Skip to main content
A customer asks, Eva answers, and the customer disappears. Inactivity follow-ups make Eva pick that conversation back up on its own, after the amount of time you decide.

What you will get

Your agent writes again to whoever stopped answering, with the number of attempts and the timings you define. Nobody on your team has to remember.

Requirements

  • An agent that is already configured. Follow-ups live in its behavior.
  • Access to that agent in your AI Agents list.
  • Publishing the changes at the end: nothing is active while it is still a draft.

Turn on follow-ups

1

Open the agent's behavior

In AI Agents, open your agent and go to the Behavior tab.
2

Find the section

At the end of the page, select Add section and choose Follow-ups & reactivation. If you already changed it, it has its own row on the page: select Open.
Add section menu with Follow-ups & reactivation marked Default.

Add section lists Follow-ups & reactivation while it still runs on its default.

3

Turn it on

Turn on the control to the right of Inactivity follow-up. The sequence and the instructions appear.
Inactivity follow-up and Allow targeted follow-ups, both switched off.

The two controls in the section; the top one turns on the inactivity sequence.

4

Tell it how it should sound

Under Instructions for Eva, describe the tone and the intent: what to remind them about, what to offer, when to close softly. Do not write the message: write the intent.
5

Build the sequence

Under Follow-up sequence, Step 1 defines how long to wait without a reply before the first message. + Add step adds the next ones.
6

Guide each step if needed

Step 1 instructions (optional) and its equivalents change the intent of that particular attempt. For example, making the last one a friendly goodbye.
7

Publish

Select Save behavior. Until that moment, nothing changes for your customers.

How many steps and how long

The sequence accepts between 1 and 5 steps. When it reaches five, + Add step disappears; the last step cannot be removed. Each step’s wait is counted from the previous message, not from the start of the conversation. It is written in hours or in days, and no step can wait more than 180 days. A three-step example: 2 hours, 1 day, 3 days. A customer who never replies gets three messages over a bit more than four days.
Follow-ups respect the agent’s business hours. If a step comes due in the middle of the night and your agent has hours configured, the message waits until it opens again.

Eva writes each message

There is no fixed text that repeats. At each step, Eva reads the whole conversation: what was discussed, what product was quoted, what was left pending. From that it writes a message for that customer at that moment. Your instructions are the guide, not the script. That is why “remind them of the design they asked for and ask whether they want to move ahead” is worth more than a phrase to copy and paste. If WhatsApp’s 24-hour window has already closed, the follow-up goes out with an approved template instead of a free-text message.

When the sequence stops

The sequence cuts itself off in three situations:
  • The customer replies. Any incoming message ends the inactivity period. The count restarts from the agent’s next reply.
  • Someone on your team takes the conversation. While a human is handling it, automatic follow-ups do not go out.
  • You turn off the control for Inactivity follow-up and publish.

The decision at the end of the sequence

When the last step goes unanswered, Eva takes one more turn to decide what happens to the opportunity: mark it lost, reschedule the follow-up, or leave it open. Eva makes the call; it is never an automatic close. How long to wait after the last step defines how much time passes before that decision. It comes at 48 hours and accepts up to 30 days. Instructions for the decision (optional) tells it what criteria to use.

Follow-ups Eva schedules on its own

Allow targeted follow-ups is a different thing, and it is turned on separately. With that tool Eva can schedule a reminder with its own objective when the conversation calls for it: picking up a quote, confirming a pending decision, checking in after a purchase. The inactivity sequence is triggered by silence. Targeted follow-ups are decided by Eva inside the conversation.

How to know it worked

  • The Unpublished changes indicator disappears after publishing.
  • In a test conversation: write to your agent, stop answering, and wait out the Step 1 time. In the Inbox you will see the follow-up arrive.
  • While a follow-up is pending, the conversation shows Next follow-up at the top. Opening it shows the Objective and the Date and time when Eva will write again.
  • On the opportunity linked to that conversation, the activity records Follow-up scheduled and later Follow-up due.
Next follow-up showing the conversation's objective and when Eva will write again.

The pending follow-up of a conversation, with its objective and its date.

If something goes wrong

Check the agent’s business hours: a follow-up that comes due outside those hours waits for opening. Also check whether someone on your team took that conversation.
That is usually because the conversation is being handled by a person. Eva does not interrupt a conversation your team already has in hand.
Leave a single step in the Follow-up sequence and remove the rest. The last one cannot be deleted; to switch everything off, turn off the control on the card.
The Instructions for Eva are too vague. Give it context about your business and a clear intent per step.

Next steps

Configure your agent's behavior

Reply instead of Eva