Use case · AI development, software, infrastructure

Anonymised, like all our use cases

AI analysis for every service call.

We developed a platform for customer service and operations, and we run it ourselves. Its AI module analyses every service call: it transcribes, summarises, checks for fraud and pre-scores the quality of the conversation. The AI models run on our own hardware.

Deployed in

Retail and e-commerce

Services

AI, software, infrastructure, operations

Modules

6 applications, one login

Languages

German, English, Ukrainian

In use since

2026

The AI prepares. The team decides.

For every call, team leads see the summary, the risk score and a pre-scored checklist. They confirm or correct it and derive training topics for individual agents.

How the AI works.

  1. Import Recordings arrive automatically from the call-centre software or by upload.
  2. Transcription Speech recognition transcribes the call, and every word gets a timestamp.
  3. Understanding A language model assigns each sentence to agent or customer and summarises the conversation.
  4. Checking Order and customer numbers are detected, matched against the ERP and checked for signs of fraud.
  5. Scoring The AI pre-scores the quality checklist and counts power words. The team confirms.

≈ 47 s

Analysis of a 3-minute call, from recording to scoring

21

Checkpoints per call, pre-scored by the AI

0–100

Risk score, justified by detected fraud signals

1

in-house GPU server instead of external AI services

One platform, six modules.

All applications share login, roles and permissions. New modules plug into the same foundation.

AI

Call analysis

Call analysis with transcript, summary, fraud check, quality check-up and training record for each agent.

AI

Service desk

Email helpdesk with channels, rules, SLAs, text templates and a spam filter. AI replies go out only after approval.

Software

Ticket system

Tickets and project boards for every team, including external requests, kept separate per client.

Software

Staff and shifts

Employees, companies and teams with contracts, shifts, work calendars and leave.

Software

Project management

Plan and track projects, with members and roles.

Platform

Login, permissions and links

One central login for all modules, roles and API access, plus quick links for the whole portal.

What makes the difference.

AI on our own hardware

Speech recognition and the language model run on our GPU server, not with a cloud provider.

People decide

The AI pre-scores, the team confirms. AI replies go out only after approval.

Connected to the ERP

Detected order and customer numbers are checked directly in the ERP.

Traceable

Every analysis stores the model version, and every change goes into the log.

Certified data protection and security

Data protection officer (IHK)

Information security officer (TÜV)

Achim Groeling, Managing Director

What we built it with

  • .NET 10
  • Blazor
  • .NET Aspire
  • PostgreSQL
  • Redis
  • Docker
  • GitLab CI
  • Ansible
  • Qwen3-ASR
  • llama.cpp
  • xUnit and Testcontainers

Where could AI give your service team a break?

In an intro call we show you the platform and work out what pays off for you.

Book an intro call