Local authoritiesAzure · Fabric · Purview · AI Foundry

A data and AI platform for local authorities: one source of truth for every resident

Zooz-tech designs and builds enterprise data platforms on Microsoft for local authorities (municipalities): one record for every resident, business and property, reports that update without manual refresh, a governed knowledge layer and AI agents for staff and residents, with data governance aligned with Amendment 13 to Israel's Privacy Protection Law. The environment, the code and the knowledge remain the property of the authority.

Source systems of a typical local authority
CRM and contact centerBilling and property taxSocial servicesEducationEngineeringBusiness licensingGISAppointment scheduling
Golden record: resident · business · property
BI reports
Knowledge and AI
APIs for channels
The challenge

Every department sees a different resident

In a local authority, data is spread across separate systems: the contact center, billing, social services, education, engineering, business licensing, GIS and scheduling. There is no single source of truth for residents, businesses and properties, management reports depend on manual refreshes, and AI use consists of point tools with no common governance.

One platform solves all three together, and prepares the authority for governed use of artificial intelligence, for both residents and staff.

Target architecture

Nine layers, all on Microsoft

Each layer has a clear role, well-known tools and is owned by the authority. For the data layer, we analyze the alternatives, Microsoft Fabric versus Azure Native, including a five-year TCO, and recommend based on the authority's needs.

Infrastructure and identityA secure cloud environment, environment separation, identities and permissionsAzure Landing Zone, Entra ID, Entra Agent ID, Key Vault, Defender for Cloud
Data ingestionDaily extraction and near-real-time updates from source systemsFabric Data Factory / ADF, Azure Functions, Service Bus, Event Grid
Storage and layersBronze, Silver and Gold: raw data, cleansed data and business data martsMicrosoft Fabric: OneLake, Lakehouse, Warehouse
Golden record (MDM)One record per resident, business and property, with cross-reference IDs for every systemAzure SQL, Spark / Functions, Power Apps
BI and reportingDashboards for management and departments, with no manual refreshPower BI (Direct Lake)
Knowledge and searchA governed index of procedures and documents, with a personal data check before ingestionAzure AI Search, Document Intelligence, Azure AI Language
AI agentsAgents for staff, and the foundation for resident service agents by text and voiceCopilot Studio, Azure AI Foundry, Azure OpenAI, Voice Live, Content Safety
APIs and accessA single gateway for exposing data to channels and systemsAzure API Management
Data governanceCatalog, classification, lineage and safe use of AIMicrosoft Purview, DSPM for AI
Our design decisions

Six decisions that make the difference

Data at rest in Israel, AI processing in Europe

Some AI services are not available in the Israel Central region. Data at rest stays in Israel, and processing runs in a European region, with no cross-region transfer.

One governed knowledge index

Instead of each tool's built-in indexing: a single index in Azure AI Search, with permission filtering and personal data checks, consumed by every agent.

Separate agents for staff and residents

Copilot Studio for internal processes. Azure AI Foundry for resident service, with custom authentication, a custom interface and a voice channel.

A golden record without a third-party engine

National ID numbers, company registration numbers and property numbers allow rule-based matching on Microsoft services, with no license for a commercial MDM engine and full authority ownership of the code and the rules.

Resilience to outages

If the AI region is unavailable, channels automatically switch to a degraded mode that runs from Israel: forms, callbacks and SMS.

Authority independence

Every component lives in the authority's own subscription, as Infrastructure as Code in its own Git, including prompts, configurations and indexes.

What the authority receives

From requirements to a working environment you own

Zooz-tech supports the authority end to end: architecture, technology and regulatory decisions, a work plan, a cost model and implementation.

1

A target architecture document and flow diagrams for every layer.

2

An analysis of alternatives for the data layer, Microsoft Fabric versus Azure Native, including a five-year TCO.

3

A data flow map and data type register for the Privacy Protection Officer: what is sent, what is stored, where and for how long.

4

A phased work plan, an effort estimate in hours and a timeline.

5

A licensing and consumption estimate for Microsoft products, with procurement staged by project phase.

6

Golden record design: matching, survivorship, the data steward process and ongoing operations.

7

The environment itself: code, configurations, indexes, documentation and operating procedures, owned by the authority.

Frequently asked questions

What is a golden record?

A golden record is a single, unified record for each resident, business and property, built from all source systems, with cross-reference IDs for every system and rules that decide which value wins when sources conflict.

Where is resident data stored?

Data at rest is stored in an Azure region in Israel. AI processing runs in a European region, because some services are not available in Israel, and the data flow map is submitted to the Privacy Protection Officer.

Microsoft Fabric or Azure Native?

Both options are viable. We analyze them against the authority's needs, including five-year cost, and the other layers (APIM, Purview, AI Search, the agents) stay the same in both.

How does this comply with Amendment 13?

Through Purview for cataloging and classification, DSPM for AI, keeping personal data out of the index, permissions through Entra, and a complete data flow map for every component.

Is your authority planning a data platform?

We would be glad to present the architecture and the path to the first phase.

Book a call