AI-supported casework

Casework with AI you can stand behind.

Lynus prepares the case and shows the sources. The caseworker reviews and signs. Every step is logged, readable in Norwegian.

Case 2026-0142 · Example case 5 of 5 steps
  1. Case read · 14 pages 09:12 · AI
  2. Research · 6 sources 09:14 · AI
  3. Roles and authority · Brønnøysundregistrene 09:14 · AI
  4. Rule check · 1 deviation logged 09:15 · AI
  5. Reviewed and signed · [Name] 09:41 · HUMAN

Sound familiar?

  • The assessment differs from one caseworker to the next.
  • Knowledge disappears when people leave.
  • Management says AI. The data protection officer says no.

How it works

From incoming case to decision, in four steps

Case 2026-0142 · Example case

  1. 01 · AI

    The case comes in

    Lynus reads the case and its attachments and places them in the workflow you have described.

    • Case read · 14 pages 09:12 · AI
  2. 02 · AI

    The case is prepared

    Research and cross-checks. Every claim shows the source it rests on.

    • Research · 6 sources 09:14 · AI
    • Roles and authority · Brønnøysundregistrene 09:14 · AI
  3. 03 · AI

    Deviations are flagged

    Anything that breaks your rules is stopped and logged. The rule has the last word.

    • Rule check · 1 deviation logged 09:15 · AI
  4. 04 · HUMAN

    The caseworker signs

    The assessment is always a person’s, with name, time and a full log.

    • Reviewed and signed · [Name] 09:41 · HUMAN

You describe the work. Lynus builds the workflow.

Tell us how you assess a case, in everyday words. Lynus proposes the steps, you adjust them, and the workflow is ready. No tool to learn.

Example: one sentence in everyday words becomes three proposed steps

Example

“We check that the applicant is registered, that the project fits the scheme and that the budget adds up.”

Proposed steps · AI

  1. The applicant is registered
  2. The project fits the scheme
  3. The budget adds up

The rules have the last word.

An instruction is a hope. A rule is a control. In Lynus your rules run as a control: anything that breaks them is stopped and logged before the caseworker sees the case.

Case 2026-0142 · Example case

Deviation from the regulations § 4 · stopped by a rule, not by AI

Trust

  • A source reference on every claim Available

    The AI reads the case, does the research and shows the source for every claim.

    Read more
  • A full log of every step Available

    The events in the case are stored in a log that can only be added to. The database rejects changes and deletions.

    Read more
  • A person signs every assessment Available

    Who reviewed and signed, and when.

    Read more
  • Case data stored at Hetzner in the Nordics Available

    Both sit on a server at Hetzner in the Nordics, set up with Coolify.

    Read more
  • Use your own Azure OpenAI, or ours Planned

    The plan is to move to Azure OpenAI in the Sweden Central region before Lynus processes real personal data.

    Read more
  • No training on your data Available

    Your data is not used to train the model.

    Read more
  • DPIA material Available on request

Read more about trust and security

Where we stand

“We are a start-up. We have no customer logos to show yet, and we are not going to pretend. What we can show is the product, in a pilot in your own casework.”

Sam Dirksz, founder. Formerly senior adviser at Innovation Norway.

How to get started

  1. A conversation
  2. We set up the first workflow together with you
  3. A pilot in your own casework