AI & Automation4 min read

Follow-Ups That Convert: How Your Agent Recovers Cold Conversations

Most COD conversations do not end with a no โ€” they end in silence. Here is how Agents schedule follow-ups that bring buyers back, within the rules.

By The Replyk Team

Silence is the default outcome

Look at any COD inbox and count how conversations actually end. A few end with a confirmed order. Almost none end with an explicit refusal. The overwhelming majority just stop โ€” the customer asked about a price, said they would think about it, got distracted by dinner or a phone call, and never wrote again. Nothing went wrong. Nobody said no. The conversation simply went cold.

Manual teams handle this badly for an understandable reason: following up on silence is tedious, low-status work with no immediate reward, and there is always a fresh, hot conversation that feels more urgent. So cold conversations accumulate unworked. For most businesses they are the single largest pool of recoverable revenue โ€” interested buyers who already know the product and the price, waiting for a nudge that never comes.

Intent decays by the hour

The value of a follow-up depends heavily on when it arrives. A buyer who asked about a product this morning still remembers the conversation, the product, and their own interest. A buyer from last week has to be re-sold from the beginning. Re-engagement works best while the original intent is still warm โ€” and warmth fades in hours, not days.

This is why follow-up timing cannot be a weekly batch job someone runs on Friday. It has to be scheduled per conversation, at the moment the conversation goes quiet, with a delay matched to the context: a few hours for a buyer who was close to confirming, a day or two for one who was earlier in the decision.

Likelihood a cold conversation revives after one follow-up
0%12.5%25%37.5%50%1 hr4 hrs12 hrs24 hrs48 hrs72 hrs

Illustrative decay pattern, not measured platform data โ€” the exact numbers vary by product and audience. The shape is the point: every hour of delay shrinks the pool of recoverable buyers.

Three kinds of scheduled outreach

During a conversation, a Replyk Agent can schedule a task to fire later โ€” and it does this judgment-first, based on how the conversation actually went. A customer who said "I get paid Friday" gets a follow-up on Friday, not a generic reminder on a fixed timer.

There are three task types, each with its own rules:

  • Scheduled message โ€” a specific message at a specific time, used when the customer named a moment ("call me after 8", "I decide tomorrow")
  • Follow-up โ€” a re-engagement touch for a conversation that went quiet, phrased naturally around where the conversation left off
  • Promotion โ€” a promotional template sent only to conversations that have never placed an order; it uses the template your business pre-configured, and the Agent never improvises the offer or picks products itself

Window-aware by design

WhatsApp's business messaging rules are strict: within 24 hours of the customer's last message you may send regular messages freely, but outside that window only pre-approved templates are allowed, and beyond 72 hours template messaging becomes a paid send. Get this wrong repeatedly and the platform can restrict your number โ€” which, for a COD business, means losing the sales channel itself.

Replyk handles the window logic automatically. A scheduled message that fires inside the 24-hour window goes out as a normal message. The same task firing outside the window converts to your approved re-engagement template instead. Paid sends beyond 72 hours happen only if you have explicitly enabled them in your settings โ€” by default, the Agent stays inside the free messaging windows. Your team schedules nothing and calculates nothing; the rules are simply enforced.

What a good follow-up sounds like

A bad follow-up is generic pressure: "Are you still interested?" It reads as automated because it is, and it gives the customer nothing new to respond to. A good follow-up continues the conversation: it references what the customer was considering, answers the hesitation they expressed, or tells them the thing they were waiting to know.

Because the Agent schedules follow-ups from inside the conversation, it has the context to write them this way โ€” it knows the product discussed, the price point reached, and the reason the customer hesitated. The follow-up arrives as the next message in a real exchange, not as a broadcast. That difference is what separates re-engagement that converts from re-engagement that gets your number blocked.

The compounding value of systematic recovery

No single follow-up is dramatic. The impact comes from the fact that every cold conversation gets one, every day, without anyone on your team remembering to do it. If your inbox produces thirty quiet conversations a day and even a modest share of them revive, that is a steady stream of confirmed orders that previously did not exist โ€” from traffic you had already paid for.

This is the quiet half of what an AI sales Agent does. The visible half is answering fast and confirming orders. The invisible half is never letting an interested buyer slip away unworked. Together they are why automated conversations out-convert manual ones even when the manual team writes perfectly good messages โ€” the Agent simply never forgets to send the next one.

Frequently Asked Questions

Do follow-ups break WhatsApp's messaging rules? No โ€” follow-ups are window-aware and use approved template messages when a conversation is outside the 24-hour window.

How does the Agent decide when to follow up? Follow-ups are scheduled based on conversation state and intent decay, so cold leads are re-engaged at the right moment.

Can I control follow-up timing? Yes โ€” scheduled tasks let you set the cadence and content of recovery messages.

Ready to automate your WhatsApp sales?

Join the COD businesses in Morocco already using Replyk to scale their confirmation operations.

Create Account