Pre-Launch Automation Readiness Checklist (Onboarding)

Pre-Launch Automation Readiness Checklist (Onboarding)

1. 01

   Map every data field the automation will read or write before connecting any tool

   Document source system, field owner, and update frequency for each field. Agencies that skip this step discover mid-campaign that a CRM field is populated by a different team on a different cadence, which breaks personalization tokens silently.

2. 02

   Confirm the client's sending domains and mailboxes are provisioned and authenticated

   SPF, DKIM, and DMARC records must resolve before the first sequence fires. A single unauthenticated sending domain can drag down reply rates across the entire client portfolio sharing that infrastructure.

3. 03

   Define the human review checkpoint for every AI-generated message before it leaves the queue

   Set a threshold: first 50 sends per sequence get manual approval, then sample 10 percent thereafter. This catches tone drift and factual errors that automated QA rules miss.

4. 04

   Test idempotency on every trigger that writes to the CRM or billing system

   Fire each trigger twice in a staging environment and verify no duplicate records appear. Unguarded retries create silent billing and CRM corruption that surfaces weeks later during client reconciliation.

5. 05

   Document the suppression list and opt-out handling path end to end

   Trace how a prospect's unsubscribe request flows from the sending tool back to the master CRM. If the path has a gap, the client risks contacting someone who already opted out, which is a compliance exposure.

6. 06

   Set volume ceilings per mailbox per day and per week before launch

   Start at 20 to 30 sends per mailbox per day for new domains and scale by 10 percent weekly only if bounce rates stay under 2 percent. Volume without a ceiling is the fastest route to a blocked domain.

7. 07

   Assign a named owner for monitoring the first 14 days of live sends

   The owner reviews bounce rates, reply sentiment, and any agent errors daily. Without a named person, anomalies sit unreported until the client notices a drop in pipeline.