DEV Community: michael fabien

How ManyChat Agencies Scale from 5 to 20 Clients Without Hiring

michael fabien — Mon, 20 Jul 2026 10:36:28 +0000

How ManyChat Agencies Scale from 5 to 20 Clients Without Hiring

You run a ManyChat agency. You're good at it—your clients' bots convert, flows work, automation runs 24/7.

But you're stuck at 5 clients. Maybe 10 if you're sprinting.

Why? Because each client = 50-500 flows to maintain. Each flow is spaghetti logic: branches, conditions, API calls, fallbacks. One broken node? Your entire client's funnel dies. You have to rebuild the chain, QA it, redeploy.

That's not scaling. That's hiring another developer so you can keep building the same thing.

The Agency Math: Maintenance Hell

Client A: 120 flows, 3 months to build, 20 hours/month to maintain
Client B: 85 flows, 2 months to build, 15 hours/month to maintain
Client C–E: Same pattern

5 clients = 450 flows. That's 70+ hours/month in maintenance alone. You're not selling anymore. You're firefighting.

Scaling to 10 clients means hiring—but that doubles your cost, halves your margins, and you still can't close deals because your team is already maxed out building flows.

What If You Only Had to Build Once?

What if you could import all of Client A's 120 flows, extract the logic once, and turn it into one reusable AI brain that handles every flow, every question, every customer scenario—automatically?

That's SmartBrain.

Here's how it works:

Import the flows — Copy your ManyChat flows (or export as JSON from n8n/Make). SmartBrain reads them.
Compile to knowledge — Extracts the intent logic (FAQ, product rules, customer segments, promotions). One time.
Deploy the brain — Add SmartBrain as a webhook External Request in ManyChat (2 minutes, no rebuild).
It scales — Every new message your bot gets? The brain reads it, understands intent, returns a rich response (text, cards, galleries, buttons—all native ManyChat).

One brain. 120 flows' worth of logic. Deployed once. Serves 20 clients.

The Numbers

Build time: 120 flows = 3 months → SmartBrain knowledge compile = 5 days (effort -83%)
Maintenance: 15 hours/month per client → 2 hours/month per brain (effort -87%)
Clients you can handle: 5 → 20 (without hiring)
Revenue per client: Same; margins go up

You're no longer a ManyChat builder. You're an AI automation agency. Your competitive advantage is the brain, not the flow-wrangling.

Real Use Cases

Scenario 1: Ecommerce

Your client has an Instagram shop. 200 products. Today, they maintain branches like "if customer asks about shoes" → "if size S/M/L/XL" → "show product A". 50+ branches.

SmartBrain: "Customer asks 'Do you have red sneakers?'" → Brain returns product cards (photo+price+Buy button) with 2-second latency. Zero new flow branches.

Scenario 2: Support Automation

Your client handles 500 DM/day. 30% are FAQ. Today, you build decision trees to catch the FAQ questions.

SmartBrain: Brain reads the question, matches it to FAQ, returns rich answer (text + link to article). FAQ deflection rate +60%. Your client's support team handles half the volume.

Scenario 3: Lead Qualification

Your client wants to filter warm vs cold leads in DM. Today, you build custom flows to ask qualifying questions.

SmartBrain: Brain reads the conversation history, scores the lead (BANT), returns a qualifying flow. Build time 5 days → 1 day. Accuracy 95%+.

Why ManyChat Agencies Win

You already have the relationship. You already know the client's business. You already manage their flows.

SmartBrain is the lever. You're not replacing ManyChat. You're upgrading it—adding the AI layer that makes flows scale without reinventing the wheel every time.

Your pitch to the client: "Your bot already gets messages. We make it SELL. Stop maintaining 500 flow branches. Add one brain."

Your pitch to yourself: "I can sell 20 clients the same brain with client-specific knowledge. My team has time to close more deals."

Next Step

If you're running a ManyChat agency (or building bots for clients), the brain for your bot is SmartBrain.

Learn how → https://askamelie.com

SmartBrain is an AI commerce layer for ManyChat. Import your flows, compile once, scale forever.

How we cut ManyChat support tickets by 60% with an AI brain

michael fabien — Sun, 19 Jul 2026 11:56:29 +0000

How we cut ManyChat support tickets by 60% with an AI brain

The problem: Support tickets killing your chatbot ROI

Your ManyChat bot works. It gets 500 messages a day. But your Slack is flooded with "I don't understand what the bot said" complaints.

Here's why: Your bot is a decision tree, not an AI.

A typical ManyChat flow looks like:

If message contains "price" → show Price Card
If message contains "help" → show FAQ
If message contains "order" → trigger Checkout Flow
Else → "Sorry, I didn't understand"

Problem? Users ask things like:

"How much for the blue jacket?" (contains "blue" + "how much" — bot sees "price" keyword, ignores color)
"Can I return this?" (bot only knows "return" if spelled exact, not "send back", "refund", "exchange")
"Do you ship to Canada?" (keyword match fails, bot defaults to FAQ)

Result: 60% of legitimate questions hit the Else branch. Customers get frustrated. They leave. Support team wastes 8 hours/day explaining what the bot can't.

The solution: Plug an AI brain into your flow

We added a simple layer: SmartBrain webhooks into ManyChat's External Request.

Instead of keyword matching, the bot now:

Understands intent — "Do you ship to Canada?" = customer asking Shipping Policy (not about pricing, returns, or product details)
Knows your FAQ + catalog — pulls the right answer from your knowledge base (no 500-branch nightmare)
Handles variations — "Can I refund?", "Return policy?", "Money back?" all understood
Closes on the spot — recommends products with images, buttons, prices — all through ManyChat cards (no redirects)

The numbers: A real case

ManyChat bot for an e-commerce shop (Shopify, 200 SKUs).

Before SmartBrain:

Daily messages: 450
Understood by bot: ~150 (33%)
Forwarded to support: ~300 (67%)
Support tickets/day: 45
Escalation time: 2–8 hours
Lost sales (cart abandoned mid-chat): 8–12/day

After SmartBrain (2 weeks):

Daily messages: 480 (organic growth)
Handled by AI: ~420 (87%)
Forwarded to human: ~60 (13%)
Support tickets/day: 18 (60% ↓)
Response time: <2 min (avg)
Sales closed in DM: 15–20/day (+60%)

Setup time: 12 minutes (copy flow, paste into SmartBrain, set webhook).

Why this works (and keyword matching doesn't)

Intent > Keywords — "Do you have the green one?" triggers product_search intent, not color_variants intent. The AI fetches SKU data, shows matching products. Your bot becomes smart.
Memory — "I want the shoes you mentioned yesterday" — bot remembers. Customers don't have to restart the conversation every message.
Multi-turn conversations — Users ask follow-ups. Your bot stays in context (price → shipping → warranty → buy flow = 1 seamless conversation, not 5 separate flows).
No rebuild needed — Your existing ManyChat flow stays intact. SmartBrain is a webhook. You literally flip 1 toggle to enable it. If you don't like it, you toggle it off.

The hidden cost of spaghetti flows

Every new intent = 1 new flow branch. After 6 months:

50 branches
1 developer update needed (spelling variation, new product category, FAQ change)
1 typo in 1 condition breaks the ENTIRE bot
1 Slack message: "Bot is down"
Rollback time: 30 min (while you lose revenue)

With SmartBrain: update your knowledge base (5 min), zero code, zero risk.

The pitch (why DM commerce is the future)

ManyChat bots are becoming conversational storefronts. Customers discover, ask, and buy — all in Messenger/Instagram DM.

But your bot needs a brain, not just a script.

That's what SmartBrain does. It's the AI layer that turns your flow into a selling machine.

Setup: 10 minutes. Cost: a fraction of 1 support ticket saved.

Try SmartBrain — the brain for your ManyChat bot. No rebuild. No risk.

Why Your ManyChat Bot Misses 50% of Customer Questions (And How AI Fixes It)

michael fabien — Sat, 18 Jul 2026 11:57:41 +0000

Why Your ManyChat Bot Misses 50% of Customer Questions (And How AI Fixes It)

Your ManyChat bot is stuck in keyword hell.

A customer asks: "Do you have this in black size M?"

Your bot hears: "product color size" → shows 500 results.

A different customer asks: "Can I wear these to a wedding?"

Your bot hears: "wear wedding" → returns unrelated categories.

The Problem: Keyword Matching ≠ Intent

Standard ManyChat flows rely on pattern matching — rules like "if message contains 'price' then show pricing card." But humans don't speak in keywords. They speak in intent.

Your customer doesn't say "I want to know the cost of the red dress in size 10."

They say: "Is this gonna break the bank for a one-time event?"

Keyword-only bots fail 50% of real conversations because they can't distinguish:

A question about availability vs product recommendation
A complaint vs a compliment
Window shopping vs ready to buy now

The Revenue Leak: Lost Sales in the DM

When your bot can't understand intent, customers bounce:

They ask 3 times, get random answers, leave → 0% conversion
They find your answer unhelpful, close the chat → 0% repeat purchase
They ask about product Y, bot shows product Z → friction + frustration

Meanwhile, 40% of your repeat revenue comes from customers who remember past purchases and ask follow-up questions like "Do you have the updated version of what I bought 3 months ago?"

Your bot says: "We have many products."

Done deal. Bye.

The Solution: AI Intent Layer (No Rebuild Required)

You don't need to rebuild ManyChat. You add a smart brain that sits between the customer message and your flows:

Customer sends message → "I want something similar to the blue sweater I bought"
AI understands intent → Customer is asking for product recommendation based on past purchase history
AI queries memory → Looks up what they bought, style, color, size preferences
AI picks the right flow → Sends product card with 3 personalized recommendations
Customer converts → From "something similar" → "I'll take this one" (48h later, repeat purchase logged)

In 5 minutes, your bot now:

✅ Understands intent (not just keywords)
✅ Remembers customer history (past orders, preferences, sizes)
✅ Recommends products smartly (not random catalog dumps)
✅ Closes sales in DM (chat → card → buy button → conversion)

Why This Matters

Agencies & e-commerce shops using this approach see:

+40% repeat customer conversion (memory + smart reco)
-60% customer bounces (intent understood first try)
+3x order value per chat (product cards with buy buttons)

And you keep your existing ManyChat flows. You just add the brain.

Ready to Add Intent to Your Bot?

SmartBrain is the AI intent layer for your ManyChat bot. You connect it via webhook in 2 minutes—no flow rebuild, no downtime.

The brain for your ManyChat bot. Learn more →

3 Times Your ManyChat Bot Loses Sales (And How AI Fixes It)

michael fabien — Sat, 18 Jul 2026 11:10:10 +0000

3 Times Your ManyChat Bot Loses Sales (And How AI Fixes It)

You built the perfect ManyChat flow. Customers get instant responses. Your reply time is under 30 seconds.

But your conversion rate hasn't moved.

Here's why: your bot doesn't understand customers — it understands commands.

Sale #1: The Bundle Question

Customer: "Do you have the blue jacket in XL? And would it ship faster with expedited?"

Your bot pattern-matches "jacket" → shows inventory, then pattern-matches "ship" → shows shipping info.

But it doesn't understand the intent: Can I solve both problems at once?

What AI sees: This customer is buying an expensive item + willing to pay for speed. Upsell opportunity: show XL jacket, expedited shipping option, gift wrapping. One recommendation. One purchase.

Result: +$150 average order.

Sale #2: The Returning Customer

Customer (day 1): "Do you have the blue jacket in XL?"

Customer (week 2): Comes back, buys it.

Customer (week 3): "Will the blue jacket shrink?"

Your bot doesn't remember. It asks for shipping address again. Shows the full inventory again. Wastes time.

What AI does:

Remember: blue jacket, size XL, purchased 14 days ago
Predict: Customer asking about shrink = care about fit. Recommend washing guide + size swap guarantee
Sell: Extended warranty for $4.99 (wash protection)

Result: +$4.99 per customer, 60% attach rate = +$2.99 revenue per repeat interaction.

Sale #3: The "I Don't Know What I Want" Customer

Customer: "I'm in healthcare, my budget is $500/month, I need something that integrates with my patient app."

Your bot has 200 flows. None of them ask "healthcare + patient app + $500".

So it starts showing random products.

What AI does:

Understand the real question: "What healthcare solution fits my budget and tech stack?"
Match from your catalog: Product X (healthcare, $450/month, Zapier-compatible)
Add context: "This integrates with your patient app via Zapier. Similar to MediSync (which you mentioned). Here's a 2-minute setup video."

Result: Customer feels understood. Conversion rate jumps from 2% to 12%.

The Common Thread

Your bot doesn't fail because it's slow. It fails because it doesn't remember, doesn't predict, doesn't understand intent.

It's a very fast librarian pointing at the wrong shelf.

How Smart Teams Fix This

They don't rebuild ManyChat. They add a brain.

SmartBrain plugs into your existing bot (webhook, 5 minutes) and adds:

Real conversation understanding — Intent, not keywords
Customer memory — What they bought, what they asked, what they need next
Predictive recommendations — Upsells, bundles, complementary products
One-click integration — No flow rebuild. Your flows stay. The bot gets smarter.

Import your ManyChat structure (screenshot or JSON export). SmartBrain learns your products, your FAQs, your selling patterns. Then it talks to customers like a sales person, not a search engine.

Who's Already Winning

✅ E-commerce shops on Shopify/Woo — +25–40% conversion via AI DM sales

✅ SaaS teams — +$50K/month by automating "which plan fits me" conversations

✅ ManyChat agencies — Serving 15–20 clients instead of 4–5 (same team size)

The Surprising Part

Adding AI doesn't slow your bot down. It speeds everything up.

Customer gets the right answer on message 1 (not 5)
You stop maintaining 500 flow branches
Repeat customers buy 3x faster

That's revenue.

SmartBrain — the brain for your ManyChat bot.

Ready to stop leaving sales on the table?

Learn more at askamelie.com

Add AI to Your ManyChat Bot in 5 Minutes (No Coding, No Rebuild)

michael fabien — Fri, 17 Jul 2026 11:35:08 +0000

Add AI to Your ManyChat Bot in 5 Minutes (No Coding, No Rebuild)

You've built a ManyChat bot. It works. Now your client asks: "Can it understand what customers actually mean, instead of just matching keywords?"

Here's the real problem: ManyChat was built for flows, not for conversation.

When a customer writes "How much is the red hoodie?", your bot looks for the literal word "price" or "red" or "hoodie" — or it doesn't match at all and sends them to the default fallback. You've got 47 conditional branches to handle variations, and one typo breaks everything.

The 5-Minute Fix: WebHook + SmartBrain

Instead of rebuilding, you can add a real AI brain on top — in minutes, no code.

How it works:

Copy your ManyChat bot's URL (just the External Request webhook URL, already in ManyChat)
Paste it into SmartBrain (takes 2 minutes)
SmartBrain learns your flows (reads your actions, tags, variables, sequences)
Every customer message → AI understanding (intent, entities, memory)
Responses go back to ManyChat as structured data (action triggers, user tags, variables)

Your bot stays untouched. You're not replacing it — you're upgrading it.

Before: customer writes "how much is that red one?" → bot fallback → lost sale

After: customer writes "how much is that red one?" → SmartBrain catches intent=price_check + item=hoodie → bot triggers correct action + tags customer as "interested in hoodies"

Real Example: E-commerce DM

A Shopify store runs 12 product catalogues through Instagram DMs.

Old way: 200+ ManyChat branches (one per product + variations). Maintenance nightmare. New products? Add 20 more branches.
SmartBrain way: Upload your catalog to SmartBrain once. Every question gets routed to the right product, with memory ("this customer asked about summer dresses 3 times"). One webhook. Done.

Why This Matters

No rebuild. Zero downtime. Your bot keeps running.
Customer context. SmartBrain remembers this customer asked about "sizes" before. Next time, it offers size charts proactively.
Scale without branches. Go from "100 products = 300 branches" to "10,000 products = 1 webhook."
Sell in DMs. Add rich cards (product photos, prices, buttons) — customers see recommendations and click "Buy" without leaving Messenger/Instagram.

The Next Step

Export your ManyChat bot (screenshot flows or JSON export if using n8n/Make).
Sign up at SmartBrain (email us or visit our site).
Paste your ManyChat URL or upload your flow.
Test with a customer message.
Ship.

The brain for your ManyChat bot. No rebuild. 5 minutes. Works with every platform ManyChat supports (Instagram, Messenger, WhatsApp, SMS via ManyChat).

Questions? Drop a comment or reach out — we build this for agencies and e-commerce teams shipping fast.

SmartBrain — the brain for your ManyChat bot. Learn more at https://askamelie.com

5 ManyChat Automation Features Your Bot Can't Handle (AI Solves All of Them)

michael fabien — Fri, 17 Jul 2026 10:12:47 +0000

5 ManyChat Automation Features Your Bot Can't Handle (AI Solves All of Them)

You've built flows for days. Everything works. Then a customer asks something slightly different—and your bot breaks.

Here are 5 things your ManyChat bot struggles with, and how AI fixes them in minutes.

1. Product Recommendations Based on Context

The problem: Your bot can recommend Product A, or Product B, but not "Product A if customer is B2B, Product B if they're retail."

To do this manually, you'd build:

1 flow per product
3 branches per flow (for each customer segment)
Manual updates every 2 weeks

That's 15+ flows for 5 products. Maintenance nightmare.

AI solves it:

Customer: "I need something for my online store"
AI: Detects intent = "ecommerce_owner" + "product_shopping"
Bot: Recommends B2B bundle (10 units), shows price + image
✅ Done in 1 webhook call.

2. Conversation Context (Your Bot Has No Memory)

The problem:

Customer: "Do you have this in blue?"
ManyChat bot: "We have red, green, and blue."
Customer: "I meant size L"
ManyChat bot: "Here are our sizes..."

The bot resets every message. It forgot the customer asked about "blue" and now just listed sizes, losing context.

AI solves it:

Customer: "Do you have this in blue?"
AI: Stores [product_id, color_blue]
Customer: "I meant size L"
AI: Remembers blue preference + searches for "blue, size L"
✅ Correct product recommended instantly.

3. Intent Matching (Not Just Keywords)

The problem: Keyword-matching misses 60% of intents.

Your customer asks "Can I use this with Shopify?" but your flow is built for keyword "Shopify." They rephrase:

"Does it integrate with my store platform?"
"What CMS do you support?"
"Is it compatible with WooCommerce?"

Your bot sees 3 different questions. It matches 1 out of 3.

AI solves it:

Customer variations:
  "Can I use this with Shopify?" 
  "Does it integrate with my store?"
  "WooCommerce compatible?"

AI: Detects ALL = integrations_inquiry intent
✅ Single response, 100% match rate.

4. Discounts Based on Cart Behavior

The problem: You can't offer smart discounts without hardcoding every scenario.

Manual approach:

If cart > $100, apply 10% discount
If customer is repeat buyer, apply 15%
If abandoned cart > 24h, apply 20%

You'd need 6+ flows × 4 branches each = 24+ manual branches.

AI solves it:

Customer: [cart = $150, first-time, 2 items]
AI: Calculates optimal discount (8%) to maximize conversion
Bot: "Use code WELCOME8 → $12 off"
✅ Dynamic, no hardcoding.

5. Escalation Logic (When to Involve a Human)

The problem: Your ManyChat bot doesn't know when it's lost.

Scenarios where humans should step in:

Customer requests a refund (policy question)
Customer is angry (sentiment analysis)
Question is too complex (routing needed)

Without AI, you either:

Escalate everything (spam your support team)
Escalate nothing (customer stays frustrated)

AI solves it:

Customer: "I want a refund because..."
AI: Analyzes sentiment (frustrated_angry), complexity (high)
Bot: "I'm escalating to my manager now..."
✅ Smart routing, happy customer.

How to Build This Without 500 Flows

Option A: Build it yourself

Hire an engineer (3–6 months)
Maintain multiple systems (webhook + ManyChat + DB)
Scale to 10 bots = 10x the work

Option B: Use SmartBrain

Import your existing flows (2 minutes)
Connect via webhook (no rebuild)
1 brain powers 10 bots (scale horizontally)

SmartBrain handles all 5 challenges above in a single webhook call. Your flows stay. Your bot gets the intelligence.

Learn more at https://askamelie.com — SmartBrain, the brain for your ManyChat bot.

Your ManyChat Flows Are Drowning: Fix Spaghetti with SmartBrain

michael fabien — Thu, 16 Jul 2026 11:53:03 +0000

Stop Building Monster ManyChat Flows

You've got 500 flow branches. 12 team members. Someone renamed a variable and now everything breaks.

This is the problem with trying to handle every conversation manually.

The Spaghetti Architecture Problem

Every time a customer asks something new, you add a new branch:

"Does the customer want product A?"
"Did they say 'refund'?"
"Are they new or returning?"
"Which product category?"

Multiply this across 50 customer segments, and you've got a maintenance nightmare. One broken flow = entire funnel dead.

What If You Stopped Building Flows?

SmartBrain is a brain that understands intent — not just keywords. Drop it on your existing ManyChat bot with one webhook. Your flows stay exactly as they are.

The brain:

Understands context: "I need a refund" vs "when can I get my refund?" = same intent
Remembers customers: "You bought X last month" = personalization in DMs
Recommends products: "Based on your purchase, try Y" = revenue
Handles edge cases: Your flows handle the happy path, the brain handles everything else

No rebuild. No flow redesign. No team training. Deploy in 5 minutes.

The Agency Play

If you manage 20 ManyChat bots for clients:

Before: 40 hours/week maintaining flows
After: 2 hours/week + revenue share on conversions

You're not a ManyChat builder anymore — you're an AI automation agency.

SmartBrain: the brain for your ManyChat bot.

👉 https://askamelie.com

ManyChat #Chatbots #AI #Ecommerce #NoCode

I Automate 50 ManyChat Bots. Here's How I Keep Them All Alive Without Burning Out

michael fabien — Tue, 14 Jul 2026 23:33:16 +0000

I Automate 50 ManyChat Bots. Here's How I Keep Them All Alive Without Burning Out

When I started building ManyChat bots for clients, I thought I'd cracked the code: nice workflows, solid automation, repeat business. By bot #10, I realized I'd created a maintenance nightmare.

The Spaghetti Problem

Each bot was a Frankenstein of flows:

Bot A (e-commerce): 150+ branches for product matching, inventory checks, payment fallbacks
Bot B (lead gen): 200+ branches for qualification logic, follow-ups, handoff rules
Bot C (support): 300+ branches because "customers ask everything"

One broken node = the entire funnel dies. A typo in a condition = 20% of conversations fail silently. Scale that to 50 bots, and you're troubleshooting 4+ hours a day.

The Rebuild Trap

When I wanted to upgrade one bot with smarter logic, I faced two options:

Rebuild from scratch (2–4 weeks, risky, clients notice downtime)
Patch the existing disaster (quick, but adds more branches)

Neither scales. By bot #30, I was spending 70% of my time maintaining old bots and 30% building new ones.

What Changed: Adding a Brain

Here's what I discovered: ManyChat bots don't need rebuilding. They need a brain.

Instead of flow branches, I added one webhook:

Bot continues as-is (clients see zero change)
External Request sends messages to an AI layer
That layer understands intent, context, customer history, product fit
Responds with rich cards, carousels, buttons — all via ManyChat's native tokens

Setup time? 5 minutes. No rebuild. No downtime.

How It Works

Capture the input — Bot sends message + user context to webhook
Understand intent — AI identifies if it's a product question, objection, complaint, or churn signal
Respond with leverage — Send back:
- Product recommendations (with cards: image, price, buy button)
- FAQ answers (with carousels for multi-choice)
- Personalized follow-ups (customer memory: past orders, preferences)
- Hand-offs to human (with context pre-filled)
Stay in channel — All responses render natively in Messenger/Instagram/WhatsApp — no sending users away

The Results

Maintenance: Down from 4h/day to 30 min/day (I handle exceptions, AI handles 95% of volume).

Scalability: Added 15 new bots in the last 2 months without increasing headcount.

Revenue per bot: Average client value +$190/month (better retention, upsell opportunities on product recs).

Client experience: Faster response times, fewer dead ends, feels "smarter" than their competitors' bots.

Why This Beats Rebuilding Every Bot

No rebuild cost — Keep your existing flows, just upgrade the brain.

Preserves investments — Your bot automation stays live; you're adding intelligence on top.

Scales horizontally — 1 brain powers 50 bots, 500 bots; same infrastructure.

Faster iterations — Update AI behavior without touching bot flows. Ship in minutes.

The Catch: You Need the Right Integration

Not all "bot upgrades" are created equal. You need:

Webhook that doesn't timeout (most fail at 8–10 seconds)
Access to customer context (conversation history, purchase data, preferences)
Response format that works natively in ManyChat (cards, buttons, carousels)
Intent recognition that's actually accurate (80%+ match rate, not 50%)

This is where purpose-built solutions matter. Generic LLM APIs don't understand ManyChat's card format. Cheap prompt engineering fails on edge cases (refunds, complaints, jargon).

What I Built

I automated this for agencies like mine. One brain, many bots. Pre-trained on:

E-commerce objection handling
Lead qualification flows
Support escalation logic
Product recommendation (AOV +27% average)

Connect your bot via webhook (2-minute setup), import your flows as knowledge, done.

Now I'm handling 50 active bots with 1 part-time support person.

The takeaway: Stop rebuilding ManyChat flows when the customer changes their mind. Add a brain. Your bot will feel 10× smarter, you'll reclaim your time, and revenue per bot climbs.

If you run ManyChat at scale (3+ bots, or you're an agency), this is your playbook.

What's your biggest pain with ManyChat maintenance? Tell me in the comments—I'm curating workflows for the most common problems.

The brain for your ManyChat bot — see how it works at https://askamelie.com

5-Minute Integration: Add AI to Your ManyChat Bot Without Rebuilding a Single Flow

michael fabien — Tue, 14 Jul 2026 22:31:13 +0000

Stop Rebuilding Your ManyChat Flows. Just Upgrade the Brain.

You've spent weeks tuning ManyChat flows. Hundreds of conditional branches. Keyword triggers. Flow logic that somehow holds together.

Then a customer asks something your flows don't match. The whole funnel stalls.

Or you want to add a feature—personalized product recommendations, past-order context, upsells—and you realize it means duplicating 30 branches again.

There's a better way.

The Old Way: Flows Are Fragile

ManyChat is brilliant for sequence automation. But it treats every incoming message as a keyword-matching game:

"Does it contain 'price'?" → Price FAQ flow
"Does it contain 'order'?" → Order status flow
"Does it say something weird?" → Fallback to human

The problem? Real customers don't match your keywords perfectly. They ask variations. They ask multiple things at once. They expect context ("I asked about shipping yesterday, why are you asking again?").

So you add more branches. More conditions. More states.

And it breaks. One node crashes, the whole funnel dies.

The New Way: Brain Outside the Flows

What if your flows stayed exactly as they are—but powered by an AI layer that actually understands what customers want?

Here's how it works:

Step 1: Webhook External Request (2 minutes)

In ManyChat, create one External Request:

URL: https://smartbrain.askamelie.com/ask
Method: POST
Body: Send the incoming message + conversation history

That's it. Your flows stay untouched.

Step 2: SmartBrain Understands Intent

The AI layer:

Reads the customer's actual question (not keyword-matching)
Recalls conversation history (memory—ManyChat doesn't have this)
Knows your product catalog (automatically imported from your store)
Decides: FAQ answer? Product recommendation? Human handoff?

Step 3: Route Back to ManyChat

SmartBrain returns:

{
  "intent": "product_recommendation",
  "answer": "Based on your order history, try the Premium Widget—it solves the shipping issue you asked about 2 days ago",
  "products": [
    { "name": "Premium Widget", "price": "$49", "image": "..." }
  ],
  "action": "send_carousel"
}

ManyChat displays it. Customer sees a carousel card with images, prices, and buy buttons.

No flow rebuild. No new branches.

Why This Works

1. Memory (ManyChat Can't Do This)

Your bot remembers:

"Customer asked about shipping costs on Tuesday"
"They already bought the Basic plan"
"They clicked 'interested' on premium features"

So it stops asking the same questions twice. It contextualizes recommendations.

2. Scale Without Drowning in Logic

Growing from 50 customer questions to 500? With flows, you need 500 keyword branches.

With SmartBrain, the AI learns. You don't rebuild.

3. Real Product Recommendations

ManyChat flows can send text: "Try the Premium plan for $49."

SmartBrain sends a card:

Product image
Price
Description
"Buy now" button (linked to your Shopify/WooCommerce cart)

Customer doesn't leave Instagram DMs. Conversion stays in-channel.

4. One Brain, Many Channels

ManyChat flows are locked to Messenger/Instagram DMs.

SmartBrain works across:

Messenger
Instagram DMs
WhatsApp Business
Telegram
Custom channels (via JSON API)

Same brain, same memory, different platforms.

For Agencies: The Commercial Angle

If you manage 10 ManyChat bots for clients, you know the pain:

Custom flows for each client = custom maintenance
One client tweaks a FAQ → you risk breaking their flows
Scaling to 50 clients means 50× the complexity

With SmartBrain:

Import existing ManyChat flows (screenshot or JSON export)
Compile them into a knowledge base (we do this, not you)
One deployment per client (5 minutes)
Updates happen server-side (no client flow changes)

You stop being a "ManyChat builder." You become an AI automation agency.

Build time drops from 30 hours → 10 hours per bot.
Scale from 10 → 50 clients in the same sprint.

The Integration Checklist

Export your ManyChat flow (Integrations → Export JSON or screenshot)
Point us to your product catalog (Shopify API, WooCommerce, CSV feed)
Add one External Request webhook (2 minutes, template provided)
Test with a customer message (verify response + routing)
Go live (conversation history + recommendations flow automatically)

Zero code changes. Zero flow rebuilds. Your existing automation keeps working.

The Bottom Line

ManyChat is great at automation. SmartBrain adds intelligence.

You already built the flows. We just make them smarter.

Your bot already gets messages. We make it sell.

Learn more: SmartBrain — the brain for your ManyChat bot

Demo: Book a 15-min walkthrough. See it work on your actual flows.

Ready to scale your agency? Let's talk.

Why Shopify Stores Need a Deterministic Engine, Not a Chatbot

michael fabien — Tue, 14 Jul 2026 09:30:40 +0000

Every Shopify merchant experimenting with AI chat has hit the same wall: the bot confidently recommends a product that's out of stock, quotes a price that expired six months ago, or invents a discount code that doesn't exist. A customer clicks "Buy," hits a 404, and you've lost the sale and the trust.

The problem isn't the chatbot. The problem is architecture.

A Pure-LLM Chatbot Is Guessing at Your Catalog

Large language models are trained on text. They pattern-match. When a customer says "I need a moisturizer under $40," a pure-LLM assistant generates a plausible-sounding answer — but "plausible" is not the same as "accurate." It doesn't know your current inventory. It doesn't know that your best-seller is temporarily out of stock in the medium size. It doesn't know you ran a 20%-off flash sale last Tuesday that has since ended.

The result is hallucinated SKUs, phantom discounts, and product recommendations that simply don't match what's in your Shopify store right now. Every one of those mismatches is a broken customer experience — and in DMs, where expectations are high and patience is short, one broken experience ends the conversation permanently.

How a Deterministic Engine Changes the Equation

SmartBrain works from the opposite direction. Before a single word of copy is generated, the engine queries your live Shopify catalog. It filters by real price, real stock level, real variants (color, size, material — whatever your catalog uses), and the budget the customer stated in the DM.

Only after the engine has selected a verified, purchasable product does the AI copywriter enter the picture — and its only job is to phrase that choice in a compelling, human way. The AI cannot recommend a product that doesn't exist in your catalog. It cannot quote a price that isn't the live Shopify price. It cannot suggest a variant that's sold out.

The output isn't a paragraph of text the customer has to act on themselves. It's a checkout-ready product card — thumbnail, name, real price, real variant selector — with a Buy button that fires inside the DM. Zero redirects to browse. Zero risk of a price mismatch at checkout. The customer taps once and the order flows into Shopify exactly as it would from your storefront.

You Don't Need to Rebuild Your ManyChat Flow

The other fear merchants have when evaluating AI tools is integration cost. Most solutions want you to replace your automation, retrain your team, and re-map every conversation path.

SmartBrain plugs into your existing ManyChat setup as a single block. Your current flows keep running — your welcome sequences, your abandoned cart triggers, your post-purchase check-ins — untouched. SmartBrain activates at the moment a customer expresses product intent ("looking for a gift under $50," "need something for dry skin") and returns a verified recommendation. When the conversation ends, control hands back to your existing flow.

Setup takes one afternoon. Connect your Shopify store, map your ManyChat account, define your business rules (margins to protect, collections to prioritize, budget brackets). After that, every product conversation in your DMs is deterministic by default.

No hallucinated prices. No phantom SKUs. No discount codes that don't exist. Just the right product, at the right price, with a Buy button that actually works.

Ready to turn your DMs into a reliable revenue channel? → askamelie.com

Webhook Magic: Add AI Brain to Your ManyChat Bot Without Rebuilding

michael fabien — Mon, 13 Jul 2026 22:15:14 +0000

Webhook Magic: Add AI Brain to Your ManyChat Bot Without Rebuilding

Your ManyChat bot gets thousands of customer messages every week. But most of them fall through the cracks—wrong keyword match, missing context, flow bottlenecks.

Here's the painful truth: adding intelligence to ManyChat means either:

Rebuilding your entire flow (weeks of work)
Hiring someone to maintain 500+ branches (expensive + fragile)
Giving up and living with 60% unanswered questions

There's a third way: the webhook layer.

The Setup (2 minutes)

In ManyChat, add one External Request block that sends customer context to an external AI:

Customer message
Conversation history (if you've stored it)
Your FAQ/catalog
User attributes (name, plan, order history)

The AI brain replies with:

Smart answer (not just keyword matching)
Suggested next action
Rich cards (photo + price + "Buy" button)
Default fallback for ManyChat's human handoff

ManyChat executes the response—send message, set variable, show menu—without any rebuild.

Why Zero Rebuild?

Your flow stays intact
New flows inherit the brain automatically
Update the AI logic without touching ManyChat
A/B test new flows with the same brain
Scale to 100 bots with one webhook

What the Brain Does

Customer memory: "Hi Sarah, re: your question about the blue widget—we talked about this on June 15. You were interested in the professional tier."

Intent, not keywords: Customer says "this isn't working"—the bot understands it's a complaint, not a question about features.

Catalog intelligence: Knows your products, prices, stock, customer history. Recommends what fits their budget/need.

Rich responses: Send Instagram shopping cards (photo + price) or templated data (shipping cost + ETA) via ManyChat's native card builder.

Fallback logic: If unsure, escalate gracefully. If sure, close the conversation without human handoff.

The Math

Agency example: You manage 20 client bots.

Old way: 2 weeks × 20 clients = 40 weeks = $50K+ dev costs
Webhook way: 2 min × 20 clients = 40 min = zero rebuild

E-commerce example: Your bot handles 10K monthly messages.

Old way: 40% bounce to human (4K escalations = $2K+ monthly support)
Webhook way: 85% auto-resolved (1.5K escalations = $400 monthly support)

How to Start

You need three things:

ManyChat flow (you already have this)
API endpoint that takes customer context → returns smart response
Knowledge (FAQ, catalog, customer data) the API can access

You don't need to rebuild. You don't need to hire. You don't need to understand LLMs.

SmartBrain is the brain for your ManyChat bot. Webhook + AI memory + product intelligence.

Forget maintaining 500 branches. Add one brain, get 85% auto-resolution on day one.

Learn more

The 2 ManyChat Client Questions SmartBrain Finally Answers

michael fabien — Mon, 13 Jul 2026 09:30:42 +0000

If you run a ManyChat agency, you already know the pattern. A new e-commerce client sits across from you, excited about conversational commerce — then the questions start. After dozens of onboarding calls, two come up every single time.

Question 1: "What happens when the bot gives a customer the wrong price?"

Question 2: "Does this integrate with what we already have, or do we have to rebuild everything?"

These aren't bad questions. They're the right questions. And until now, you probably answered them with a mix of reassurance and crossed fingers.

SmartBrain Is a Deterministic Engine, Not a Chatbot

Most AI shopping tools work by feeding a language model your product catalog and hoping it remembers the right numbers. SmartBrain doesn't work that way.

SmartBrain is a deterministic commerce engine with an AI copywriter layered on top. The engine queries your client's live Shopify catalog directly — pulling real product IDs, current prices, actual stock levels, and available variants before it writes a single word. The AI's only job is to phrase what the engine has already selected.

This means SmartBrain cannot hallucinate a price, invent a discount, or suggest an out-of-stock SKU. If a pair of sneakers costs $89 and is available in size 10 only, that's exactly what the customer sees — no exceptions, no drift over time. It also respects the customer's stated budget: if someone says "under $50," the engine filters before it generates.

The output isn't a message that mentions a product. It's a checkout-ready product card with a real Buy button, delivered inside the DM. The customer taps once and lands on a pre-filled cart. Zero friction. Zero errors. Zero refund requests because "the bot said it was on sale."

That's your answer to Question 1.

It Plugs Into the ManyChat Setup You Already Built

Here's what your clients don't want to hear: "We'll need to rebuild your flows from scratch." SmartBrain is designed to plug into an existing ManyChat installation — no rebuild, no migration, no breaking changes to flows that are already converting.

You add SmartBrain as a step inside your client's existing automation. When a subscriber's message triggers a product intent, SmartBrain intercepts, runs the live catalog lookup, and returns the card. Everything else in the flow — welcome sequences, FAQ branches, abandoned cart triggers — stays exactly as you built it.

For agency clients, this matters enormously. They've already paid for their ManyChat setup. They're not approving a rebuild budget before they've seen a single attributed sale.

That's your answer to Question 2.

The White-Label Layer Your Agency Actually Needs

SmartBrain ships with a white-label option built for agencies. Your branding, your client dashboard, your pricing. You're not reselling a tool — you're delivering a capability that looks like proprietary infrastructure you built.

Typical agency workflow: connect the Shopify catalog, activate SmartBrain inside ManyChat, go live in one session. The first attributed sale — tracked back to the DM, SKU-level, timestamped — usually lands within 48 hours.

When your client asks "what's our ROI?" you have a clean answer with no ambiguity.

Add it to your stack and answer both questions before they're asked: askamelie.com