%%{init: {'theme': 'base', 'themeVariables': { 'primaryColor': '#e0f2fe', 'primaryTextColor': '#0f172a', 'primaryBorderColor': '#0284c7', 'lineColor': '#64748b', 'secondaryColor': '#f0fdf4', 'tertiaryColor': '#fff'}}}%%
flowchart TD
Start([Landing Page Form Submission]) -->|Webhook| Router{Make / Zapier Router}
Router -->|Path A| CC[Stripe Checkout]
Router -->|Path B| PP[PayPal Smart Button]
Router -->|Path C| MoMo[Manual MoMo ID Input]
CC -->|checkout.session.completed| Ledger[(Google Sheets Ledger)]
PP -->|Instant Payment Notification| Ledger
MoMo --> Gemini[Gemini API gpt-4o-mini]
Gemini -->|Structured JSON Format| Ledger
Ledger --> StatusCheck{Row Status?}
StatusCheck -->|Paid| Deliver[Auto-Deliver Product via Postmark/Mailgun]
StatusCheck -->|Pending Verification| Alert[Slack/Telegram Admin Alert]
Alert --> Manual[Manual Ledger Check]
Manual -->|1-Click Approval Webhook| Approve[Update Row Status to Paid]
Approve --> Deliver
No-Backend AI Payment Automation Pipeline
An automated pipeline routing Stripe, PayPal, and Mobile Money (MoMo) transactions using Make.com, Gemini, and Google Sheets.
This guide provides a comprehensive blueprint for building a scalable, completely serverless payment verification and logging infrastructure. The system leverages automated API webhooks for card payments alongside an AI agent to clean, parse, and flag manual Mobile Money transactions.
1 System Architecture
2 Step-by-Step Implementation Guide
2.1 Step 1: Capture Front-End Ingestion Data
2.1.1 Infrastructure Selection
Utilize form frameworks that support logic-routing and native webhook outputs (e.g., Tally.so, Typeform, or embedded Framer Forms).
2.1.2 Form Configuration Properties
Configure fields strictly adhering to the following structure:
customer_name(Text)customer_email(Email, Required)product_tier(Dropdown/Radio)payment_method(Dropdown:Stripe,PayPal,Mobile Money)momo_transaction_id(Conditional Text Field: Visible only ifpayment_methodequalsMobile Money).
2.1.3 The Ingestion Trigger
Set up a POST Request webhook mechanism on your form builder. Route the JSON payload directly to a custom webhook URL generated inside your Make.com or Zapier automation scenario.
2.2 Step 2: Establish the Routing Logic
Add a Router Module directly after your webhook listener inside Make.com to orchestrate workflow splitting:
2.2.1 Path A: Credit Card / Standard (Stripe)
Generate and serve a hosted Stripe Checkout Link. Pause execution down this path until Stripe shoots a checkout.session.completed event to your handler.
2.2.2 Path B: Digital Wallet (PayPal)
Deploy a PayPal smart button on your site. Configure the automation to wait for an Instant Payment Notification (IPN) webhook payload matching the customer’s email.
2.2.3 Path C: Mobile Money Manual Settlement (MoMo)
When a customer inputs an arbitrary manual receipt ID string, immediately catch the payload and route it straight to the Gemini integration module.
2.3 Step 3: Implement AI Structuring & Data Validation
Manual inputs introduce high error margins (copy-paste truncated digits, localized telecom prefixes, or malicious fraud strings). You must sanitize this data before logging it.
2.3.1 Module Configuration
- Model:
gemini-3.5-flash - Response Format:
json_object
2.3.2 System Prompt Configuration
You are a specialized financial operations parser. Analyze the following user-submitted Mobile Money (MoMo) transaction text payload.
[USER SUBMISSION]: "{{Step1.momo_transaction_id}}"
Tasks:
1. Isolate and extract the alphanumeric transaction ID. Remove spaces, dashes, or extraneous terms (e.g., "ID:", "TXN", "Ref").
2. Validate if the string profile aligns with standard network structures (e.g., MTN MoMo typically matches 10-11 digits or specific alphanumeric profiles, Orange Money uses unique string lengths).
3. Guess the carrier network based on known structural footprints.
Output a strictly valid JSON object complying with this schema:
{
"is_valid_format": boolean,
"cleaned_id": "string",
"carrier_guess": "string"
}
2.4 Step 4: Populate the Visual Database Ledger
Create a Google Sheet (or Airtable Base) to perform the role of your system ledger. Map properties meticulously to chronological rows:
| Column Name | Value Source | Data Type | Notes |
|---|---|---|---|
| Timestamp | {now} |
DateTime | System date |
| Customer Email | Form Submission | String | Primary identity key |
| Payment Method | Form Submission | String | Stripe / PayPal / MoMo |
| Raw Input ID | Form Submission | String | Preserved for auditing |
| Cleaned ID | Gemini Output | String | Evaluated string |
| Status | Conditional Router | String | Paid (Stripe/PayPal) OR Pending Verification (MoMo) |
2.5 Step 5: Configure Notifications and Product Delivery
graph TD
A[Log Row in Google Sheets] --> B{Check Status}
B -- Paid --> C[Trigger Postmark/Mailgun API]
C --> D[Instantly Deliver Product Link]
B -- Pending Verification --> E[Webhook to Slack/Telegram]
E --> F[Admin Checks Phone Ledger]
F --> G[Click 'Approve' Button via Webhook]
G --> H[Update Sheet to Paid & Send Product]
2.5.1 Automatic Delivery Path (Paid)
If Status evaluates directly to Paid, route downstream directly to a Mailgun or Postmark module to fire out transactional product assets.
2.5.2 Human-in-the-Loop Verification Path (Pending Verification)
Create a notification module targeting an internal Slack or Telegram endpoint.
Structure the alert payload clearly: > ⚠️ New Manual MoMo Payment Verification Required > * User:
{Email}> * Declared ID:{Cleaned_ID}> * Inferred Carrier:{Carrier_Guess}> >Append an active interactive webhook button in the notification labeled “Approve Transaction”.
Clicking this button sends a payload back to a secondary execution hook in Make.com, which targets the corresponding Google Sheet row, shifts
StatustoPaid, and releases the delivery email.
3 Technology Stack Matrix
| Component | Recommended Tool | Core Architectural Justification |
|---|---|---|
| Automation Engine | Make.com | Substantially more cost-efficient handling of data routers, conditional filters, and loops than Zapier. |
| Payment Gateways | Stripe & PayPal | Eliminates custom endpoints; offloads security and hosting directly to secure payment pages. |
| AI Parsing Layer | Gemini API (gpt-4o-mini) |
Fast inference speeds, exceptionally cheap unit economics for token passing, supports absolute structural JSON integrity. |
| No-Code Database | Google Sheets / Airtable | Easily interfaces with modern integrations; simplifies visual inspection for support and operations teams. |
4 Operational Risk Management
Because users can key in fictitious sequences to attempt access, never automate automatic fulfillment for manual Mobile Money strings. Retain strict structural boundary blocks: keep them pinned at Pending Verification until an operator manually checks their physical phone ledger balance.
Network latency or transient upstream API lockouts can kill webhooks. Ensure that Queueing / Advanced Error Handling is checked inside Make.com settings. This saves failed data points into a dead-letter queue rather than dropping customer contexts completely.
Ensure your landing page terms of service communicate that payment references are safely processed via encrypted third-party validation layers to conform with localized consumer privacy frameworks.