From Zero to Building AI Apps in One Evening

Hey there! Roy here. Last night, we ran a workshop that changed how 200+ people think about building AI applications. By the end of our session, folks who’d never written a line of code were building news aggregators, job matching systems, and crypto trading platforms.

Here’s a recording of our Workshop: Please watch while following along below.

The Setup: What You’ll Need (10 Minutes)

Before we dive into the chaos of my workshop (yes, my internet cut out three times), let’s get you set up:

1. N8N Account: Go to n8n.io, create a free account. You get 5 workflows and unlimited executions.
2. OpenAI API Key: Visit platform.openai.com, sign in, go to API Keys on the left sidebar, create a new secret key. You get $5 free credits.
3. Lovable Account: Sign up at lovable.dev. You get 35 credits daily on the free plan.
4. Claude: Visit Claude.ai and sign up for a free account.

Project 1: The Saudi Gazette News Filter That Actually Works

During the workshop, I wanted to show everyone how to build something immediately useful. So we created an AI system that reads Saudi news, filters for topics you care about, and emails you summaries. Here’s exactly how we built it:

Building Your First N8N Workflow (30 Minutes)

Step 1: Create Your Workflow Canvas

• Open N8N, click “New Workflow”
• You’ll see an empty canvas with just a trigger node

Step 2: Set Your Trigger

• Keep the default “Manual” trigger for testing
• This means you’ll click “Execute Workflow” to run it
• (Later you can change this to run daily at midnight)

Step 3: Add Your News Source

• Click the + button, search for “RSS Read”
• Connect it to your trigger by dragging from one node to the other
• For URL, I used: https://saudigazette.com.sa/rss.xml
• Click “Execute” – you should see 50+ articles appear

Step 4: The Critical Limit Node

• Add a “Limit” node between RSS and the next step
• Set “Max Items” to 5
• This is crucial – without it, you’ll process (and pay for) 50+ API calls
• I learned this the hard way when I sent myself 50 emails

Step 5: Connect Your AI Brain

• Add “Basic LLM Chain” node
• Click the sub-node area and add “OpenAI Chat Model”
• Under credentials, click “Create New”
• Paste your OpenAI API key
• Choose model: gpt-3.5-turbo (cheapest option)
• In the prompt field, select “Define below”
• Paste this prompt: Your job is to read through the content and only focus on news related to food. If the article is not related to food, reply with one word: “irrelevant” If related, summarize in 2-3 sentences. Include only facts. Title: {{$json.title}} Content: {{$json.content}} Link: {{$json.link}}

Step 6: Add Smart Filtering

• Add “Edit Fields” node after the LLM
• Set to “Manual Mapping”
• Add these fields:
  • title: {{$('Limit').item.json.title}}
  • link: {{$('Limit').item.json.link}}
  • summary: {{$json.text}}
  • date: {{$('Limit').item.json.pubDate}}

Step 7: The Decision Point

• Add “IF” node
• Set condition: {{$json.summary}} “is not equal to” irrelevant
• This splits your flow: relevant articles go to true, irrelevant to false

Step 8: Send Your Curated News

• Add “Gmail” node to the “true” branch
• Connect your Gmail (it’ll open OAuth window)
• Set recipient to your email
• Subject: “Food News Alert”
• Message:

{{$json.summary}}

Read more: {{$json.link}}
Published: {{$json.date}}

Step 9: Test It

• Click “Execute Workflow”
• Watch it process 5 articles
• Check your email – you should have 1-3 relevant articles

During the workshop, when we switched from “AI” to “food” as our topic, we got 3 articles about restaurant closures and food safety violations in Saudi Arabia – all published that day. The magic moment was watching 50 articles get filtered down to just the 3 that mattered.

Project 2: The Resume Reality Check (The Top-Down Revolution)

Here’s where I blew everyone’s minds. Instead of building node by node, we used Claude to generate an entire workflow. Watch this:

The Claude Workflow Generator Method

Step 1: Open Claude and paste this exact prompt:

Create an N8N workflow that:
- Allows me to upload a file (my resume)
- Has it reviewed by an AI agent who is a professional recruiter
- Has it reviewed by a traditional applicant tracking system
- Gives me a summary of both reports
- Then sends both reports to a job coach AI assistant
- Who will review my resume and make specific recommendations
- Give me an interface where I can make changes to my resume and resubmit

Your output should be copy-pasteable directly into N8N with nothing changed.
You are also a professional tech auditor. Audit your own response to ensure the workflow will work and is properly connected.

Step 2: Claude’s Response Claude generated a complete workflow with 15+ nodes including webhook triggers, multiple AI reviewers, merge nodes, and feedback loops. Here’s what to do with it:

Step 3: Import to N8N

• Copy Claude’s JSON output
• In N8N, create new workflow
• Click the three dots menu → Import from File → Paste the JSON
• You’ll see the complete workflow appear

Step 4: Fix the Red X’s Each red X needs fixing. Here’s how:

• Click each OpenAI node
• Add your API credentials
• For each prompt field that’s empty, ask Claude: “What’s the prompt for the recruiter review node?”
• Claude will give you prompts like:

You are an experienced technical recruiter. Review this resume:
{{$json.resumeText}}
Provide feedback on:
1. Overall presentation
2. Keyword optimization
3. Achievement quantification
4. Missing skills for target roles

Step 5: Connect the Webhook

• The webhook node shows a URL like: https://your-n8n.app/webhook/abc123
• Copy this URL – you’ll need it for your Lovable interface

This approach got us 80% complete in 30 seconds versus 30 minutes of manual building.

Project 3: Building the Ka.nz Recruiting Platform

Now for the grand finale – building a complete application like our Ka.nz platform. This is where Lovable shines.

The Professional Product Spec Method

Step 1: Brain Dump Your Idea I started with this messy note:

Recruiting app
Recruiter types: "find me civil engineers with 10+ years experience building skyscrapers in Saudi"
Shows top candidates instantly
Can schedule interviews
Gives interview questions
Tracks everything

Step 2: Transform It With Claude Paste this prompt along with your brain dump:

You are a professional product manager and software developer.
Your specialty is working with Lovable.
Write a complete spec that Lovable can ingest to build an app with the fewest number of prompts.
I want the app to be complete and polished.
Fill in the blanks where detail is needed.
Use best UX practices.
Let the color palette follow green tones.

[Paste your brain dump here]

Step 3: Claude Creates Your Product Bible Claude generated 20 pages including:

• Brand identity (colors: #10B981 primary, #F3F4F6 background)
• Typography (Inter font family)
• Component library specs
• Page layouts
• Authentication flow
• Database schema
• API endpoints

Step 4: Feed Lovable

• Open Lovable, create new project
• Paste Claude’s entire spec
• Lovable responds: “This is comprehensive. I’ll build in this order: 1. Design system, 2. Landing page, 3. Authentication, 4. Dashboard…”
• Type: “Start”

Step 5: Watch the Magic Over 5-10 minutes, Lovable builds:

• Complete design system
• Landing page with Arabic/English toggle
• Sign up/login with Google OAuth
• Dashboard with job management
• Candidate search interface
• Interview scheduling
• Payment integration

Step 6: Connect to N8N Remember that webhook URL from the resume workflow?

• In Lovable: “Add file upload to dashboard that posts to [webhook URL]”
• Lovable adds the upload component
• When users upload resumes, N8N processes them
• Results return to your Lovable app

The Integration: Making It All Work Together

During the workshop, we connected everything:

1. Lovable Frontend → User uploads resume
2. N8N Webhook → Receives file
3. AI Processing → Reviews resume
4. Database → Stores results
5. Lovable Display → Shows recommendations

The actual connection code Lovable generates:

javascript

const handleFileUpload = async (file) => {
  const formData = new FormData();
  formData.append('file', file);
  
  const response = await fetch('https://your-n8n.app/webhook/abc123', {
    method: 'POST',
    body: formData
  });
  
  const results = await response.json();
  setRecommendations(results);
};

The Reality Check Moments

Let me share what actually happened during the workshop:

The 50 Email Disaster: I forgot to connect Gmail AFTER the IF node, so it tried to send an email for every article. Always double-check your connections.

The PIN Feature Discovery: After running the workflow three times and burning through API credits, we discovered the PIN feature. Click the three dots on any node → PIN. This saves the output so you don’t re-run expensive operations.

The Internet Crashes: My connection dropped three times. Each time, participants kept building because once you understand the pattern, you don’t need hand-holding.

Your Turn: Tonight’s Assignment

1. Hour 1: Build the news filter exactly as I described. When it works, change “food” to your interest.
2. Hour 2: Use Claude to generate a workflow for your own problem. Maybe expense tracking, meeting scheduling, or content creation.
3. Hour 3: Create a simple Lovable app. Start with: “Build a personal dashboard that tracks my [whatever you care about]”

The Bottom Line

Six months ago, what we built would have cost $50,000 and a dev team. Last night, 50+ non-coders built functional AI applications. Not prototypes – actual working apps. One participant’s pet emergency advisor is already getting users. Another’s crypto portfolio tracker is being used by their investment club.

The barriers haven’t lowered – they’ve disappeared. The only question is: what problem in your life will you solve first?