DEV Community

michael fabien
michael fabien

Posted on • Originally published at askamelie.com

Turn Your ManyChat Bot Into an AI Product Recommendation Engine (In 5 Minutes)

#manychat #chatbots #ecommerce #ai

The Problem: Your Bot Gets 100 Messages/Day But Sells Like It Has 0 Context

You've built a beautiful ManyChat flow. Customers message, bot replies, and... nothing.

Why? Because your bot is keyword-matching, not understanding.

Customer: "Do you have this in blue?"
Bot: (doesn't know your catalogue)

Customer: "I saw this on Instagram 3 days ago"
Bot: (forgets every interaction)

Customer: "Is this still available?"
Bot: (checks a spreadsheet, misses the context that they asked yesterday)

Result: 100 DMs, 2 sales. Meanwhile, your competitor's bot recommends the exact blue version, remembers the customer, and closes the loop.

The Solution: Add a Brain, Not a New Bot

Here's the pattern ManyChat agencies are using right now:

  1. Keep your existing bot (no rebuild, no hours wasted retraining your team)
  2. Plug in a knowledge layer (your product catalogue, pricing, inventory, past orders)
  3. Route smart questions through an AI brain (webhook External Request, 2 minutes setup)
  4. Get rich responses (recommendation cards with image + price + Buy button, sent back as ManyChat cards)

It's not about replacing ManyChat. It's about upgrading it.

What Changes

Before:

  • Customer: "Recommend something"
  • Bot: (picks random item from flow hardcode)

After:

  • Customer: "Recommend something under $50"
  • Brain: (reads catalogue, checks inventory, picks top 3 matches, sends card with photo + price + link)
  • Customer: (clicks "Buy" button, flow captures cart)

No new UI. No flow rebuild. Just smarter responses.

For Agencies: This Is Your Margin Play

Instead of:

  • Spending 20 hours per client building spaghetti flows
  • Maintaining 500 flow branches across 10 clients
  • Rebuilding every time they want to change pricing

You now:

  • Copy/paste one webhook integration (5 min)
  • Upload their product feed (bulk import)
  • Let the brain handle 80% of questions
  • Scale to 50 clients without hiring

Time/client drops from 20h→2h. Margin flips from 30%→70%.

How It Works

  1. Import your ManyChat flow (screenshot or export JSON from n8n/Make)
  2. Upload your product database (CSV, Shopify, WooCommerce)
  3. Add one webhook to your bot (copy-paste External Request URL)
  4. Brain handles the rest: intent matching, product lookup, recommendations, inventory check

Existing customers see no change. New ones immediately get smarter responses.

The Numbers

Teams using this pattern report:

  • Repeat customer rate: +35% (bot remembers preferences)
  • Average order value: +$40 (recommendations work)
  • Manual follow-ups: -60% (bot doesn't miss questions)
  • Support time: -50% (bot actually solves things)

Try It

The brain for your ManyChat bot is live. No credit card, no rebuild, no training docs. Import your flow, upload your catalogue, plug in the webhook.

→ Start here

"Your bot already gets messages — we make it SELL."

Originally published on Ask Amelie

Top comments (0)