Cutover plan

Dual-run, evidence, then the rest. Heroic cutovers are how recordings go missing.

1. Inventory

Write down numbers, queues, scripts or flows, skills, desktops, reports, recording paths, and the WFM feed. If it is not on the list, it will be rediscovered during the outage.

2. Map objects

Use the bridge table. A skill group, a CSQ, and a hunt pilot are not three names for one Connect queue. Say which Connect object each one becomes.

3. Build the instance

Region, identity, hours, queues, routing profiles, and flows whose error branches play a prompt. Logging on. One Lambda alias associated.

4. Pilot number

Claim or port one number. Keep the old platform able to take that traffic back. Porting uses the same Support path the quota page describes.

5. Dual-run

Move a small share. Compare one ContactId, one recording object in S3, one CTR, and one screen pop. Then compare them again the next morning.

6. Prove the agent network

From a real agent PC, not a jump host: TCP 443, UDP 3478, and an error rate under 2 percent across more than 200 calls.

7. Prove emergency calling

For US CCP agents, validate the address, store it, and read it back with 933. Do not place a live 911 call as a test.

8. Shift and watch

Increase traffic only after the pilot evidence exists. Keep the rollback number for a full business cycle.

9. Decommission

Turn off the old route only after recordings and CTR delivery have been seen for the contacts that moved.

Source: Routing profiles, used as the object map

Sources and go deeper