Migrate from Zendesk to Support Oasis

The playbook

1. 1

   Day 1 to 3: Account setup and connections

   Provision the Support Oasis tenant. Connect your primary email channel and your help center. Add live chat to a staging surface. Wire Shopify or any CRM you rely on. Most teams complete this in a single sitting.

2. 2

   Day 4 to 7: Roles, SLAs, and routing

   Bring agents in with the right roles. Define SLAs per channel and priority. Set up routing rules. The defaults cover most teams, so this is usually a half-day exercise rather than a sprint.

3. 3

   Day 8 to 14: AI configuration and tone

   Turn on AI Copilot for drafts. Connect your help center as a source. Pick one ticket category for AI Agent autonomous resolution. Read the first 50 drafts. Tune confidence thresholds and tone until quality matches your bar.

4. 4

   Day 15 to 21: Run new tickets in parallel

   Route new tickets to Support Oasis while existing tickets continue resolving in Zendesk. Two weeks of parallel operation gives you confidence without risking active customer threads.

5. 5

   Day 22 to 28: Historical import

   Import historical conversations and customer records. Verify search and analytics line up. Most teams find this is the longest-running step but the lowest-risk one because the imported data is read-only.

6. 6

   Day 29 to 30: Full cutover

   Switch DNS and email forwarding so all new traffic goes to Support Oasis. Disable Zendesk seats once verified. Keep a Zendesk read-only seat for one cycle if you want belt-and-suspenders access to historical data.

Common pitfalls

• Underestimating the AI bill on Zendesk

  Most teams pricing the move discover their Zendesk AI usage was billed differently than expected. Pull three months of invoices first. Compare line by line, not headline price.

• Trying to migrate triggers and macros 1:1

  Zendesk triggers map cleanly to Support Oasis automations in spirit, not in shape. Rebuild from intent rather than translating mechanically. The result is usually simpler.

• Skipping the parallel run

  The two-week parallel period seems redundant. It is not. It catches edge cases that a tabletop walkthrough misses, with zero risk to live customers.