evidoria

← Back to browse

Good practice

Public AI — Austria's Federal Administration Programme for GovGPT and Four Other AI Tools

Austria · Vienna · See the Austria profile

Austria bundled five concrete AI tools, led by the sovereign GovGPT assistant for 180,000 federal employees, under one "Public AI" programme launched March 2026. Only one tool is live so far, and no usage or productivity results have yet been published for any of them.

44 %
Civil servants due to retire within 13 years (next 13 years)
180,000 federal employees
Target users for GovGPT (targeted H1 2026)
13,000 users
Target users for KI im ELAK (targeted H2 2026)
6,000+ pages
Training material pages covered by KI-SUN (at launch)
1 of 5 planned tools
Tools live at time of reporting (as of publication)
Public AI — Austria's Federal Administration Programme for GovGPT and Four Other AI Tools

Details

Maturity
Pilot
Promoter
Federal Chancellery of Austria — Digitalisation State Secretariat, with the Federal Computing Center (Bundesrechenzentrum)
Period
Announced March 2026; phased rollout of five tools through end of 2026 and beyond
Keywords
generative AI, sovereign AI infrastructure, civil-service productivity, e-government

Context

Austria faces a looming civil-service staffing crisis: 44% of civil servants are due to retire within 13 years, motivating a shared federal AI infrastructure rather than fragmented per-ministry tools.

Objectives

Public AI bundles all of the federal administration's AI projects onto one shared, sovereign infrastructure operated by the Federal Computing Center (Bundesrechenzentrum).

Activities

Launched 19 March 2026 at the 'KI-Konklave' event by State Secretary for Digitalisation Alexander Pröll, the programme comprises five applications with staggered rollout dates: GovGPT, a sovereign generative-AI chat and summarisation tool for all 180,000 federal employees (targeted H1 2026); KI im ELAK, adding AI query/summarise functions to the electronic file system for around 13,000 users (H2 2026); KAPA, supporting drafting of responses to parliamentary inquiries under a human-in-the-loop principle (H2 2026); KI-SUN, a personal knowledge assistant covering more than 6,000 pages of federal training material, already live; and a planned agentic AI layer for the following year.

Results

As of publication, only KI-SUN has actually launched; GovGPT, KI im ELAK and KAPA remain pre-release.

Conclusions

The case is a real, multi-sourced, government-wide commitment of infrastructure and target user counts, but not yet an evaluated deployment — neither the government's own announcement nor independent ORF coverage cites usage figures, error rates, time savings or user-satisfaction data for any component.

Implementation
Indicative cost
High (€500k–€5M) — No budget or cost figures disclosed; infrastructure is operated centrally by the Federal Computing Center rather than funded per-ministry.
Time to results
Medium (1–3 years) — Launched 19 March 2026; GovGPT targeted for H1 2026, KI im ELAK and KAPA for H2 2026, with an agentic AI layer planned for the following year.
Staffing & skills
Federal Computing Center (Bundesrechenzentrum) — shared infrastructure operator, Digitalisation State Secretariat (political ownership, State Secretary Alexander Pröll)

Conditions for success

  • shared, sovereign, cross-ministerial infrastructure instead of per-ministry point solutions
  • an explicit human-in-the-loop principle for higher-stakes tools such as KAPA
  • a staggered rollout starting with a lower-risk personal knowledge assistant (KI-SUN)

Common failure modes

  • four of five planned tools remain pre-release with no usage or performance data yet
  • no independent oversight or audit mechanism yet public

Where it fits

Governance type
federal government, centralised infrastructure
Scale
national, targets 180,000 federal employees
Income level
high-income

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful