More than 100,000 Czech citizens each month now file benefit claims, check payment status, and ask questions of an AI assistant named Eva — all through a single cloud-based portal called Jenda. According to a Microsoft customer story published this month, the trio of Jenda, Eva, and a backend analytics stack built on Azure is processing a volume of welfare interactions that would have required armies of civil servants just a few years ago.

The numbers are eye-catching, but the real story is how a government agency moved from paper forms and manual transcription to a cloud-native, AI-augmented system in under three years — and the practical lessons that journey holds for any organization eyeing similar digital leaps.

From Paper Forms to a Cloud Trio in Under Three Years

The Czech Ministry of Labour and Social Affairs launched its first digital tool for welfare delivery in August 2022, when a temporary child allowance program pushed it to create a basic online application portal. That portal, which later became the Jenda Client Zone, now centralizes a far wider set of benefits: housing allowance, parental benefits, and — as of 2025 — a mobile app and forms related to unemployment and employment services.

Behind the public-facing portal sit two other components that transform it from a simple web form into an intelligent service:

  • Eva, a conversational AI assistant built with Azure OpenAI and orchestrated through Azure AI Foundry. Eva handles routine questions that used to flood call centers, guiding applicants through eligibility checks and form-filling steps. Microsoft says this has cut call volumes and freed staff for complex cases.
  • An analytics platform that ingests operational logs from Jenda, Eva, and physical branch offices into Azure Data Lake Storage and analyzes them with Azure Databricks. Dashboards show branch-by-branch performance, bottlenecks, and resource needs in near real time, allowing managers to reassign staff or adjust hours dynamically.

On the intake side, Azure AI Document Intelligence automatically extracts data from paper forms — still a reality for many applicants — and feeds structured information directly into the processing pipeline, eliminating manual rekeying that used to consume significant staff time.

The Ministry did not build these pieces in isolation. Rather, it stitched together purpose-built Azure services that are documented and commercially available: AI Foundry for agent orchestration, Document Intelligence for form parsing, Data Lake and Databricks for the data layer, and Microsoft Defender for Cloud for security. This is not a custom code behemoth but a considered assembly of cloud-native building blocks — a pattern Microsoft itself promotes as a "cloud-first" government approach.

What This Means for Citizens and IT Leaders

For the average Czech claimant, the shift is tangible. Applying for a housing allowance no longer demands a trip to a state office; the Jenda portal and mobile app accept applications, uploads of supporting documents, and proactive status notifications. The AI assistant Eva offers round-the-clock guidance, even if it remains deliberately constrained to informational support rather than definitive eligibility rulings.

But the impact extends well beyond citizen convenience. The project demonstrates how legacy-heavy public agencies can adopt modern cloud services to achieve measurable efficiency gains — and it surfaces hard-won lessons for IT leaders considering similar transformations.

For government IT shops
The Czech case underscores that cloud migration in the public sector is as much about procurement discipline and governance as it is about technology. An independent audit by the country’s Supreme Audit Office (NKÚ) previously flagged procedural shortcomings and contract irregularities in the Ministry’s digital projects, including aspects of Jenda’s development. Those findings are a warning: even technically sound solutions can falter if contracts, maintenance plans, and public accountability measures are not nailed down early.

For Windows and Azure practitioners
The technical choices illustrate a mature, well-architected pattern that many enterprise and mid-market teams can borrow. Using Azure AI Foundry to host and orchestrate a conversational agent, tapping prebuilt Document Intelligence models for form extraction, and building an analytics lakehouse with Data Lake and Databricks is a path that minimizes custom coding and accelerates time-to-value. However, practitioners should note that each service requires careful configuration: AI agents must be grounded in policy documents to reduce hallucinations; Document Intelligence needs human-in-the-loop review for exceptions; and the data platform necessitates role-based access controls to comply with data protection regulations like GDPR.

For citizens who depend on welfare services
The new system is designed to shrink the wait between application and payment. By automating data extraction and routing, the Ministry says it has accelerated benefit dispersal — a claim that is plausible given the volume of forms processed, though independent verification of exact time savings is not yet public. What is verifiable: the portal successfully handled spikes. During the initial child allowance push, the system processed applications from nearly 600,000 families, proving its capacity to withstand peak loads that would overwhelm paper-based alternatives.

How We Got Here: A Timeline of Incremental Buildout

The path from initial concept to the current tripartite platform reveals a step-by-step digitalization strategy:

  • August 2022: The Ministry launches a basic online portal for a one-off child allowance payment. It serves as a functional MVP, proving that citizens will adopt digital channels en masse.
  • 2023–2024: The portal expands into the Jenda Client Zone, adding housing and parental benefits. Eva the assistant is introduced, initially with a limited knowledge base. Azure AI Document Intelligence begins to replace manual transcription of paper forms.
  • 2025: A mobile app arrives, along with new forms for unemployment services. The analytics platform goes live, unifying data feeds and delivering dashboards. Microsoft publishes a customer story highlighting the 100,000-interactions-per-month milestone.

Throughout this period, the project weathered both praise and scrutiny. Local news outlets reported intermittent DDoS attacks that briefly disrupted Jenda, prompting the Ministry to harden its cloud defenses. And the NKÚ audit — while not focused exclusively on the portal — cast a shadow over procurement practices, reminding observers that public trust depends on transparency even when the technology works.

What to Do Now: Practical Steps for Similar Transformations

If your organization is considering a comparable leap, the Czech experience offers a checklist of do’s and don’ts:

  1. Start with a high-volume, low-risk workload. The Ministry’s initial one-off child allowance gave it a safe proving ground before expanding to ongoing, high-stakes benefits.
  2. Use cloud-native AI services, but tame the risks. Azure AI Foundry and Document Intelligence are powerful, but they require strict guardrails. Constrain the chatbot to answering from an approved knowledge base, and never let it make final eligibility decisions without human override.
  3. Plan your procurement and contracts for public scrutiny. Conduct open tenders, demand exit strategies and data portability from your cloud vendor, and publish technical rationales. The NKÚ’s criticisms show that even successful projects can be undermined by opaque contracting.
  4. Treat security as a continuous exercise. Government portals are prime targets for DDoS attacks. Run regular drills, employ layered defenses (like Microsoft Defender for Cloud plus a CDN/WAF), and maintain tested failover procedures.
  5. Pair citizen-facing tools with internal analytics. The analytics layer is what enables continuous improvement. Without it, you are running a static website, not a learning system.
  6. Communicate openly with the public. Publish non-sensitive performance metrics and error rates; give users a clear channel to report inaccuracies in the assistant’s responses. Trust is scarce currency in e-government.

What to Watch Next

The Czech system will likely continue evolving. If the Ministry follows through on its analytics foundation, we could soon see predictive models that forecast demand spikes or detect fraud patterns. The upcoming challenge is to maintain momentum while addressing the procurement and governance weaknesses that audits exposed. Other European governments are watching closely — the Jenda-and-Eva blueprint, with its mix of Azure services, could easily be replicated in nations that already use Microsoft cloud infrastructure. The test will be whether they also replicate the institutional discipline required to make it stick.

For everyday users and IT professionals alike, the story is a reminder that cloud transformation in government is no longer a futuristic concept. It is happening, with real throughput and real lessons, right now.