Use case · Software, AI, operations
Anonymised, like all our use cases
Every mailbox, one service desk.
A company with several brands receives customer requests in many mailboxes. We built a service desk that collects every email as a ticket, uses AI to route it to the right channel and suggests draft replies. The team checks, adds and sends.
Deployed in
Customer service, several brands
Services
Software, AI, operations
Roles
Support, team lead, admins
Languages
German, English, Ukrainian
The AI sorts and suggests. The team replies.
Incoming emails become tickets in the right channel, with a recognised customer number and a summary. One click creates a draft reply with a confidence score. Each sender decides whether an unchanged AI draft may go out.
How a ticket runs.
- Intake The desk collects emails from every mailbox. Fixed rules apply first, anything suspicious goes into spam quarantine.
- Route The AI assigns the email to a channel and recognises the customer number. Watchdogs move special cases to their own channels.
- Reply Templates, signatures per sender and AI drafts at the push of a button. Whoever works on a ticket blocks sending for others.
- Review AI replies sent unchanged stay in controlling until someone marks them as reviewed.
- Evaluate Insights show backlog, first responses and trends, only as totals per team and channel.
1 click
to an AI draft with a confidence score
3
roles with clear visibility rules
5 min
is the most the insights can lag behind
0
figures on individual people
Insights instead of Excel exports.
A dedicated page shows what is happening right now: open tickets, missed first responses, replies due. Below, sentences explain the biggest changes and charts show the trends. There are deliberately no figures on individuals.
One application, six areas.
From the inbox to the report in one place, on the same platform as the ticket system, shift planner and AI agent.
Software
Ticket list
Channels and teams as tiles, search, filters and multi-select across all results, plus the archive.
AI
Replies with AI
Draft and summary at the push of a button. A separate prompt per channel for tone, permitted actions and escalation.
Automation
Rules and watchdogs
Email rules based on fixed patterns, watchdogs with an AI question, spam quarantine with restore.
Software
Controlling
Review AI replies that were sent and mark them as reviewed, one at a time or many at once.
Software
Activity
A log of every status change, assignment and spam flag, live on the cockpit.
Platform
Administration
Channels, teams, response times per status, signatures, senders and mailboxes.
What makes the difference.
People send
The sender decides whether an AI draft may go out unchanged. Whatever went out that way ends up in controlling.
Team totals, not surveillance
Reports show totals per team and channel, no figures on individual people.
No duplicate work
Whoever replies blocks sending for others. Other tickets from the same address are listed right below.
Visibility by role
Support sees their teams’ tickets, team leads see all. The same rule applies to figures, search and activity.
What we built it with
- .NET 10
- Blazor
- .NET Aspire
- PostgreSQL
- Redis
- MailKit
- Microsoft Graph
- Docker
- GitLab CI
- Ansible
- OpenTelemetry
Where are the emails piling up?
In an intro call we look at how your requests can reach the right team faster.