Local AI

Pilot

An AI assistant that keeps your data in the company

A language model runs on your server next to the portal. It triages tickets, answers from your policies and summarises long threads, and the text goes nowhere.

Features

What the assistant does

Capabilities arrive in the portal as the pilot progresses. Ticket triage and document answers come first.

01

Ticket triage

From the ticket text the assistant suggests the team, category and priority and finds similar solved cases. The agent confirms with one click.

02

Reply drafts

For routine questions the assistant drafts a reply from the help centre and past solutions. A person always sends it.

03

Summaries

A long ticket or document thread folds into a few lines: what happened, what's done, what's left.

04

What changed in a version

During approval the assistant explains in plain words how the new contract version differs from the last.

05

Ask your policies

Staff ask in their own words and get an answer from internal documents with a link to the clause.

06

Writing help

Drafts of news posts and announcements, style fixes and translation.

Principles

Helpful, but under control

  • The model runs on your server. Ticket and document text never goes to cloud services.
  • Access rights are respected. The assistant answers only from documents the employee can already open.
  • A person decides. The assistant suggests but never sends or approves anything on its own.
  • You see where answers come from. Every answer shows its sources, and assistant actions are marked in the history.
  • Can be switched off per module. The administrator turns the assistant on where it's needed.

Questions

Questions about AI

Which model is used?

An open language model that runs on your server. We pick the specific model and server requirements for your data volume and tasks during the pilot.

Do I need a GPU?

A server with a GPU gives fast answers. Small models also run on CPU, but slower.

Is the model trained on our data?

No. The model isn't fine‑tuned; it reads the relevant documents at question time and sends nothing outside.

When will the assistant be generally available?

It's in pilot now: the model is being built into the portal and tested on real tickets. Write to us if you'd like to join the pilot.

Bring it all into one window

We'll show every module on a demo stand with sample data, walk you through the local AI pilot and talk through installing it in your infrastructure.

[email protected]