Document Processing · IT Operations Automation
The BA needed a privacy-compliant way to reduce the manual effort of turning Requests for Change and user stories into high-quality Jira tickets under increasing workload and limited staff capacity.
Reported impact
The BA automated most ticket creation work, reduced manual effort on summarizing and formatting, and established a reusable on-premises AI agent pattern that fits public-sector privacy and security requirements.
The German Federal Employment Agency worked with Capgemini to automate RFC and user-story conversion into Jira tickets using a coordinated on-premises multi-agent architecture. Incoming change requests are first analyzed by a reader agent, then broken into steps by a planner agent, transformed into structured Jira content by a creator agent, and checked for consistency and duplicates by a reviewer agent before human approval. The deployment was built to manage large RFC documents and LLM token limits while fitting the agency’s privacy and security requirements. After pilot integration with selected departments, the system was fully connected to the agency’s local Jira environment and orchestrated with CrewAI using privacy-compliant models including Aleph Alpha, LLaMA, and Mistral. This lets the BA keep all data on premises while automating repetitive formatting, summarization, and copy-paste work so staff can focus on oversight and decision-making.
RFCs and user stories are read and extracted by the reader agent, decomposed by the planner agent, turned into structured Jira tickets by the creator agent, checked by the reviewer agent, and then sent to human reviewers before entering the agency’s local Jira instance.
"Capgemini and the BA developed a multi-agent AI system that automatically transforms Requests for Change and user stories into Jira tickets, supported by human oversight."
"Privacy-compliant models such as Aleph Alpha, LLaMA, and Mistral are orchestrated via CrewAI, an open-source platform for multi-agent coordination."
"Today, the system is fully integrated into the BA’s local Jira instance and operates entirely on-premises."
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Capgemini Sample Jun 9, 2026 Federal Ministry of Digitalization and State Modernization (BMDS) BMDS needed to reduce long, complex planning and approval procedures for infrastructure projects that routinely exceeded legal timelines and burdened case workers with documentation-heavy manual workflows. more than six months beyond the legally defined limit 
Mistral Sample May 27, 2026 HTX HTX needed AI that understood Singapore's public-safety context, could be trained on sensitive data inside a secure environment, and would remain under government control rather than depending on generic off-the-shelf… 123B-parameter natively multimodal and multilingual foundation model AI Forge · AI Studio · NGINE 
OpenAI Sample Mar 21, 2025 Commonwealth of Pennsylvania Pennsylvania wanted to reduce time spent on drafting, research, summarization, IT support, procurement, contract analysis, and other administrative work across state agencies while keeping human oversight, transparency,… 175 employees from 14 agencies used ChatGPT Enterprise ChatGPT Enterprise 