# Utility AMI Modernisation — Enterprise Architecture Case

## Context
A regional utility needed to replace brittle point-to-point meter integrations connecting its head-end system, MDMS, SAP IS-U, outage management, data platform, and customer channels.

## Architecture challenge
- 2.4 million meters producing interval, event, and command traffic
- Duplicate reads and delayed billing caused by inconsistent contracts
- Tight coupling between operational technology and enterprise systems
- Limited replay, traceability, and service-level visibility

## Target architecture
The target introduces an event backbone between AMI/MDMS and downstream consumers. Canonical, versioned events separate producers from billing, outage, analytics, and customer-service workloads. An API layer handles commands and synchronous queries. A replay store supports controlled recovery without republishing from operational source systems.

## Key decisions
1. Kafka topics are partitioned by meter identifier to preserve per-meter ordering.
2. Avro contracts use backward-compatible evolution enforced in CI/CD.
3. Each event carries event ID, correlation ID, source timestamp, schema version, and idempotency key.
4. OT traffic crosses a controlled integration zone; consumers never connect directly to the head-end system.
5. Dead-letter topics are operational queues with ownership, alerts, and replay procedures.

## Transition roadmap
- **Wave 1 — Foundation:** platform, schema registry, observability, security controls
- **Wave 2 — Meter reads:** interval and register reads to MDMS and analytics
- **Wave 3 — Events:** outage/tamper events to OMS and customer operations
- **Wave 4 — Commands:** governed remote-connect and service-command APIs
- **Wave 5 — Retirement:** remove point-to-point interfaces after reconciliation

## Expected measures
- End-to-end event latency and successful delivery rate
- Duplicate and invalid event rate
- Mean time to detect and recover failed messages
- Interfaces retired and consumers migrated
- Billing exceptions attributable to integration quality

*This is a representative, anonymised EAOT reference case. Figures and system names illustrate a realistic enterprise scenario rather than a named client engagement.*