Skip to main content
TEC-02 · Technology & Software

Customer Success Intelligence

Outcome​

Give the customer success manager a defensible account health picture that ties product usage against entitlement, support burden, contract commitments and stakeholder sentiment into a specific renewal or expansion play grounded in the account's own evidence.

Portfolio attributeValue
Primary personaCustomer success manager
Primary aggregateCustomerAccountHealth
Cosmos DB containers1
Grounding dataCRM, usage, support, contracts, calls
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
ProductivityAccount evidence discovery time15-25% reduction(baseline median account-evidence search minutes - assisted median minutes) / baseline median minutes * 100
Quality or riskEvidence-complete risk and action narratives5-10 percentage-point improvementassisted percentage meeting the account-evidence completeness rubric - baseline percentage

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: CRM, usage, support, contracts, and calls support faster preparation of grounded customer-success narratives.

Pilot guardrail: Humans approve renewal, escalation, commercial, and customer-commitment actions.

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 account evidence discovery time and evidence-complete risk and action narratives 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. Signal consolidation across systemsAccount data connectorsUsageFact → consolidatedSignalSetPolicy controlled
2. Health scoring with driver attributionCustomer success intelligenceconsolidatedSignalSet → healthScoreWithDriversPolicy controlled
3. Commitment and objection retrievalRetrieval orchestratorInteractionChunk → evidenceBundlePolicy controlled
4. Play recommendationCustomer success intelligenceevidenceBundle → AccountPlayPolicy controlled
5. CSM and manager commitment to the playCustomer success manager and CS leaderAccountPlay → committedPlayHuman required

Technical architecture​

Technical architecture

Cosmos DB data model​

ContainerHierarchical partition keyDocumentsSearch paths
customerAccountHealths/tenantId, /accountIdAccount, UsageFact, SupportSignal, StakeholderRecord, InteractionChunk, AccountPlay/chunkVectorEmbedding, /rationaleVectorEmbedding, /chunkVector, /rationaleVector

Query contract​

QueryContainerScopePurpose
RenewalsAtRiskInWindowcustomerAccountHealthscross-partition-boundedList accounts renewing inside the horizon whose health score sits in a risk band. Intentional cross-partition book-of-business view: bounded by the tenantId prefix, one CSM's assignment, a renewal horizon and the elevated churn-risk bands.
EntitlementUtilizationGapcustomerAccountHealthssingle-partitionExpose features where consumption is far under or over entitlement, driving recovery or expansion.
ChampionDepartureRiskcustomerAccountHealthssingle-partitionSurface influential champions with no meaningful engagement inside their expected cadence.
CommitmentEvidenceRetrievalcustomerAccountHealthssingle-partitionRetrieve the call or contract passage where a success criterion or commitment was made.

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/technology-software/customer-success-intelligence
az login
./Deploy.ps1 -Runtime python -Location eastus2

Change python to csharp to deploy the .NET implementation.

Explore the implementation​

Return to all accelerators