Guidebook

Back

THE COMPLETE AGENTIC AI & AUTOMATION
COMPANY GUIDEBOOK

AGENTIC AI &

AUTOMATION

COMPANY GUIDEBOOK

Workflow Automation • Agentic AI • Vibe Coding

From Zero Knowledge → Running Company

Practical Edition | 2025

📌 What This Guidebook Covers

CHAPTER 0

Your Master Roadmap

The complete path from beginner to automation agency owner

Before diving into individual chapters, internalize this end-to-end learning path. Each phase builds on the previous one. Do NOT skip phases — the foundation is what separates freelancers who charge ₹10,000 from agencies charging ₹2,00,000+.

| PHASE | WHAT TO DO / LEARN |

Phase 1 (Week 1-2) Business knowledge, workflow architecture thinking, APIs, JSON basics
Phase 2 (Week 3-4) n8n mastery — every node type, triggers, HTTP, OpenAI node
Phase 3 (Week 5-6) Make.com + Gumloop + templates library — build reusable systems
Phase 4 (Week 7-8) Agentic AI — LangChain agents, CrewAI multi-agent systems
Phase 5 (Week 9-10) Vibe Coding — Cursor, Lovable, Bolt, v0 for client dashboards
Phase 6 (Week 11-12) Build demos, choose niche, close first client
Phase 7 (Month 4+) Scale: retainers, SaaS products, hire team

🔥 The One Mental Model That Changes Everything

Stop thinking like a tool user. Start thinking like a SYSTEMS ARCHITECT.

Every business problem = an input → process → output flow.

Your job is to automate that flow using AI + automation tools.

The more businesses you understand, the more problems you can solve.

TOOLS ARE INTERCHANGEABLE. Business thinking is your moat.

CHAPTER 1

Foundation Knowledge

What you MUST know before touching any tool

1.1 — Why Foundation Comes First

Most beginners jump straight into n8n tutorials. They build one workflow, get stuck, and quit. The ones who succeed spend 2 weeks understanding HOW systems think before building anything. This chapter is your competitive advantage.

1.2 — Workflow Architecture Thinking

Workflow architecture means understanding how information flows through a system — from trigger (something happens) to actions (things that happen because of it) to output (the final result delivered to a person or system).

The 3-Layer Architecture Model

| LAYER | WHAT IT MEANS |

Layer 1: Triggers What starts the workflow? Form submit, webhook, schedule, email received, CRM update
Layer 2: Logic & AI IF conditions, data transformation, AI classification, enrichment, decision trees
Layer 3: Actions Send WhatsApp, create CRM record, book calendar, send email, update sheet

Practical Exercise: Draw These 3 Workflows on Paper

💡 Pro Tip: The Paper-First Rule

NEVER open n8n or Make.com before drawing your workflow on paper first.

Draw boxes for each step. Draw arrows showing data flow.

Write what data enters each box and what leaves it.

Only when you can explain it to someone else on paper — THEN open the tool.

1.3 — APIs and JSON: The Language of Automation

APIs (Application Programming Interfaces) are how software applications talk to each other. Every modern tool — WhatsApp, Stripe, Google Calendar, HubSpot, GPT-4 — exposes an API. When you send a webhook or make an HTTP request, you are calling an API.

API Concepts You Must Know

JSON — The Universal Data Language

JSON (JavaScript Object Notation) is how data is structured in almost every API call. If you cannot read and write JSON, you cannot do automation professionally. 

// Example JSON: A lead that comes in from a form
{
"name": "Rahul Sharma",
"email": "rahul@gmail.com",
"phone": "+91-9876543210",
"source": "Facebook Ad",
"budget": 50000,
"interested_in": "Real Estate",
"timestamp": "2025-01-15T10:30:00Z"
}
// Accessing nested data
// In n8n: {{ $json.name }} → Rahul Sharma
// In n8n: {{ $json.budget }} → 50000

JSON Practice Tasks

1.4 — Business Process Knowledge (CRITICAL MUST-HAVE)

This is the skill 90% of automation builders completely ignore — and it is exactly why they stay stuck charging low rates. If you understand how a business actually runs, you will spot automation opportunities immediately in any sales call.

Core Business Processes You Must Understand

| BUSINESS AREA | KEY PROCESSES TO KNOW |

Sales & Lead Gen Lead capture → qualification → follow-up → proposal → close → onboarding
Marketing Funnels Ad click → landing page → opt-in → email sequence → offer → sale
Operations Order → fulfillment → delivery → customer service → review → repeat
Customer Success Onboarding → check-in calls → milestone tracking → renewal → upsell
Finance Invoice → payment → reconciliation → reporting → tax preparation
HR & Hiring Job post → applicant tracking → interview scheduling → offer → onboarding

The 5-Question Business Audit (Use in Every Client Call)

🏆 Industry Deep-Dives: Know These 4 Industries Cold

CHAPTER 2

n8n Mastery

Every node, every workflow, real data

2.1 — Why n8n is Your Primary Tool

n8n is open-source, self-hostable, and has native AI/LLM nodes built in. Unlike Zapier (expensive per task) or Make.com (limited AI), n8n lets you build enterprise-grade agentic workflows for nearly free when self-hosted. Master this tool first.

📺 YouTube Learning Resource

2.2 — n8n Installation (Get Running in 10 Minutes)

Option A: Local (for learning)

# Install Node.js first (v18+), then:
npm install n8n -g
n8n start
# Open browser → http://localhost:5678

Option B: Cloud (for client work)

2.3 — Every Node Type Explained

n8n has 400+ nodes. Here are the ones you will use for 90% of real work:

TRIGGER NODES (How Workflows Start)

| NODE | WHAT IT DOES + USE CASE |

Webhook Trigger Receives HTTP POST/GET requests. Use: Form submissions, payment notifications, CRM webhooks. URL: yourdomain.com/webhook/abc123
Schedule Trigger Runs on cron schedule. Use: Daily reports, weekly emails, hourly data sync. Set: Every day at 9am, every Monday, every 1 hour
Email Trigger (IMAP) Watches inbox for new emails. Use: Auto-reply system, lead from email, support ticket creation
Telegram Trigger Activates when bot receives message. Use: Internal team bot, customer support bot
RSS Feed Trigger Monitors RSS feed for new content. Use: News monitoring, content aggregation

ACTION NODES (Do Things)

| NODE | WHAT IT DOES + USE CASE |

HTTP Request Make ANY API call to ANY service. THE most important node. Use: WhatsApp API, custom APIs, web scraping
Send Email (SMTP) Send emails via Gmail, Outlook, SendGrid. Use: Lead follow-up, notifications, reports
Google Sheets Read/write spreadsheet data. Use: CRM tracking, lead lists, reporting
Airtable Database operations. Use: Project tracking, CRM, inventory
HubSpot / Salesforce Native CRM integrations. Use: Create contacts, update deals, log activities
Telegram Send messages, files, photos via Telegram bot. Use: Team alerts, client updates
Slack Send messages to channels. Use: Internal alerts, lead notifications
Cal.com / Calendly Book appointments. Use: Auto-booking when lead qualifies
WhatsApp (via API) Send WhatsApp via Twilio/WATI/360Dialog. Use: Lead nurture, order updates

LOGIC & FLOW NODES (Make Decisions)

| NODE | WHAT IT DOES + USE CASE |

IF Node Branch based on condition. Use: IF budget > 50000 → send to sales team, ELSE → send nurture email
Switch Node Multiple branches (like IF-ELSE IF). Use: Route leads by city, by product, by source
Merge Node Combine data from multiple paths. Use: Merge lead data + enrichment data + AI analysis
Split In Batches Process large lists in chunks. Use: Send 1000 emails without overwhelming API
Wait Node Pause workflow for time/event. Use: Wait 2 hours before follow-up, wait for webhook response
Code Node (JS) Write JavaScript for custom logic. Use: Complex data transformation, calculations, string manipulation
Set Node Create/modify data fields. Use: Rename fields, set default values, build new JSON objects
Filter Node Remove items that don't match. Use: Only process leads from India, only orders above ₹500

AI NODES (The Money Nodes)

| NODE | WHAT IT DOES + USE CASE |

OpenAI Node Direct access to GPT-4/GPT-4o. Use: Lead classification, email writing, sentiment analysis
AI Agent Node Full autonomous agent with tools. Use: Research agent, customer service agent
Chat Memory Store conversation history. Use: Multi-turn chatbot that remembers context
Vector Store Store + search embeddings. Use: Knowledge base, document Q&A, FAQ bot
Document Loader Load PDF/Word/text files. Use: Train custom knowledge base
Text Splitter Break long text into chunks. Use: Prepare documents for vector storage
Embeddings Convert text to vectors. Use: Semantic search, similarity matching
Output Parser Force AI to return structured JSON. Use: Get reliable data from AI instead of free text

2.4 — The HTTP Node: Deep Dive

The HTTP Request node is your Swiss Army knife. ANY service with an API can be connected using this node. You do NOT need a dedicated n8n integration — if it has an API, you can connect it. 

// Example: Sending a WhatsApp message via WATI API
Method: POST
URL: https://live-server.wati.io/api/v1/sendTemplateMessage
Headers:
Authorization: Bearer YOUR_WATI_API_KEY
Content-Type: application/json
Body (JSON):
{
"template_name": "lead_followup",
"broadcast_name": "lead_{{ $json.id }}",
"receivers": [
{
"whatsappNumber": "{{ $json.phone }}",
"customParams": [
{ "name": "name", "value": "{{ $json.name }}" }
]
}
]
}

2.5 — n8n Expressions: Dynamic Data

Expressions let you use data from previous nodes dynamically. Everything inside {{ }} is evaluated as code.

| EXPRESSION | WHAT IT DOES |

{{ $json.email }} Gets the 'email' field from current item
{{ $json.name.toUpperCase() }} Converts name to uppercase
{{ $now.toISO() }} Current timestamp in ISO format
{{ $node['Set'].json.data }} Gets data from a specific node named 'Set'
{{ $items().length }} Count of items in current batch
{{ $json.price * 1.18 }} Calculates price with 18% GST
{{ $json.phone.replace('+91', '') }} Removes country code from phone

2.6 — 5 Workflows to Build Right Now

Workflow 1: Lead Capture → AI Qualify → WhatsApp

Workflow 2: Daily Sales Report to WhatsApp/Telegram

Workflow 3: E-commerce Order → WhatsApp Updates

Workflow 4: AI Customer Support Bot

Workflow 5: Content Repurposing Automation

⚡ n8n Pro Tips

CHAPTER 3

Other Automation Platforms

Make.com, Gumloop, Zapier & Templates

3.1 — Platform Comparison

| PLATFORM | BEST FOR + PROS/CONS |

n8n BEST: Complex AI workflows, self-hosted. PRO: Free, flexible, AI native. CON: Steeper learning curve
Make.com BEST: Visual workflows, client-friendly. PRO: Beautiful UI, 1000+ apps. CON: Costs per operation
Gumloop BEST: AI-first automation, LLM workflows. PRO: Built for AI, simple. CON: Newer, fewer integrations
Zapier BEST: Quick simple automations. PRO: Easiest to use, most integrations. CON: Expensive at scale
Activepieces BEST: Open-source Zapier alternative. PRO: Free, self-hosted. CON: Fewer nodes than n8n

3.2 — Make.com (Formerly Integromat)

Key Concepts in Make.com

Make.com vs n8n: When to Use Which

Use Make.com WHEN: Client prefers visual tool, budget allows, no need for AI agents

Use n8n WHEN: AI agents needed, self-hosting required, complex logic, cost-sensitive

Use BOTH: Offer Make.com setup as a service, migrate complex ones to n8n for cost savings

3.3 — Gumloop

Gumloop is built specifically for AI-first automation. Think of it as n8n but designed from the ground up for LLM workflows. It shines when you need to chain multiple AI calls together with document processing.

Best Gumloop Use Cases

3.4 — Template Library: Your Revenue Multiplier

Templates are pre-built workflows you can sell or deploy repeatedly. A good template library is the difference between a freelancer and an agency. Build these templates and sell them as products OR as included services.

Must-Build Templates

| TEMPLATE NAME | WHAT IT DOES + TARGET CLIENT |

Real Estate Lead Machine Facebook Ad → Form → AI Qualify → WhatsApp → Calendar booking. Sell to: real estate agents
E-commerce WA Bot Order → WhatsApp updates → review request → win-back. Sell to: Shopify/WooCommerce stores
Coaching Sales System Webinar signup → Email sequence → Call booking → Enrollment. Sell to: coaches, consultants
AI Customer Support Bot WhatsApp bot that answers FAQs, looks up orders, escalates to human. Sell to: any business
Daily Report Automation Collect data from multiple sources → AI summary → send to owner. Sell to: any business
Appointment Reminder Upcoming bookings → WhatsApp reminder 24h + 1h before. Sell to: clinics, salons, gyms
Lead Nurture Sequence Cold lead → 7-day WhatsApp + email sequence → re-engage. Sell to: sales teams
Social Media Repurposer Blog/YouTube → AI rewrite → schedule on LinkedIn/Instagram. Sell to: content creators

💰 Template Monetization Strategy

CHAPTER 4

Agentic AI

Build AI agents that work autonomously

4.1 — What is an AI Agent?

An AI agent is not just a chatbot. An agent can: (1) Understand a goal, (2) Decide which tools to use to achieve it, (3) Execute those tools, (4) Observe the result, (5) Decide the next step — all without human intervention.

| CHATBOT (OLD) | AI AGENT (NEW) |

Follows fixed script Decides actions dynamically
Cannot use tools Can call APIs, search web, write code
One-shot response Multi-step reasoning and execution
Forgets context Maintains memory across sessions
You define every step Agent decides its own steps

4.2 — The Agent Architecture

Core Components

4.3 — LangChain: The Agent Framework

LangChain is the most widely used framework for building AI agents. It provides abstractions for tools, memory, chains, and agents that work with any LLM. 

# Install LangChain
pip install langchain langchain-openai
from langchain_openai import ChatOpenAI
from langchain.agents import create_react_agent, AgentExecutor
from langchain.tools import Tool

# Define tools the agent can use

tools = [
Tool(name='search_crm', func=search_crm_function, description="Search CRM for customer data. Input: customer name or email"),
Tool(name='send_whatsapp', func=send_whatsapp_function, description="Send WhatsApp message. Input: phone number and message"),
]

# Create the agent

llm = ChatOpenAI(model='gpt-4o', temperature=0)
agent = create_react_agent(llm, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

# Run the agent

result = agent_executor.invoke({"input": "Qualify this lead..."})

4.4 — CrewAI: Multi-Agent Systems

CrewAI lets you build teams of specialized AI agents that collaborate. Think: one agent researches a lead, another writes a personalized email, another schedules a call — all working together automatically. 

from crewai import Agent, Task, Crew
lead_researcher = Agent(role='Lead Research Specialist', goal='Research company', tools=[web_search], llm=llm)
outreach_writer = Agent(role='Personalization Expert', goal='Write cold email', tools=[email_tool], llm=llm)
research_task = Task(description='Research {company} pain points', agent=lead_researcher)
outreach_task = Task(description='Write personalized cold email', agent=outreach_writer)
crew = Crew(agents=[lead_researcher, outreach_writer], tasks=[research_task, outreach_task])
crew.kickoff(inputs={'company': 'Tata Motors'})

4.5 — Building an AI Sales Agent (Full Example)

What This Agent Does

Building It in n8n (No-Code Approach)

🤖 Agent Use Cases by Industry

CHAPTER 5

Vibe Coding

Build full apps using AI prompts — zero coding required

5.1 — What is Vibe Coding?

Vibe Coding means building software applications by describing what you want in plain English (or Hindi) to an AI, and the AI writes the code for you. You focus on WHAT to build. The AI handles HOW.

This completely changes the game for automation agencies — you can now offer custom dashboards, CRM systems, client portals, and SaaS products WITHOUT hiring developers.

| OLD WAY | VIBE CODING WAY |

Hire developer (₹50,000–₹2,00,000) Use AI to build in hours
Weeks to build MVP Hours to first working version
Pay for every change Prompt to iterate instantly
Technical knowledge required Business thinking is enough
Limited by developer capacity Build 10x more deliverables

5.2 — The Vibe Coding Toolkit

| TOOL | BEST FOR + HOW TO USE |

Cursor Best for full-stack apps. AI pair programmer inside VS Code. Use for React, Node.js, Python apps
Lovable.dev Best for web apps fast. Describe app in chat → get deployable React app in minutes
Bolt.new Best for instant prototypes. Browser-based, no install. Great for client demos same-day
v0 by Vercel Best for UI components. Describe the UI → get clean React/Tailwind code instantly
Replit AI Best for backend + APIs. Build and host Python/Node backends with AI assistance
GitHub Copilot Best for existing codebase. Autocompletion and chat inside any editor

5.3 — Cursor: The Professional's Tool

Cursor is VS Code with Claude/GPT built in. This is your primary tool for building production-quality client projects. Here is how to use it effectively:

Cursor Keyboard Shortcuts

The .cursorrules File: Your Secret Weapon

Create a .cursorrules file in your project root to give Cursor context about your project. This makes every prompt 10x more accurate. 

// .cursorrules file example for a CRM project
You are building a CRM system for a real estate agency.
Tech stack: React (Vite), Tailwind CSS, Supabase for database, n8n for automation.
All UI must be in Indian English. Currency in INR.
Database has these tables: leads, contacts, properties, deals, activities.
Always write TypeScript. Use shadcn/ui components.
Every form must have validation. Every table must be sortable and filterable.

5.4 — Vibe Coding Prompts: In-Depth Examples

The quality of your output depends entirely on the quality of your prompt. Here are production-ready prompt examples:

Prompt Example 1: Real Estate CRM Dashboard

📝 FULL PROMPT — Build a Real Estate CRM dashboard using React, Tailwind CSS, and mock data.

PAGES NEEDED:

DESIGN: Professional, clean theme, sidebar navigation, responsive.

Prompt Example 2: Automation Agency Website

📝 FULL PROMPT — Build a professional landing page for an AI automation agency targeting Indian businesses.

SECTIONS:

Prompt Example 3: WhatsApp Chatbot UI Builder

📝 FULL PROMPT — Build a drag-and-drop chatbot flow builder for WhatsApp automation.

FEATURES:

Use React Flow library for the canvas. Tailwind for styling.

Prompt Example 4: Iterating on Existing Code

📝 ITERATION PROMPT — The leads table currently shows all leads in a flat list.

CHANGES NEEDED:

5.5 — The Lovable.dev Workflow

5.6 — Bolt.new for Rapid Prototyping

Bolt.new runs entirely in your browser — no installs. Perfect for same-day client demos. Open bolt.new, describe your app, get a working prototype in 15 minutes.

🚀 Vibe Coding Best Practices

CHAPTER 6

Starting Your Company

From zero to first paying client

6.1 — The 4-Step Launch Sequence

| STEP | WHAT TO DO |

Step 1: Learn (Month 1-3) Follow this guidebook. Build all practice projects. Get hands dirty daily.
Step 2: Build Demos (Month 2-3) Build 3 industry demos: Real Estate, E-commerce, Coaching. Record Loom videos.
Step 3: Choose Niche (Month 3) Pick ONE industry to focus on first. Depth > breadth in early stage.
Step 4: Get Clients (Month 3+) Daily outreach: LinkedIn DMs, WhatsApp cold outreach, referrals, content

6.2 — Choosing Your Niche

Niching down feels scary but it is the fastest path to clients. 'I do automation for real estate agents' gets more meetings than 'I do automation for all businesses'.

6.3 — Outreach System (Get Clients Without Ads)

LinkedIn Outreach (B2B)

WhatsApp Cold Outreach (Local Businesses)

Content-Led Inbound (Long Term)

6.4 — The Discovery Call: Closing ₹50,000+ Deals

Call Structure (20-30 Minutes)

💬 Closing Language That Works

"Based on what you told me, you're losing about [X] leads per month to manual follow-up. If we automate this, you'll recover those leads. That alone pays for the system in [X] weeks."

"I'll build the first workflow and show you the output. If you love it, we proceed. If not, you owe me nothing. Fair?"

"Most clients in real estate see ROI within 30 days — either recovered deals or saved time. My setup fee is ₹X."

6.5 — Pricing Your Services

| SERVICE | PRICING RANGE |

Single workflow automation ₹10,000 – ₹30,000 setup
Complete lead automation system (3-5 workflows) ₹30,000 – ₹75,000 setup
Agentic AI chatbot (WhatsApp/web) ₹50,000 – ₹1,50,000 setup
Custom CRM + automations (vibe coding + n8n) ₹75,000 – ₹2,00,000 setup
Full automation audit + build for enterprise ₹1,50,000 – ₹5,00,000+
Monthly retainer (maintenance + new workflows) ₹10,000 – ₹50,000/month
Sell templates (digital products) ₹2,000 – ₹25,000 per template

📌 Pricing Psychology

CHAPTER 7

Scaling to Agency & SaaS

From freelancer to systems-driven business

7.1 — Phase 1: Stable Freelancer (₹1–3L/month)

7.2 — Phase 2: Small Agency (₹5–15L/month)

7.3 — Phase 3: SaaS Product (₹20L+/month potential)

The highest leverage move: turn your best workflow template into a hosted SaaS product. Instead of building custom every time, clients pay a monthly subscription.

SaaS Ideas from Automation Services

🎯 Final 90-Day Action Plan

The automation economy is not coming — it is here.

The only question is whether you build it or get replaced by it.

Now go build. 🚀