· By the AgentSDR team
How do you automate sales follow-ups?
Write the follow-up steps once, set the wait between them, and let the software schedule each message and stop the moment the lead replies, asks to stop or books. Decide separately whether each message sends by itself or waits for a person to approve it.
This guide covers the decisions behind that setup: how many steps, how much space, how to mix channels, what ends a sequence and what happens after it. We use AgentSDR’s rules as the concrete example and give no reply-rate statistics, because the numbers that circulate online come from different markets and lists and will not predict yours. Measure your own.
Which kind of follow-up are you automating?
There are two, and they need different handling.
| Kind | Campaign follow-up | CRM follow-up |
|---|---|---|
| Goes to | Someone who never replied | Someone who replied, then went quiet |
| Starts when | The first message goes unanswered | You answer their reply |
| Written from | A template with merge fields | The conversation so far |
| Typical goal | Earn a first reply | Move to a meeting or the next stage |
A campaign follow-up can be a short template. A CRM follow-up that ignores what the person said reads as a robot. The second kind is where AI drafting helps most, because the message depends on the thread.
How many follow-ups should you send?
Pick a number you can defend, and let your data move it. A few rules of thumb hold without statistics:
- Start short. Three to five touches after the first message is a common shape for cold outreach. Begin with two or three for CRM follow-ups and add more once you see what works.
- Give each step a new reason to write. A new angle, a question, a case study or a small useful link beats “just checking in” every time.
- Read the evidence per step. If step five adds almost no replies and some unsubscribes, remove it.
The limit in the tool is rarely the constraint. AgentSDR allows up to 30 steps in a CRM sequence, which exists so long, slow cadences are possible. It is not a target.
How far apart should follow-ups be?
Space them wider as the sequence goes on. A common pattern is a couple of days between the first touches, then a week, then longer. Follow-ups a week apart usually read better than daily ones, and a hot lead and a cold one deserve different rhythms.
Spacing is only half of timing. Where a message lands in the week matters too. Sending hours and gaps keep automation from looking like automation:
- Send inside the recipient’s working hours, in a time zone you set.
- Keep a random gap between sends from the same account. For email in AgentSDR the default is 18 to 24 minutes per mailbox, and a mailbox sends only inside its own hours. See the Google Workspace cold email guide.
- Respect a daily limit per mailbox or account. When a message comes due outside the hours or over the limit, it waits for the next open slot.
- For CRM follow-ups, AgentSDR adds its own gap between follow-ups from the same sending account, 30 minutes by default and set in the organization’s Sending rules.
When should automation stop?
Most follow-up damage comes from messages that went out after they should have stopped. Make the stop rules explicit, and check your tool applies them without a person remembering.
| Situation | What the automation should do |
|---|---|
| Did they reply on any channel? | Stop the sequence and handle the reply |
| Did they ask to stop or unsubscribe? | Mark Do Not Contact on every channel |
| Did the email bounce? | Stop and fix or remove the address |
| Did they book a meeting? | Move to the meeting stage and switch to meeting reminders |
| Is it an out-of-office reply? | Pause, resume on the return date |
| Is it outside sending hours or over the daily limit? | Wait for the next open slot |
| Have the steps run out? | Move to a slow loop, or close the record |
A reply on one channel should stop pending steps on every channel. Someone who answered your LinkedIn message and then gets a cold email three days later has every reason to be annoyed. In AgentSDR a reply stops that person’s pending email, LinkedIn and WhatsApp campaign steps and cancels pending CRM follow-ups, and a new reply cancels unsent drafts that the conversation has outrun.
Should follow-ups use more than one channel?
Often, yes, with care. A person who ignores your emails may answer a short LinkedIn message, and the reverse. A multichannel follow-up works when each channel has its own limits and the lead is treated as one person across them. It fails when three tools each chase the same lead without knowing about the others.
- Keep one record per person across channels, so a reply anywhere is visible everywhere. That is the idea behind an omnichannel CRM.
- Follow the channel the conversation is on. In AgentSDR a CRM sequence is not tied to one channel, and the lead’s conversation decides where the message goes.
- Keep each channel inside its own guardrails: email sending hours and daily limits, LinkedIn daily limits per account, and WhatsApp rules such as no new chats for 24 hours after linking a number and a gap between sends.
- Never contact someone marked Do Not Contact on any channel.
What happens after the last follow-up?
A sequence that ends in silence leaves a record idle, and idle records are forgotten. Two choices fit: close the record, or hand it to a slow loop.
AgentSDR’s answer is the “never idle” rule. A sequence can name another sequence to follow when its steps run out, and the record continues with that sequence from its first follow-up. A sequence may name itself. The Long-term loop in the default follow-up template does this: one useful touch every 21 days until the person replies or says stop. It does not apply to Do Not Contact, closed or paused records.
A slow loop only works if each touch is useful. A product update, a relevant case study or a new question is a reason to write. A monthly “any update?” is not, and the loop should end if it keeps going unanswered. Read the loop’s wording before you switch it on.
How should you handle out-of-office replies?
An out-of-office reply is not a reply from a person. It needs no answer, and it should not end the sequence. It should not clog the queue either.
In AgentSDR, ingest first stops the person’s outreach when a reply arrives, and writes down what it stopped. When the classifier reads the reply as out of office, that outreach is put back: the next message goes out when it was due, or on the return day from the notice if that is later. No draft is written, and the record never reaches the Action required queue. If an earlier reply from the person is still unanswered, the normal reply flow runs instead.
How does AgentSDR automate CRM follow-ups?
The rules, in the order they apply:
- A lead replies and the AI classifies the reply. The stage’s sequence drafts the immediate reply.
- A person sends it, the AI draft or their own. Sending the reply, or confirming the stage afterwards, starts the stage’s follow-ups. A reply typed outside AgentSDR does not.
- Each follow-up is drafted when it comes due, from the conversation and the step’s goal. A sequence can have up to 30 steps, each with a wait in minutes, hours or days.
- By default each draft waits in Action required for a person. A sequence can be set to send follow-ups automatically. That setting is off by default.
- Automatic sending takes at most one draft per sending account at a time, inside that account’s hours and the follow-up gap. A draft that cannot be sent within 48 hours, or that a person takes over, goes to Action required.
- The immediate reply can also be sent automatically after a 15-minute hold, as a separate setting, also off by default.
- When the steps run out, a “then” sequence can take over, such as the 21-day Long-term loop.
For the wider system, read automated CRM, AI email replies and the AI CRM page. If you are comparing tools, see the best AI CRM roundup, and for the basics of the category, what an AI CRM is.


