AI in jatid: worker-first, human-reviewed, minimal-risk
This page is our EU AI Act (Regulation (EU) 2024/1689) transparency statement, required for limited-risk AI systems from August 2026. It describes what AI jatid uses, what it does not use, and how a human stays in control of every AI output.
What AI jatid uses — and does not use
jatid does not deploy AI to make or influence employment decisions without a human deciding. That is not a policy promise — it is how the features are built.
✅ What jatid uses AI for
- AI work summary (M21) — turns your own tracked time into a draft narrative you edit before sharing. Worker-initiated, worker-reviewed, never auto-sent.
- Personal focus coach (M19) — trend lines from your own focus-block data, shown only to you.
- Auto-categorised focus blocks (M4) — statistical grouping of app-focus time into blocks, which you review and can reclassify.
🚫 What jatid does NOT use AI for
- No manager-facing automated scoring that affects employment decisions without human review (GDPR Art. 22; EU AI Act).
- No emotion recognition in the workplace.
- No social scoring of workers.
- No manipulative or deceptive techniques designed to distort behaviour.
- No AI output is ever auto-sent to a manager — a human always chooses to share.
Modules in the monitoring catalogue flagged as EU AI Act-relevant (see /settings/monitoring for the full catalogue): M13, M15 — every one of them requires a DPIA attestation and worker-first review before it can be enabled, and none is suggested for any business bundle by default.
Risk classification under the EU AI Act
jatid's AI features are minimal-risk or limited-risk under the EU AI Act, not high-risk. jatid does not deploy any of the practices the Act prohibits outright.
🟢 Minimal-risk: worker-only assistive tools
M19 (focus coach) and M21 (AI work summary) are assistive tools the worker uses on their own data, for their own benefit. Nobody else sees the output unless the worker chooses to share it.
🟡 Limited-risk: transparency obligation
Where an AI system generates content a person reads (M21's draft narrative), the Act requires that people are told it is AI-generated. We do — see "Transparency" below.
⛔ Prohibited practices — jatid does not build these
No emotion recognition in the workplace, no social scoring, no manipulative or subliminal techniques, no biometric categorisation to infer sensitive traits. These are explicit refusals recorded in our module catalogue, not gaps we haven't gotten to.
Human oversight guarantee
Any AI output a manager could ever see is advisory only. A human decides what happens next — jatid never auto-fires, auto-promotes, auto-schedules, or auto-flags a worker based on an AI output alone.
Workers see and correct their own AI-derived data
Everything an AI feature derives from your data is visible to you at /my-data. You can edit, dispute, or delete it — the same rights as any other data we hold about you.
Every AI-touching moment is logged
When an AI feature generates output, we record that it happened and that a human reviewed it. Admins can review this oversight log for their organisation to demonstrate compliance to an auditor or Datatilsynet.
Transparency — you are told when content is AI-generated
The M21 weekly summary is clearly labelled as an AI-generated draft. It appears in an editable text box, not a finished document — you review, correct, and rewrite before you send it anywhere. jatid never presents AI output as if a human wrote it without your involvement.
Questions? DPIA or AI Act assessment requests?
For DPIA documentation, AI Act conformity questions, or Datatilsynet inspection prep — write directly to dpo@jatid.dk.
Last reviewed: 2026-07-11 · Trust & Privacy · Worker disclosure