaS replacement Servicescustom software, AI implementation Read time4 min
Custom CRM with tickets, tasks and AI support
Replaces Zendesk
Stack
Laravel PHP MySQL OpenAI
Since then in daily use, being expanded
01

The challenge

SaaS software like Zendesk is a fine start. But as soon as you grow, work with several people at once and want to expand further, you hit the limits of what the vendor designed for you.

The price rises, you pay for features you never use, you have to train your own AI agent inside their system, and every link to your other systems costs another API and another limit. And in the end all your data sits in someone else's format and system.

mockup
support · tickets
Tickets tuneurgency1
Export does not workKappers Collectief · Basis
High
Login screen loads slowlyJansen Retail · Pro
High
Add new userStudio Noord · Pro
Normal
Question about May invoiceDe Vries B.V. · Standard
Low
Screen 1 Ticket overview. Incoming tickets, linked to customer and subscription data. 1 filters by urgency.
02

What we built

We built our own ticket system, fully integrated into the CRM we already used for customers, subscriptions and invoicing. No more monthly bill, just a one-time build plus hosting. We knew exactly what we wanted, could connect it to our own way of working immediately, and have been expanding it steadily since.

mockup
ticket #2043
Export does not workKappers Collectief · Basis
boltCreate task1
arrow_forwardTicket → Task created
Time tracking00:452
auto_awesomeAI summary Customer cannot create an export. Next step: check the export setting and re-enable the connection. 3
Screen 2 Ticket becomes a task. 1 a task in one click, with 2 its own time tracking and 3 an AI summary of the next step.
mockup
thread · 20 messages
· 17 messages collapsed ·
summarizeAI summary
checkCustomer wants an extra user plus a shared overview.
checkApproval for the extra cost has already been given.
checkOnly waiting on the start date.
Screen 3 AI summary. A thread of 20 messages, reduced to the core.
mockup
compose reply
menu_bookInternal knowledgewhat we already know
travel_exploreLive researchsearches the web
Hello, you add an extra user via Settings. We shared the overview and scheduled the start date.
auto_awesomeProposal generatedSend
Screen 4 Compose reply. Based on internal knowledge, or on live research.
03

What the system does

Incoming tickets automatically scored for urgency, with direct escalation to the right people on high-urgency alerts
One ticket becomes a task in one click with its own time tracking, including an AI summary of the next step
Long email threads summarised to the core in one click
Compose replies in two ways: based on own knowledge, or based on live web research
All ticket data in our own database and control, never dependent on a third party's export format

The result

ThenZendesk
closeMonthly bill, regardless of use
closeLocked to the vendor's format and limits
closeTraining an AI agent only works inside their system
closeEvery link to other systems costs a separate API
closeData ultimately sits in someone else's system
NowOur own system
check_circleOne-time build plus hosting, no monthly bill
check_circleFully custom, precisely matched to our way of working
check_circleAI trained directly on own ticket data
check_circleDirectly integrated into our own CRM
check_circleAll data in our own database and control
A few days of build time · 3 to 4 people use it daily · ~10 tickets per day, with AI active

We once built this as a challenge: how far can you get with AI. Within a few days it was already more useful than what we had built up at Zendesk for years.

Mitchell van Honk
Mitchell van Honk · founder, Mantella
Get in touch
Sound familiar? We just build it ourselves. Book a call →