Relationship Manager Intelligence Copilot
Outcome
Prepare a relationship manager for every client conversation by merging holdings, transaction behaviour, CRM history, and approved research into a cited briefing, with suitability review before any recommendation is shared.
| Portfolio attribute | Value |
|---|---|
| Primary persona | Relationship manager |
| Primary aggregate | ClientRelationship |
| Cosmos DB containers | 2 |
| Grounding data | CRM, holdings, transactions, approved research |
| 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 | Client preparation time | 10-18% reduction | (baseline median preparation minutes - assisted median preparation minutes) / baseline median preparation minutes * 100 |
| Quality or risk | First-pass acceptance | 5-10 percentage-point improvement | assisted percentage accepted without material rework - baseline percentage accepted without material rework |
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: Relationship managers can prepare cited account narratives faster without attributing direct revenue or suitability decisions to the assistant.
Pilot guardrail: Keep advice within approved CRM, holdings, transaction, and research evidence; preserve suitability controls.
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 client preparation time and first-pass acceptance 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. Assemble relationship and holdings context | Briefing service | Client group id and upcoming meeting → Consolidated relationship snapshot | Policy controlled |
| 2. Detect portfolio drift and cash build-up | Analytics agent | Positions versus mandate targets → Drift and concentration findings | Policy controlled |
| 3. Ground ideas in approved research | Research retrieval agent | Findings and client objectives → NextBestActionCandidate set with citations | Policy controlled |
| 4. Run suitability and conflict screening | Suitability engine | Candidate actions and client risk profile → Screened candidates with blocking flags | Policy controlled |
| 5. Approve the client-facing briefing | Relationship manager and supervisor | Screened briefing pack → Released briefing and logged advice record | Human required |
Technical architecture

Cosmos DB data model
| Container | Hierarchical partition key | Documents | Search paths |
|---|---|---|---|
clientRelationships | /tenantId, /clientGroupId | ClientRelationship, PortfolioPosition, InteractionNote, NextBestActionCandidate | /summaryVectorEmbedding, /rationaleVectorEmbedding, /summaryVector, /rationaleVector |
approvedResearchLibrary | /tenantId, /researchId | ResearchNote | /bodyVectorEmbedding, /bodyVector |
Query contract
| Query | Container | Scope | Purpose |
|---|---|---|---|
| Relationship briefing assembly | clientRelationships | single-partition | Single partition read of relationship, positions, and recent interactions for the meeting pack. |
| Concentration exposure check | clientRelationships | single-partition | Identify positions breaching single-issuer or asset-class concentration limits. |
| Approved research retrieval | approvedResearchLibrary | cross-partition-bounded | Knowledge-container query against approvedResearchLibrary (bounded fan-out across the /tenantId + /researchId partitions named by the shortlist): restrict cited commentary to research that passed compliance approval, as a bounded fan-out over the research ids returned by vector shortlisting. |
| Pending suitability review queue | clientRelationships | cross-partition-bounded | Bounded cross-partition control-plane query: list proposed actions waiting on supervisor sign-off across the whole coverage book. It cannot carry the second partition key because a supervisor queue spans every client group the desk covers, so it runs as a tenant-scoped scheduled sweep off the interactive path, bounded by the pendingReview state and a 14-day createdUtc window, executed with MaxItemCount paging and continuation tokens. |
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/banking-capital-markets/relationship-manager-intelligence-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