Many organisations still run a system that works — just not with anything modern. Rewriting it can take years. Bridging it with a well-designed API layer often unlocks value in weeks.
Start with ownership and failure modes
Before writing connectors, document:
- Which system owns each field
- What “in sync” means for each entity
- How you detect and recover from partial failures
- Who gets alerted when a sync stalls
Prefer contracts over cleverness
Stable schemas, idempotent writes, retries with backoff, and audit logs beat one-off scripts every time. Treat the bridge as a product: versioned, monitored, and owned.
Keep humans in the loop where it matters
Not every record should sync automatically. Validation rules and exception queues protect billing, compliance, and customer-facing data while still eliminating routine double-entry.
Measure the win
Track re-keyed records, sync latency, and mismatch rate before and after. Those numbers make the business case for the next integration — and prove the bridge was worth building.