There are two types of agency when it comes to follow-up. The first type sends a proposal and then stares at their inbox for three days, hoping for a reply. When it does not come, they send one awkward “just checking in” email and then give up. The second type buys a sales automation tool, sets up a seven-email drip campaign, and blasts every lead with the same generic sequence until they unsubscribe or block the domain.
Both approaches lose deals. The first loses them through neglect. The second loses them through irritation. The agencies winning work are the ones who follow up consistently, personally, and based on what the prospect actually does, not on an arbitrary timeline.
AI makes this possible without requiring a dedicated sales team. It is a key part of automating agency sales from lead to signed proposal.
Why behaviour-based beats time-based
Traditional follow-up sequences are time-based. Day 1: send the proposal. Day 3: check in. Day 7: nudge. Day 14: last chance. The problem is obvious. These messages have no relationship to what the prospect is actually thinking or doing.
Behaviour-based sequences respond to signals. They trigger when a prospect takes (or does not take) a specific action. This means every message feels relevant, because it is.
Here are the behaviour triggers that matter for agency sales.
Opened proposal but did not sign. The prospect has read your proposal (tools like PandaDoc and Proposify track this). They spent 12 minutes on it, looked at the pricing page three times, but did not respond. This is a buying signal. They are interested but something is holding them back.
The follow-up: “I noticed you had a chance to review the proposal. Happy to walk through any section in more detail or adjust the phasing if the timeline needs to work differently. What would be most useful?”
Visited your website after going quiet. A prospect who went cold two weeks ago just visited your case studies page. Something has reignited their interest.
The follow-up: “Wanted to share a recent project we completed for [similar company] that might be relevant to what you are considering. Worth a quick call this week?”
Attended an event or webinar. They took time out of their day to listen to you speak. Do not waste it with a generic “thanks for attending” email. Reference something specific: “You mentioned [specific question] during the Q&A. We have seen that exact challenge at [client]. Here is how we approached it.”
Engaged on LinkedIn. Liked your post, commented on an article, viewed your profile. These are soft signals, but they indicate the prospect has you on their radar. Do not sell. Add value. Share an article, make an introduction, or comment thoughtfully on their content first.
Downloaded content or read multiple articles. They are in research mode. This is the time to offer something that moves them from research to conversation: “Given your interest in [topic], we have put together a briefing on how agencies like yours are approaching this. Happy to share it over a 15-minute call.”
Personalisation at scale with AI
The magic of AI is not that it writes follow-up emails. It is that it writes follow-up emails that sound like you wrote them individually.
Here is how to set this up.
Step 1: Write your voice guide. Take 10 emails you have sent to prospects that felt natural and effective. Feed them into Claude and ask it to analyse the writing style, tone, and personality, then create a style guide. Our prompt engineering guide covers how to do this well. This becomes the foundation for every AI-generated message. Without it, your emails will sound competent but generic.
Step 2: Create message templates for each trigger. For each behaviour trigger above, write a template that includes:
- The trigger condition (what the prospect did)
- The message objective (what you want to happen next)
- Variable fields (prospect name, company, specific content they engaged with, relevant case study)
- Your style guide
Feed these into Claude and generate 3-4 variations of each. Test them. Refine the ones that perform best.
Step 3: Build the sequences in your tooling.
- Lemlist handles behaviour-triggered sequences well and supports AI-powered personalisation, dynamically adjusting messages based on engagement.
- Apollo combines prospecting, enrichment, and sequences in one platform. Strong for outbound-heavy agencies.
- HubSpot sequences work natively with HubSpot’s CRM tracking, so website visits, email opens, and form submissions automatically trigger the right follow-up.
These tools integrate with the broader automation stack you should already have in place.
Step 4: Add the human layer. Build “pause points” into your sequences where a human reviews the AI draft before it sends. The trigger fires, the AI drafts the message, it lands in your queue for a 60-second review, and then it goes. The prospect does not know (or care) whether AI drafted the message. They care whether it is relevant to them.
Timing: when to send and when to stop
Timing is as important as content.
Response windows. After a proposal send, the highest-converting follow-up window is 48-72 hours. Not 24 hours (too eager) and not 7 days (too late). If they have opened the proposal, follow up sooner. If they have not opened it, give it the full 72 hours.
Frequency. For warm prospects, 3-4 touchpoints over 3 weeks. For cooler prospects (downloaded content, attended an event), 2-3 touchpoints over 4-6 weeks.
The breakup message. After your sequence runs its course, send one final honest message: “I have reached out a few times and it seems like the timing might not be right. No hard feelings.” This converts surprisingly often. Prospects who were too busy feel guilty and respond. Either way, you leave the door open.
When to stop. If they do not engage with any of your messages, move them to a long-term nurture list rather than continuing the active sequence. Your lead scoring system should automatically downgrade them.
What to say (and what not to say)
Do say:
- Something specific about their business or situation
- Something valuable (an insight, a relevant case study, an introduction)
- A clear, low-pressure next step (“15-minute call” not “let’s discuss a partnership”)
Do not say:
- “Just checking in” (says nothing, adds nothing)
- “I wanted to follow up on my previous email” (they know, they chose not to reply)
- “We would love the opportunity to…” (passive, vague, forgettable)
- Anything that could apply to any prospect (if you could send the same email to 100 people, it is not personalised)
Measuring what works
Track these metrics: open rate by trigger type (which behaviours lead to the highest opens), reply rate by message variation (A/B test and let the data pick the winner), and conversion to call (the only metric that really matters).
Good follow-up sequences should convert 15-25% of engaged prospects into calls. If you are below 10%, your messages are too generic or your timing is off.
Follow-up is not about persistence. It is about relevance. Build sequences that respond to what your prospects are doing, write messages that sound like a real person sent them, and know when to stop. That combination, powered by AI but guided by human judgement, is what separates agencies that win work from agencies that wait for the phone to ring.
This is part of The Pitch Stack, a series on agency sales and new business strategy. Subscribe to the newsletter to get new articles weekly.
