The WhatsApp journey that books demos

Stage one: the first reply, under sixty seconds

In India the sale happens on WhatsApp and it is largely decided in the opening exchange. That makes reply latency a conversion metric rather than a service metric, which is a genuinely different way to run the channel. We hold a 33 second average across every number we operate, day or night, holiday or not.

The critical design decision is what that first reply contains. A menu tree is the wrong answer — it tells the customer they are talking to software that cannot help them and it burns the one moment of maximum intent. The agent has to actually answer the question asked: is this available in my city, what does it really cost, does it work for my situation.

That means the agent needs your catalogue, pricing logic, offer stack, delivery policy and the twenty objections your team hears every week loaded before it goes live. When it does not know something, it says so and escalates rather than inventing an answer — a confidently wrong price costs more than a slow reply, and that constraint is built in rather than hoped for.

• Answer the actual question first; never open with a menu.
• Target the reply inside the first minute — we hold 33 seconds average.
• Load catalogue, pricing rules, delivery and refund policy, and the top twenty objections.
• Hard constraint: escalate on uncertainty rather than guess.

Stage two: qualify inside the conversation, not with a form

The instinct is to fire three qualifying questions the moment someone replies. It converts badly, because it inverts the exchange — the customer came to get information and is instead being processed.

Qualification works when it is embedded in genuinely useful answers. A question about availability naturally surfaces city. A question about pricing naturally surfaces budget band and use case. The agent gathers what it needs while doing the thing the customer came for, and the qualification data lands as a by-product rather than a toll.

What you are scoring toward is a simple decision: does this person need a human conversation, do they need nurturing, or do they need nothing. Getting that classification right is what makes 1,840 qualified leads a day meaningful rather than a large number of rows. An unqualified lead handed to a rep is worse than no lead, because it costs the rep an hour and teaches them to distrust the queue.

Stage three: journeys for the not-yets

Most inbound is not ready today, and the whole economic case for the channel rests on what happens to those people. Handled properly they are your cheapest pipeline; handled badly they become a list you broadcast at until they block you.

The always-on journeys do this work: welcome, abandoned cart, post-purchase, win-back, re-order reminders. Each is templated, approved and measured on revenue rather than delivery rate — which is the metric distinction that separates a journey from a spray. And critically, every one of them drops the customer into a live conversation the moment they reply, rather than into another automated step.

Cadence is the part most brands get wrong in the other direction. A journey that touches someone every second day for a fortnight does not feel attentive, it feels like a leak in their inbox, and the block that follows costs you the contact permanently. We space nurture against observed behaviour rather than against a calendar — a browse, a price question, a returned visit moves someone forward a step; silence moves them into a slower track rather than a louder one.

The design rule we hold to is that no automated message should be un-repliable. If someone answers a win-back nudge with a question, the agent is right there, in context, with the full history of the thread. A journey that cannot absorb a reply is a newsletter.

• Welcome, abandoned cart, post-purchase, win-back and re-order as standing journeys.
• Every journey measured on revenue, never on delivery or open rate.
• Any reply to any automated message opens a live conversation with full context.
• Dated re-entry for not-nows rather than a graveyard list.

Stage four: booking the meeting without a handoff gap

The gap between intent and calendar is where most funnels lose the deal. Someone says yes, they get told a team member will reach out, and the momentum built over the previous four minutes evaporates into an unspecified future.

The agent books directly. When intent crosses the threshold it offers real slots and confirms the meeting inside the same thread, then hands the conversation to a human with the qualification data and full history attached. Nobody re-asks what the customer already answered, which is the single most common way a warm conversation goes cold.

There is a smaller mechanic inside this that is worth stealing regardless of what tooling you run. Offer two or three concrete slots rather than a scheduling link. A link is a task the customer has to go and do; two named times are a decision they can make in the thread they are already in, with one thumb, in eight seconds. The conversion difference between those two designs is larger than almost any copy change you will test this quarter.

Where a call is the better next step, the journey passes to the Voice Engine instead — sub-800ms Hinglish voice-to-voice, dialled while the person is still on your page. And where the lead came in through outbound rather than inbound, the Outreach Engine is running the same contact record with shared suppression across email, WhatsApp and LinkedIn, so the customer never gets hit twice in a day by two systems that do not know about each other.

The anatomy of 412 demos a day

The headline number is 412 demos booked a day, and it is worth decomposing because the shape matters more than the total. Upstream of it sits 1,840 leads qualified a day and 7,420 AI calls a day, running across WhatsApp and Voice as one motion rather than as separate channels reporting separately.

What makes the number hold is not the volume at the top. It is that roughly 80% of conversations resolve without a human, so the human capacity that exists gets spent entirely on the conversations that need judgement. Every reply is classified — interested, objection, wrong person, not now — and routed accordingly. Not-nows go into a dated re-entry track rather than a graveyard, which is where a meaningful share of next quarter's demos come from.

None of this appears in week one, and you should be suspicious of anyone who says it does. The agent launches supervised, with a human watching every thread for the first fortnight while corrections feed straight back into its instructions. Accuracy climbs fast in that window, then journeys and broadcasts switch on and escalation thresholds relax. From there it is a monthly loop: read the transcripts, patch the gaps, add the next use case.

The one number worth instrumenting from day one is time-to-first-reply, because it moves everything downstream and it is the cheapest thing to fix. Most brands discover their real median is measured in hours, and simply closing that gap changes the conversion rate of traffic they are already paying for.