The Problem with Raw LinkedIn Data
LinkedIn profiles are inconsistent. You'll see:
Companies: "Apple Inc.", "APPLE COMPUTER", "Apple (Cupertino Office)"
Names: "John (He/Him)", "SARAH", "Dr. Michael Smith Jr."
Job titles: "Senior VP of Biz Dev & Strategic Partnerships (Remote) - We're hiring!"
Using this data directly makes your messages look robotic. That's where Botdog's AI variables come in.
How to Create AI Variables
Step 1: Start a new campaign
Step 2: Click the "AI variables" button
Step 3: Either use one of our templates or build your own
You can click "Use template" to see our pre-built prompts and learn from them.
Our 3 Core Templates
Clean Company Name We already remove basic suffixes like "LLC" and "GmbH" with our standard {{currentCompany}} variable, but AI handles complex cases much better. It can turn "Microsoft Corporation - Seattle Division" into just "Microsoft" or handle international variations that fixed rules miss.
Clean First Name Standard {{firstName}} already removes emojis, but AI goes further. It handles titles, pronouns, multiple names, and weird capitalization that rules-based cleaning misses.
Clean Job Title Raw job titles are often bloated with keywords, locations, and hiring pitches. AI extracts the actual role: "Rockstar Senior Software Engineer (Remote, $200K, We're hiring!)" becomes "Senior Software Engineer."
See our templates by clicking on "Use template
What LinkedIn Data AI Can Analyze
The AI system reads LinkedIn profiles like a human would, analyzing:
Professional Information
Job title and seniority level
Employment history and career progression
Industry experience and specializations
Company Data
Company size (startup, SMB, enterprise)
Industry and business type
Growth stage and funding status
Profile Content
About section summary
Skills, endorsements, and recommendations
Publishing activity and engagement levels
Education & Location
Degrees, certifications, and institutions
Current location and remote work status
Contact information and linked platforms
AI Limitations to Remember
AI can only work with publicly visible LinkedIn data. It can't determine someone's budget, whether they're actively buying, their quarterly targets, or personal preferences not mentioned in their profile.
Keep your AI prompts focused on concrete, observable data like job titles, company size, or listed skills rather than trying to guess subjective factors.
Creative AI Variable Ideas
Industry Simplified: "Information Technology & Services" → "Tech"
Seniority Level: Extract "Senior," "Lead," "Director" from messy titles
Location Cleaned: "San Francisco Bay Area, California, United States" → "San Francisco"
Company Size: Turn employee count into "startup," "mid-size," or "enterprise"
Product Personalization: Create prompts with your main value propositions and ask AI to personalize based on their role and company
Suggestion Generator: "I see you work at a LinkedIn automation company - have you considered running ads? Here are tactics that work well for similar companies..."
Jokes/Poems: People try this, but it's risky - AI humor often feels obviously generated
Sky's the limit with what you can create!
Prompt Engineering Basics
Be specific. Don't say "clean this company name." List exactly what to remove: legal suffixes, locations, departments, etc.
Give examples. Show input and desired output:
"Apple Inc." → "Apple"
"Google LLC (Mountain View)" → "Google"
Handle edge cases. What happens with empty fields? Non-English text? Decide upfront.
Test thoroughly. Try your prompts on weird LinkedIn profiles to catch issues.
For advanced techniques, check the Prompting Guide and OpenAI's best practices.
Getting Started
Replace your existing merge tags with AI versions in a few test campaigns. You'll immediately see cleaner, more professional messages that get better responses.
Start with company and job title cleaning - those have the biggest impact. Then experiment with custom variables for your specific use case.
AI merge tags aren't magic, but they solve the data quality problem that makes most automated outreach look obviously automated. Clean data = better responses.