The five-minute rule: what a WhatsApp-first follow-up stack looks like

Every clinic we take over promises a fast follow-up. Almost none deliver one, because the follow-up is somebody's second job, done between consultations, in whatever tab the CRM lives in.

A five-minute response is not a discipline problem. It is a system problem. Here is the shape of the system that actually holds.

Why five minutes

The data on response times is not marketing folklore. Across the accounts we run, the difference between a lead answered in five minutes and a lead answered in an hour is roughly a 40% drop in booked consultations. At two hours it is closer to 60%.

The reason is behavioural, not technical. A hair transplant patient submits four to six enquiries in one browsing session, on their phone, at night. Whoever replies first controls the conversation. Whoever replies fourth is a footnote.

Routing before humans

Before any human touches a lead, the system decides who owns it.

• By market, so a UAE lead does not sit in a UK inbox at 3am.
• By language, obvious, still gets skipped.
• By availability, round-robin only among agents actually online, checked against a shared calendar or Slack status.
• By source, so a Meta lead and a Google brand-search lead do not get the same opening message.

A round-robin that does not check availability is worse than no routing at all. It puts leads in the queue of the agent who happens to be on holiday.

The template library

One template per market, per source, per common objection. Not one master script.

For most clinics this ends up around 20 to 30 templates, all under 40 words, all written in the same voice, all with three fields the agent fills in by hand: name, referenced goal, next available slot.

The 40-word cap matters. Anything longer reads as marketing, and the reply rate drops.

AI first-touch, out of hours

Between 10pm and 8am local, no human is going to hit the five-minute window. This is where an AI concierge earns its cost.

The scope is deliberately narrow:

• Acknowledge the lead in under 30 seconds.
• Ask two qualifying questions, Norwood scale and travel window.
• Offer two specific slots in the next 72 hours, held tentatively.
• Hand the thread to a human at 8am with the qualification already done.

An AI that tries to close the booking overnight will lose the lead. An AI that opens the conversation, holds the room, and hands over cleanly will hold roughly 70% of what a human would have converted in the same window. That is not the ceiling. That is the floor.

The human handoff

The handoff is where most stacks break. The AI hands the thread over, the human opens a fresh CRM window, has no context, asks the qualifying questions again, and the patient churns.

The fix is boring: one thread, one interface, the AI writes a two-line summary at the top ("Norwood 4, wants 3,000 grafts, travel window mid-November"), the human picks up in the same WhatsApp thread with the patient's name and their stated goal. No re-qualification.

What we measure

Not response time in aggregate. Response time by hour of day, so the 3pm queue does not hide the 10pm collapse.

Alongside it:

• Reply rate on first message, split by template.
• Time to booked consultation, from first message to slot held.
• Conversation length, in messages. Anything above 20 messages to book is a template problem, not a patient problem.

What we ignore

• Email drip sequences. The patient is on WhatsApp. Email is where enquiries go to die.
• Chatbots that pretend to be human. Patients notice within two messages, and the trust cost is worse than the speed benefit.
• Long qualification forms bolted on before the human touches the lead. Anything more than three questions upfront doubles drop-off.

The number that matters

Percentage of leads answered inside five minutes, 24 hours a day, seven days a week. Track it as one number.

If that number is below 80%, no amount of ad optimisation upstream will fix the pipeline. The leaks are all after the click.