Sales Execution Copilot
Outcome
Prepare the field sales representative for each retail account call with a grounded pre-call plan: what the account bought and stopped buying, which planogram gaps and promotion compliance issues exist, and what the trade terms permit them to offer, all cited to the account record and the trade playbook.
| Portfolio attribute | Value |
|---|---|
| Primary persona | Field sales representative |
| Primary aggregate | RetailAccountCallPlan |
| Cosmos DB containers | 2 |
| Grounding data | retailer history, assortment, promotions, playbooks |
| Development estimate | USD 60-220 monthly dev estimate |
| Runtime options | Python FastAPI and C# ASP.NET Core |
Productivity hypothesis
These measures reflect the workflow, evidence, decision boundary, and industry risk for this use case.
| Impact type | Measure | Conservative internal assumption | Customer measurement |
|---|---|---|---|
| Productivity | Account preparation time | 10-20% reduction | (baseline median preparation minutes - assisted median preparation minutes) / baseline median preparation minutes * 100 |
| Quality or risk | Evidence discovery time | 15-25% 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: Field-sales teams can prepare accounts faster from retailer history, assortment, promotions, and execution guidance.
Pilot guardrail: Keep proposed actions within approved retailer policy, trade terms, promotions, and playbooks.
Customer validation plan
- Baseline the current workflow with representative users and tasks.
- Run the accelerator on the same task mix with human review retained.
- Measure account preparation time and evidence discovery time with the documented formulas.
- 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

| Stage | Actor | Input → output | Approval |
|---|---|---|---|
| 1. Account performance analysis | Sales execution copilot | OrderHistoryFact → performanceGapSet | Policy controlled |
| 2. Distribution void and compliance detection | Assortment analyzer | AssortmentPosition → voidAndComplianceFindings | Policy controlled |
| 3. Trade terms grounding | Retrieval orchestrator | voidAndComplianceFindings → TradePlaybookChunk | Policy controlled |
| 4. Pre-call plan generation | Sales execution copilot | TradePlaybookChunk → CallPlan | Policy controlled |
| 5. Representative call execution and commitment capture | Field sales representative | CallPlan → CallOutcome | Human required |
Technical architecture

Cosmos DB data model
| Container | Hierarchical partition key | Documents | Search paths |
|---|---|---|---|
retailAccountCallPlans | /tenantId, /accountId | RetailAccount, AssortmentPosition, OrderHistoryFact, CallPlan, CallOutcome | /planVectorEmbedding, /planVector |
tradePlaybookLibrary | /tenantId, /termsTier | TradePlaybookChunk | /chunkVectorEmbedding, /chunkVector |
Query contract
| Query | Container | Scope | Purpose |
|---|---|---|---|
| DistributionVoidsForAccount | retailAccountCallPlans | single-partition | List authorized SKUs that are not on shelf so the representative can open the call with a specific gap. |
| DecliningSkuTrend | retailAccountCallPlans | single-partition | Find SKUs whose ordered cases fell materially across the trailing periods for this account. |
| PermittedOfferRetrieval | tradePlaybookLibrary | single-partition | Retrieve only the trade terms and promotion rules that apply to this account's terms tier and the topic at hand. |
| OpenFollowUpsByTerritory | retailAccountCallPlans | cross-partition-bounded | Show commitments and follow-ups owed across the representative's route. Intentional cross-partition rep worklist: bounded by the tenantId prefix, one representative and a due-date horizon. |
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

Clone the source repository, authenticate with Azure, and run:
Set-Location accelerators/consumer-packaged-goods/sales-execution-copilot
az login
./Deploy.ps1 -Runtime python -Location eastus2
Change python to csharp to deploy the .NET implementation.
Explore the implementation
- Launch signed-in demo
- Implementation article
- Domain data model
- Workflow and API contract
- Python application
- C# application
- Domain container Bicep
- Platform Bicep
- Cost and configuration
- Source repository