When someone goes quiet, the agent nudges them on the channel they were already using. Each channel has its own schedule and its own units, because a two-day gap is normal in email and rude on WhatsApp.
Async channels only. Website chat and live phone calls don't get follow-ups - a live conversation either continues or ends.
Email follow-ups use day offsets from the last contact. The default is day 2, day 5, day 10: a check-in, something useful, then a soft close.
WhatsApp has two layers because Meta's rules split at 24 hours from the customer's last message:
| Layer | Unit | What's sent |
|---|---|---|
| Inside the 24-hour window | Hours | Free-form messages the agent writes, like any other reply. Set offsets such as 2h, 6h, 12h. |
| Outside the window | Days | Only a Meta-approved template can be sent. Splice schedules the templates you configured. |
Hour offsets are ignored where they can't apply
Templates are synced from your connected WhatsApp Business Account - Splice never invents a template name. For each step you pick an approved template and its language, map each numbered placeholder to a lead or agent field (or fixed text), and optionally add a condition so the step only runs when it applies.
Free-form outside the window fails, then falls back
Instagram follows the same in-window model with hour offsets, but Meta provides no template equivalent. Once the window closes, the conversation waits for the customer to write again.
| Event | What happens |
|---|---|
| The customer replies | Pending steps are cancelled. A new sequence starts if needed. |
| The conversation reaches a terminal stage | Auto follow-up switches off by itself and the automation card says which stage stopped it. |
| A human takes over | The agent stops, including its follow-ups. |
| You pause follow-ups on that conversation | Nothing further is scheduled for that thread. |
| The agent is paused | Everything stops workspace-wide for that agent. |
A follow-up that ignores what was agreed on a call is worse than no follow-up. Where meeting context is available, a drafted follow-up is checked against the last meeting before it goes:
| Situation | Outcome |
|---|---|
| The draft contradicts a decision from the call | Escalated for human review - shown as 'Needs review - contradicts call.' |
| The rep owes the next step and the draft ignores it | Skipped, with 'Paused - waiting on rep next step.' |
| The agent owes the next step but the draft doesn't reference it | Held for context rather than sent. |
| A meeting happened but its transcript hasn't landed yet | Held, asking for call notes so it can be re-scored. |
| Nothing conflicts | Sends normally. |
The full logic is in Follow-up Sequences.
The automation card in the inbox side panel has an auto follow-up switch that affects that one conversation only. To change the schedule for everyone, edit the agent's follow-up settings. To stop everything at once, pause the agent.