evidoria

← Back to browse

Good practice

Agentic AI Hub — Germany's Federal Pilot for Agentic AI in Municipal Administration

Germany · Berlin · See the Germany profile

Germany's federal digital ministry (BMDS) piloted agentic AI across 19 municipalities with 9 AI startups: GDPR-request processing cut over 90%, care-application processing cut 45%, and 600 administrative processes modeled in 12 weeks, with 95% municipal satisfaction.

>90 %
GDPR request processing time reduction
28 hours
Time saved per GDPR case
45 %
Care-assistance application processing reduction
31 %
Active processing time reduction (care applications)
78 %
Meeting-protocol post-processing reduction (45 to 10 minutes)
1 EUR
Cost per AI-assisted care pre-screening
600
Administrative processes modelled (12 weeks)
264
Process models produced (Flensburg)
70 %
Solutions in operational use at pilot end
65 %
Solutions implemented within 4 weeks
80 %
Solutions processing real municipal data
95 %
Municipal satisfaction
15 of 18 cases
Staff reporting they felt secure using AI tools
Agentic AI Hub — Germany's Federal Pilot for Agentic AI in Municipal Administration

Details

Maturity
Pilot
Promoter
Bundesministerium für Digitales und Staatsmodernisierung (BMDS)
Period
March–May 2026 (three-month pilot)
Keywords
agentic AI, municipal government, administrative process automation, GDPR compliance, digital public administration

Context

Between March and May 2026, Germany's Federal Ministry for Digital Affairs and State Modernisation (BMDS) ran a three-month 'Agentic AI Hub' pilot pairing 19 municipalities with 9 AI startups across 20 pilot projects, selected from nearly 600 initial submissions.

Objectives

To test agentic AI tools on real municipal administrative processes, including GDPR information requests, care-assistance applications, meeting-protocol drafting, and process modelling.

Activities

20 pilot projects tested agentic AI on specific administrative workflows; separately, seven municipalities used AI tools to model roughly 600 administrative processes over 12 weeks, work that conventionally takes three or more weeks per process.

Results

GDPR information-request processing time fell by more than 90% (saving about 28 hours per case); 'Hilfe zur Pflege' care-assistance application processing fell 45% overall (active processing time down 31%); meeting-protocol drafting cut manual post-processing from 45 to 10 minutes (78% reduction); AI-assisted pre-screening of care applications cost around €1 per request; Flensburg alone produced 264 process models. By pilot end, 70% of tested solutions were in operational use, 65% had been implemented within four weeks, 80% were processing real municipal data, municipalities reported 95% satisfaction, and staff in 15 of 18 cases said they felt 'secure' using the tools.

Conclusions

BMDS assessed that roughly half of the tested solutions could transfer to other municipalities and most pilots had already started follow-on projects, but a full evaluation report and a planned 2027 dynamic procurement system to scale nationally were still pending at the pilot's conclusion.

Implementation

Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.

Data sources

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

Attachments

Similar practices you may find useful