Sharing a workflow I've been running for a while that combines Claude/ChatGPT with bulk WhatsApp sending — might be useful for anyone doing personalized outreach.
The problem with WhatsApp templates
Most people writing WhatsApp message templates do it manually, which means every message sounds either too generic or too stiff. The sweet spot is a message that reads like a human wrote it for that specific person — but you need to generate 200 versions of it without spending 3 hours on copy.
The prompt stack
Step 1 — generate the base template. I use this with Claude:
Write a WhatsApp message template for [use case: appointment reminder / sales outreach / order update].
The message should:
- Be under 160 characters
- Include {{name}} and {{detail}} as placeholders
- Sound like it was typed by a person, not a CRM
- No emojis unless the context calls for it
- End with a soft CTA, not a hard sell Step 2 — stress test it. I then ask: Read this WhatsApp template back to me as if you are the recipient seeing it for the first time. Does it feel personal or does it feel like a mass message? What would you change? This almost always catches something — usually the opening line. Step 3 — variation generation. For different segments (cold leads vs existing customers vs lapsed users) I run: Give me 3 variations of this template for:
- Someone who's never heard of us
- A returning customer
- Someone who signed up but never used the product Where the CSV comes in Once I have the template locked, I drop it into Premium Sender with a CSV — name, phone, and whatever {{detail}} field the template uses (appointment time, product name, etc). Each person gets their own version, messages go out through WhatsApp Web, and I get a delivery report after. The prompt work upfront is what makes the actual send feel human at scale. Without it you're just blasting the same stiff sentence to everyone. Anyone else running a similar stack? Curious what prompts people use to QA message tone before sending to a list.
Top comments (0)