How to Run a WhatsApp Campaign Using an AI Employee

In seven minutes this video shows the end-to-end flow: picking a Meta-approved template, uploading a contact list, letting the AI employee send it, and watching the AI handle replies as they come in — including the one that books a demo on the spot.

Most WhatsApp campaign tools stop at the send: they fire the template, log the delivery and read-receipts, and leave the replies to a human agent (or worse, a shared inbox that nobody checks). The result is the same every time — a decent open rate, a terrible conversion rate, and a pile of unanswered "yes, interested" replies.

Running the same campaign through an AI employee flips that: the AI reads every reply, understands the intent in natural language (English, Hindi, Hinglish, Tamil, Telugu — whichever your audience uses), qualifies the lead against your criteria, and takes the next step automatically — book the meeting, send the payment link, update the CRM stage, or route to a human when there’s a real reason to.

A Nexshift WhatsApp AI campaign always has the same three pieces. Get these right and the campaign works; miss one and it stalls at exactly the point you’d expect.

1. 1

   A Meta-approved template

   WhatsApp Business API only allows outbound sends via a template that Meta has pre-approved. Templates fall into Marketing, Utility, Service and Authentication categories, each with its own pricing and policy rules. Nexshift ships standard templates for common use cases (EMI reminders, class notifications, offer alerts, re-engagement) and will submit custom ones for approval — usually under an hour for Utility/Service.

2. 2

   A contact list

   Either uploaded as CSV, pulled from Nexshift CRM by stage/tag/filter, or captured live via webhook from TagMango, your landing page, Shopify, etc. Opt-in state is enforced — a number that didn’t opt in won’t be sent to, even if it’s on the list.

3. 3

   An AI employee attached to replies

   This is the piece most tools don’t have. You configure the AI employee once (identity, tone, qualification criteria, allowed next-steps, escalation triggers), and every reply across every future campaign is handled by it. The AI uses the same prompt stack as our voice agents, with your product knowledge-base attached via RAG.

If you watched the video, you already saw this flow. Here it is written down so you can follow along the first time.

1. 1

   Pick or submit a template

   Open the Campaigns tab in Nexshift CRM, click New Campaign, choose from your approved templates or submit a new one. For a new template, provide the body text with variables (`{{1}}`, `{{2}}`), header/footer, buttons (if any), and the category. Meta approval is near-instant for Utility and Service; Marketing takes longer.

2. 2

   Upload or select the audience

   Upload a CSV with phone, name, and any merge variables (loan amount, due date, order id, etc.). Or filter your existing leads: stage = "new lead" + source = "landing page" + language = "Hindi". The CRM shows the count and flags opt-out violations before you proceed.

3. 3

   Attach the AI employee

   Select the AI employee that will handle replies for this campaign. If this is your first campaign, Nexshift’s team configures it with you — identity, voice, qualification questions, and the actions it can take. On subsequent campaigns, you reuse the same AI employee.

4. 4

   Dry-run first

   Flip the dry-run switch and send. The CRM logs every "would send" without touching WhatsApp — you verify the merged copy on 10 random leads before anything leaves your Business number. This has saved us (and our customers) more than once.

5. 5

   Go live and watch the lead cards

   Flip off dry-run, hit Send. Delivery and read-receipts stream in; replies land on the lead card with the AI employee’s response underneath. Qualified leads move stages; escalated ones hit a human in your CRM with full context.

This is the part most prospects want to see. Here’s a concrete example from a jewellery campaign — a Diwali offer blast where the template said "Reply with your preferred visit day for an exclusive gold-rate preview".

• Customer replies: *"I want gold rate for 22K today"* — AI employee answers with today’s rate (pulled from a live price feed), offers a visit booking.
• Customer replies: *"Nahi chahiye, bass dekh raha tha"* (not interested, just looking) — AI employee thanks them politely, updates CRM lead to "cold", stops the campaign for this contact.
• Customer replies: *"Saturday 11am"* — AI employee confirms the slot, writes a booking into the showroom calendar, sends a confirmation message with the address and a map link.
• Customer replies: *"Can I pay in EMI?"* — AI employee answers the EMI question using the product KB, then offers to connect them with a sales person; if they say yes, it escalates to a human in the CRM with the full chat so far.

Every one of these outcomes lands on the lead card — stage change, task, CRM note, calendar event. The sales lead doesn’t have to read 2000 WhatsApp replies to find the 50 that matter; they just open the "Hot" stage and work the shortlist.

A quick orientation on what charges apply, because WhatsApp Business API pricing confuses almost everyone the first time.

A few policy limits worth knowing before you launch: you can only send to numbers that have opted in (we enforce this at upload time); Marketing templates have a 24-hour frequency cap per contact; and any reply opens a 24-hour service window during which you can send free-form messages — this is the window the AI employee works in.