I’ve set up LiSeller’s automated follow-up sequence and it’s working technically—the messages are going out on schedule—but I’m getting this nagging feeling that they sound robotic, even though they’re supposed to be personalized.
Here’s what I’m doing: I have a 4-message sequence with AI personalization enabled. The first message has a customized hook based on the prospect’s company/industry. The second and third messages are more generic (which I know is part of the problem), and the fourth is a soft breakup message.
The issue is that the second and third messages feel like they’re following a template formula. Like, the AI inserts a different company name or detail, but the structure is identical, so any prospect who got message 1 and then message 2 can obviously tell the second one is automated.
I’ve also noticed that my follow-ups don’t really reference what happened in the previous message. They feel like standalone touches rather than part of an actual conversation. So a prospect might reply to my first message, and then my automated follow-up goes out treating it like a fresh restart.
I want my automation to feel natural, like I’m actually continuing a conversation, not just firing off pre-written templates. How are you all handling this? Should I be manually managing some follow-ups instead of automating them, or is there a way to make automation feel more natural?
The structure is the problem. Most automated sequences follow the same three-act structure: (1) hook, (2) value pitch, (3) CTA. Prospects recognize this pattern immediately, and it signals automation.
I changed my sequence to break the pattern. My messages now follow different structures:
- Message 1: Hook + one question
- Message 2: Data point or social proof (totally different format)
- Message 3: Case study or reverse ask (“I actually see in your industry”)
- Message 4: Soft exit + calendar link
Each message feels different structurally, so even though they’re technically automated, the variation masks the automation. Prospects don’t feel like they’re in a robotic sequence; it feels like I’m sending different types of messages over time.
Vary your structure, not just your details.
You need conditional logic in your sequence. If a prospect replies to message 1, don’t send message 2 at all—send a different message acknowledging their reply. LiSeller supports conditional sequences where you can say: “If replied, send sequence B. If no reply, send sequence C.”
I built this logic into all my sequences and it completely changed the feel. Now, people who engage early get a conversation; people who don’t get a standard follow-up. It feels way more natural because it actually responds to prospect behavior instead of ignoring it.
This is a technical fix that takes 20 minutes to set up and changes everything.
Also be careful with message frequency. If you’re sending 4 messages too quickly (like 2 days apart), they feel like spam even if they’re personalized. Space your follow-ups out: Day 0, Day 3-4, Day 7-8, Day 14+. The longer gaps make automation feel less aggressive and more like natural touchpoints.
Plus, LinkedIn’s algorithm punishes repetitive messaging from the same account. Longer gaps keep you safer.
In recruiting, I’ve learned that the second message is where most sequences fail because people default to a generic pitch. Instead, I make message 2 about them—new research I found, a news item about their company, or a genuine question based on their profile.
Message 2 should never be about your product. It should prove you’re still thinking about them specifically. That’s what makes automation feel human—when the AI is clearly finding new reasons to reach out, not just cycling through a pre-written pitch.
Real talk: I went hybrid. Message 1 and 2 are fully automated and AI-personalized. Messages 3 and 4, I manually draft based on what the prospect has (or hasn’t) done in response to messages 1 and 2.
It sounds like more work, but it’s actually efficient because I’m only manually working with prospects who made it past message 2 (the hot ones). The tire-kickers drop by then. For the ones still engaged, a manual touch makes all the difference.
You don’t need all automation to feel human. Just the parts that matter most.
One thing in LiSeller you might be underusing: custom prompt fields. You can set a prompt for each message in the sequence that instructs the AI on tone and structure. So message 1 prompt says “conversational hook,” message 2 says “educational value add,” message 3 says “curiosity-driven question.”
This makes each message structurally different by design, not by accident. The AI will follow the prompt, so your sequence won’t feel repetitive. Try customizing the prompts per message rather than using one global tone throughout.