Skip to main content
GOV-01 · Government & Public Sector

Citizen Services Navigator

Outcome​

Help a resident understand which public benefit or licensing programs they qualify for, what evidence each program requires, and where their submitted application currently stands, using only published program rules and the resident's own case record.

Portfolio attributeValue
Primary personaCitizen
Primary aggregateResidentServiceCase
Cosmos DB containers2
Grounding dataprogram rules, forms, services, case status
Development estimateUSD 60-220 monthly dev estimate
Runtime optionsPython FastAPI and C# ASP.NET Core

Productivity hypothesis​

These measures reflect the workflow, evidence, decision boundary, and industry risk for this use case.

Impact typeMeasureConservative internal assumptionCustomer measurement
ProductivityIntake-to-response time8-14% reduction(baseline median intake-to-response minutes - assisted median minutes) / baseline median minutes * 100
Quality or riskCase rework3-6% reduction(baseline reopened or corrected case rate - assisted rate) / baseline rate * 100

Important: Conservative internal pilot assumptions only; not public benchmarks, commitments, or expected customer results. Validate against a comparable baseline or matched control cohort with mandatory human review.

Rationale: Program rules, documents, case status, and service information can support faster and clearer citizen navigation.

Pilot guardrail: Caseworkers retain eligibility authority; preserve accessibility, language support, and evidence tracking.

Customer validation plan​

  1. Baseline the current workflow with representative users and tasks.
  2. Run the accelerator on the same task mix with human review retained.
  3. Measure intake-to-response time and case rework with the documented formulas.
  4. Replace internal assumptions with customer-observed results before a business case.

Why Azure Cosmos DB​

This AI solution needs an operational data plane, not only a vector index:

  • Hybrid vector and full-text retrieval combines semantic recall with exact identifiers, codes, names, and domain language.
  • Hierarchical partition keys and metadata predicates keep tenant and domain boundaries in the normal query path.
  • Operational JSON, source chunks, workflow and agent state, feedback, and audit records remain together for low-latency reads and governed writes.
  • Change feed reacts to source and operational updates without polling.
  • Autoscale throughput, configurable consistency, and multi-region distribution support variable demand and enterprise availability.

Domain workflow​

Domain workflow

StageActorInput → outputApproval
1. Plain-language needs intakeResident with citizen services navigatorresidentQuestion → ServiceCasePolicy controlled
2. Program rule retrievalRetrieval orchestratorServiceCase → EligibilityRuleChunkPolicy controlled
3. Advisory eligibility screeningCitizen services navigatorEligibilityRuleChunk → EligibilityAssessmentPolicy controlled
4. Evidence collection and verificationResident and intake clerkEligibilityAssessment → EvidenceItemHuman required
5. Caseworker determination and noticeEligibility caseworkerEvidenceItem → determinationNoticeHuman required

Technical architecture​

Technical architecture

Cosmos DB data model​

ContainerHierarchical partition keyDocumentsSearch paths
residentServiceCases/tenantId, /caseIdServiceCase, EvidenceItem, EligibilityAssessment, CaseNote/rationaleVectorEmbedding, /rationaleVector
benefitPolicyLibrary/tenantId, /programCodeEligibilityRuleChunk/chunkVectorEmbedding, /chunkVector

Query contract​

QueryContainerScopePurpose
ResidentCaseStatusresidentServiceCasessingle-partitionReturn the resident's current case state and next required action for the status page.
OutstandingEvidenceChecklistresidentServiceCasessingle-partitionList documents the resident still owes before determination can proceed.
EffectiveProgramRuleRetrievalbenefitPolicyLibrarysingle-partitionRetrieve only the program rule text in force on the application date, in plain language where available.
CaseInteractionHistoryresidentServiceCasessingle-partitionReconstruct the full cited interaction history for an appeal or a public records request.

Python and C# services route each document type to its declared container and hierarchical partition key. ETag-based writes protect mutable aggregates, while append-only records and bounded cross-partition queries preserve scale.

Azure deployment​

Azure deployment

Clone the source repository, authenticate with Azure, and run:

Set-Location accelerators/government-public-sector/citizen-services-navigator
az login
./Deploy.ps1 -Runtime python -Location eastus2

Change python to csharp to deploy the .NET implementation.

Explore the implementation​

Return to all accelerators