Skip to main content
NON-02 · Nonprofit & Social Impact

Grant Development Studio

Outcome​

Let the fundraising team move from a grant call to a submission-ready proposal faster by matching funder priorities against the organization's evidence base, reusing approved language and outcome data, and exposing every eligibility or budget condition that would disqualify the application.

Portfolio attributeValue
Primary personaFundraising team
Primary aggregateGrantOpportunity
Cosmos DB containers2
Grounding datagrant calls, evidence, budgets, outcomes, proposals
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
ProductivityProposal readiness time12-22% reduction(baseline median proposal-readiness minutes - assisted median minutes) / baseline median minutes * 100
Quality or riskEvidence discovery time20-35% reduction(baseline median evidence-search minutes - assisted median evidence-search minutes) / baseline median evidence-search minutes * 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: Grant requirements, organizational evidence, budgets, and prior materials can be assembled faster into compliant drafts.

Pilot guardrail: Use verified evidence and require named reviewers for proposal, budget, and organizational commitments.

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 proposal readiness time and evidence discovery time 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. Opportunity intake and requirement extractionGrant development studionotionOfFundingOpportunity → FunderRequirementChunkPolicy controlled
2. Eligibility and fit screeningGrant development studioFunderRequirementChunk → EligibilityFindingPolicy controlled
3. Go or no-go decisionDevelopment directorEligibilityFinding → goNoGoDecisionHuman required
4. Narrative and budget draftingGrant development studio with grant writerEvidenceAsset → NarrativeSectionPolicy controlled
5. Finance and executive approval before submissionFinance director and executive directorBudgetLine → submittedProposalHuman required

Technical architecture​

Technical architecture

Cosmos DB data model​

ContainerHierarchical partition keyDocumentsSearch paths
grantOpportunities/tenantId, /opportunityIdGrantOpportunity, FunderRequirementChunk, NarrativeSection, BudgetLine, EligibilityFinding/opportunityVectorEmbedding, /chunkVectorEmbedding, /sectionVectorEmbedding, /opportunityVector, /chunkVector, /sectionVector
outcomeEvidenceLibrary/tenantId, /programCodeEvidenceAsset/assetVectorEmbedding, /assetVector

Query contract​

QueryContainerScopePurpose
EliminatingRequirementsUnmetgrantOpportunitiessingle-partitionSurface funder conditions that would disqualify the application if unresolved.
VerifiedEvidenceRetrievaloutcomeEvidenceLibrarycross-partition-boundedMatch a narrative requirement to outcome evidence that is verified and inside its measurement currency window.
HighWeightScoringCriteriagrantOpportunitiessingle-partitionPrioritize drafting effort on the funder's most heavily scored criteria.
BudgetLinesWithRestrictionsgrantOpportunitiessingle-partitionList budget lines carrying unresolved allowability or indirect-rate findings before finance approval.

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/nonprofit-social-impact/grant-development-studio
az login
./Deploy.ps1 -Runtime python -Location eastus2

Change python to csharp to deploy the .NET implementation.

Explore the implementation​

Return to all accelerators