Building Your First Automation
Hands-on walkthrough: build a real AI-powered workflow.
Building Your First Automation
Time to get hands-on. We'll walk through building a practical AI-powered automation that you can adapt for your own business.
The "Smart Lead Processor" Workflow
This automation takes new form submissions, uses AI to analyze and score them, creates CRM records, and drafts personalized follow-ups.
What You'll Build
- 1.Trigger: New form submission (Typeform, Google Forms, or any form tool)
- 2.AI Step 1: Classify and score the lead
- 3.AI Step 2: Draft a personalized response
- 4.Action: Create a record in your CRM / Google Sheet
- 5.Action: Send the drafted email (or queue for human review)
- 6.Action: Notify your team on Slack
Step-by-Step in Zapier
Step 1: Set Up the Trigger
- •Choose your form tool (Typeform, Google Forms, Jotform, etc.)
- •Select trigger: "New Form Submission"
- •Connect your account and test with a sample submission
Step 2: AI Classification
Add a Zapier AI action:
- •Action: "AI Text"
- •Prompt:
Analyze this form submission and provide:
1. Lead Score (1-100): Based on buying intent, company size, and urgency
2. Category: hot-lead, warm-lead, cold-lead, or not-a-lead
3. Key interests: What products/services are they interested in?
4. Recommended next step: What should our sales team do?
5. One-sentence summary: For the CRM notes field
Form data:
Name: {{name}}
Company: {{company}}
Email: {{email}}
Message: {{message}}
Budget: {{budget}}
Respond in JSON format.Step 3: AI Email Draft
Add another AI action:
- •Prompt:
Draft a personalized follow-up email to this lead:
Name: {{name}}
Company: {{company}}
Their message: {{message}}
Lead category: {{classification from Step 2}}
Rules:
- If hot-lead: Enthusiastic, mention scheduling a call this week
- If warm-lead: Helpful, share a relevant case study link
- If cold-lead: Soft touch, offer a free resource
Tone: Professional but warm. Under 150 words.
Sign off as: [Your Name], [Your Title]Step 4: Create CRM Record
- •Tool: HubSpot / Salesforce / Google Sheets
- •Map fields: name, email, company, lead score, category, summary
Step 5: Send or Queue Email
- •For hot leads: Send immediately via Gmail
- •For warm/cold leads: Create a draft in Gmail for human review
Step 6: Team Notification
- •Post to Slack: "New {{category}} lead: {{name}} from {{company}}. Score: {{score}}. Summary: {{summary}}"
Step-by-Step in n8n
The same workflow in n8n uses:
- 1.Webhook node or Form Trigger node
- 2.AI Agent node with Claude/GPT for classification
- 3.AI Agent node for email drafting
- 4.Google Sheets node or CRM node for record creation
- 5.Gmail node for email
- 6.Slack node for notification
The visual workflow builder lets you see the entire flow and debug each step.
Testing Your Automation
Test Checklist
- •[ ] Submit a test form with a hot lead profile → Does it get classified correctly?
- •[ ] Submit a test form with minimal information → Does it handle missing data gracefully?
- •[ ] Submit a test form in a different language → Does AI still classify correctly?
- •[ ] Check the CRM record → Are all fields populated correctly?
- •[ ] Read the drafted email → Is it personalized and appropriate?
- •[ ] Check Slack → Did the notification arrive with correct info?
Common Issues and Fixes
| Issue | Cause | Fix |
|---|---|---|
| AI returns wrong format | Prompt too vague | Add "Respond ONLY in JSON format: {score: number, category: string, ...}" |
| Empty fields in CRM | Data mapping error | Check that each CRM field maps to the correct step output |
| Email sounds generic | Not enough personalization data | Include more form fields in the email drafting prompt |
| Workflow runs slowly | Too many AI calls | Combine classification and email draft into one AI call |
| Duplicate records | Trigger fires multiple times | Add a deduplication check (search CRM before creating) |
Adapting This Template
This lead processor template can be adapted for:
- •Support ticket routing: Classify tickets instead of leads
- •Content moderation: Classify user-generated content
- •Job applications: Score and route resumes
- •Customer feedback: Analyze and route feedback to the right team
- •Expense reports: Extract data from receipts and categorize expenses
Exercises
0/3Build (or detailed plan) an AI-powered automation workflow for your business. Document: the trigger, each step with the specific tool and AI prompt, data mapping between steps, and how you would test it. If you have access to Zapier, Make, or n8n — actually build it!
Hint: Start with the simplest version that works. You can always add branches, error handling, and additional steps later. The Smart Lead Processor template is a great starting point.
When an AI step in a workflow returns data in the wrong format, what is the best fix?
List 3 workflows at your company that could benefit from an AI classification step. For each one, define: what data is being classified, what categories you would use, and what action happens for each category.
Hint: Think about any process where someone reads something and decides what to do with it. That decision step is a classification task AI can automate.