CRM Frontier Firm Showcase — a reference implementation for a modern CRM built with human-agent teams on top of Dynamics 365 / Dataverse / Power Platform / Copilot Studio / Azure AI Foundry. The illustrated vertical is insurance; the architecture patterns generalise to any regulated B2C/B2B CRM.
The question that shapes a CRM programme running past 2038 is not "can this platform satisfy today's requirements?" but "can it enable the requirements that do not exist yet?" This repository is the engineering surface behind that answer: the design decisions, the data model, the integration contracts, the agent definitions and the pipeline that produce the Dynamics 365 solution the customer sees running in the sandbox.
GitHub-native, agent-driven delivery with end-to-end traceability from requirement → ADR → solution change → test → evidence is the delivery model we propose and the thing we demonstrate live.
| Mode | What it is | Where it lives |
|---|---|---|
| RUN | The working solution, live | Sandbox environment |
| DESIGN | Decisions, models, contracts | docs/, docs/adr/ |
| BUILD | Agents producing a real change | .github/agents/, solution/ |
RUN always comes first in any demonstration. DESIGN and BUILD are the answer to "and how did it get there — and what happens at the next release?"
One customer, one change, cascading — see docs/ideas/UC-01-relocation-across-jurisdictions/.
The Smith household relocates across a jurisdiction boundary. One address change must re-rate motor, contents and natural-hazard cover; withdraw and re-originate building cover because the jurisdiction changed; reassign the household to a different General Agent; and unwind the multi-product discount on the remaining policies.
Every architectural conversation is validated against this one thread.
Every contribution — human or Copilot — follows the same rules:
- SUPERPOWERS_CONTRACT.md — binding operating contract.
- AGENTS.md — 6 runtime agents (
AG-F-##) + 10 engineering agents (AG-E-##). - CONTRIBUTING.md — ground rules and workflow.
- .github/copilot-instructions.md — repo-wide Copilot rules.
- .github/instructions/superpowers.instructions.md — path-scoped rules loaded automatically.
- .github/agents/ — one Copilot custom agent per engineering role.
- .github/chatmodes/ — the same roles as chat modes.
Start here:
- docs/DESIGN-PRINCIPLES.md — read before any design decision. Each principle has a failure test.
- docs/PRD.md — product intent, use cases, non-goals.
- docs/SD.md — solution design and target architecture.
- docs/BACKLOG.md — working backlog by epic.
- docs/COPILOT-BUILD-GUIDE.md — how PRs move through this repo.
Full index: docs/README.md.
README.md AGENTS.md SUPERPOWERS_CONTRACT.md CONTRIBUTING.md
.github/
copilot-instructions.md
agents/*.agent.md 10 engineering agents (AG-E-01..10)
chatmodes/*.chatmode.md matching chatmodes
instructions/ path-scoped rules
prompts/ reusable prompts
workflows/ ci-solution · cd-infra · cd-solution-dev/test
ISSUE_TEMPLATE/ governance escalation etc.
pull_request_template.md
docs/
SD · DATA · INTEGRATION · EXTENSIBILITY · PROCESSES · AI · ANALYTICS
OPERATIONS · SHARED-RESPONSIBILITY · LICENSING · COMPLIANCE · SECURITY
DESIGN-PRINCIPLES · PERSONAS-JOURNEY · TEST · BACKLOG · COPILOT-BUILD-GUIDE
PRD · GLOSSARY · ENVIRONMENTS
adr/ architecture decision records (0001..0018)
ideas · specs · plans · sprints · reviews
api/ integration contracts (OpenAPI, event schemas)
data/scenarios/ the golden-thread fixture (synthetic only)
solution/ unmanaged Dataverse solution, source-controlled
infra/terraform/ IaC for tenant, environments, Entra, GitHub
Every capability section carries one of:
✅in the offer➕additional licence required🧩configuration / own build — no additional licence🗺️roadmap — not yet productive
The consolidated view is docs/LICENSING.md. An unflagged capability is a finding.
- Public — anonymised demo. No real customer identity, policy or claims data. The illustrated customer is fictional (Contoso Insurance); the illustrated household is fictional (the Smith household).
- All commercial and licensing statements are indicative in a demo context; in a real engagement they must be confirmed against the offer before being asserted.
- Not legal advice — data-protection and regulatory positions in docs/COMPLIANCE.md are illustrative; the customer's legal function validates them in a real programme.