Challenge Tasks
Table of contents
- Challenge Tasks
Overview
These challenges extend your cold email system with advanced features. Each builds on the AI Research & Personalization pattern you’ve learned.
Challenge 1: Multi-Channel Outreach
Objective
Extend your workflow to send personalised messages across multiple channels based on prospect preferences.
Requirements
- Research which channels the prospect is active on
- Generate platform-specific content (LinkedIn, Twitter/X, Email)
- Route to appropriate channel based on research
- Maintain consistent messaging across platforms
Why This Matters
Most professionals have preferred communication channels. Reaching them where they’re most active increases response rates significantly.
What You’ll Learn
- Multi-platform integration patterns
- Content adaptation for different channels
- Dynamic routing based on research insights
- Fallback logic when preferred channels aren’t available
Success Criteria
- ✅ Identifies prospect’s preferred communication channel
- ✅ Generates platform-appropriate content (280 chars for Twitter, professional for LinkedIn)
- ✅ Successfully sends via detected channel
- ✅ Falls back to email if no social presence found
Challenge 2: A/B Testing System
Objective
Implement A/B testing to optimise your email performance over time.
Requirements
- Create two email variations for each outreach
- Randomly assign prospects to version A or B
- Track performance metrics for each version
- Automatically select winning version after threshold
Why This Matters
What works for one audience might not work for another. A/B testing lets you systematically improve your messaging based on real data, not guesses.
What You’ll Learn
- Statistical testing in automated workflows
- Random assignment and traffic splitting
- Performance tracking and analysis
- Data-driven optimisation patterns
Success Criteria
- ✅ Generates two distinct email versions
- ✅ Evenly distributes prospects between versions
- ✅ Tracks version performance in Google Sheets
- ✅ Identifies statistically significant winner after 50+ sends
Challenge 3: Smart Follow-Up Sequence
Objective
Build an automated follow-up system that sends contextual messages if no response is received.
Requirements
- Track email opens and responses
- Send follow-up after 3 days of no response
- Each follow-up references previous context
- Maximum of 3 follow-ups then stop
Why This Matters
Most successful cold outreach requires 3-5 touches. Automated follow-ups dramatically increase response rates whilst respecting prospect’s time and attention.
What You’ll Learn
- Time-based automation with Wait nodes
- Gmail API for checking email responses
- Contextual message generation referencing history
- Stopping conditions to avoid spam
Success Criteria
- ✅ Waits appropriate time before follow-up
- ✅ Detects if prospect has responded
- ✅ Generates contextual follow-up referencing original
- ✅ Stops after 3 follow-ups or response
Challenge 4: Industry-Specific Templates
Objective
Create specialised email templates for different industries, automatically selected based on prospect’s company.
Requirements
- Identify prospect’s industry during research
- Maintain library of industry-specific approaches
- Select appropriate template and terminology
- Include industry-specific value propositions
Why This Matters
Generic emails feel like spam. Industry-specific language shows expertise and relevance, dramatically increasing engagement from qualified prospects.
What You’ll Learn
- Industry classification using AI research
- Template management and dynamic selection
- Context-aware content generation
- Domain-specific personalisation
Success Criteria
- ✅ Correctly identifies prospect’s industry
- ✅ Applies industry-specific template
- ✅ Uses appropriate terminology and pain points
- ✅ Shows measurably better engagement rates
Challenge 5: Sentiment-Aware Responses
Objective
Adjust email tone and approach based on prospect’s recent public sentiment (from social media, news).
Requirements
- Analyse sentiment from prospect’s recent posts/news
- Adjust email tone to match or complement
- Reference specific content that shows awareness
- Maintain authenticity whilst adapting
Why This Matters
Reaching out to someone celebrating a win requires different tone than someone facing challenges. Sentiment awareness makes outreach feel timely and empathetic, not robotic.
What You’ll Learn
- Sentiment analysis in research workflows
- Dynamic tone adjustment based on context
- Emotional intelligence in automation
- Balance between personalisation and authenticity
Success Criteria
- ✅ Accurately detects prospect sentiment
- ✅ Generates tone-appropriate emails
- ✅ References specific content naturally
- ✅ Improves response rates vs generic approach
Challenge 6: Multi-Tool Research Agent
Objective
Expand your research agent’s capabilities by connecting multiple tools, letting the AI decide which tools to use based on the research needs.
Requirements
- Add multiple research tools to the AI agent:
- Perplexity (web search) - already connected ✓
- Code node (for data processing/calculations)
- HTTP Request (to query LinkedIn API, company databases, or CRM)
- Google Search (alternative search engine)
- Give the AI freedom to choose which tools to use
- Compare results when AI has access to multiple sources
- Log which tools were used for each research request
Why This Matters
This is the future of AI workflows: Multi-tool agents that autonomously orchestrate research across different sources. Today’s simple Perplexity search becomes tomorrow’s sophisticated research agent that:
- Searches multiple sources
- Cross-references information
- Validates facts across platforms
- Chooses optimal data sources for each query
You’re learning agentic workflows - the foundation of AI automation in 2025+
What You’ll Learn
- Function calling patterns: How AI decides which tool to use
- Multi-tool orchestration: Connecting 3+ tools to one agent
- Tool selection logic: Understanding when AI picks different tools
- HTTP Request tool: Querying external APIs (LinkedIn, Clearbit, Hunter.io)
- Code tool: Processing and analysing data returned from tools
- Debugging multi-tool agents: Understanding tool execution logs
Implementation Guide
Step 1: Add HTTP Request Tool (LinkedIn Company Lookup)
- In your AI Agent - Research Prospect node, click ”+” under Tools
- Select “HTTP Request”
- Configure:
- Method: GET
- URL:
https://nubela.co/proxycurl/api/linkedin/company(or similar LinkedIn API) - Authentication: Add your API key
- Update your agent prompt to mention this tool is available
Step 2: Add Code Tool (For Data Analysis)
- Click ”+” under Tools again
- Select “Code”
- This allows the AI to run calculations, parse JSON, or analyse data
Step 3: Update Your Research Prompt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
You are an expert researcher with access to multiple tools:
- Perplexity: Web search for recent news and general information
- HTTP Request: Query LinkedIn API for professional background
- Code: Analyse and process data, run calculations
Research the following person/company intelligently:
Name:
Company:
IMPORTANT: Choose the RIGHT tools for each subtask:
- For recent news/achievements: Use Perplexity
- For professional background/company info: Use HTTP Request (LinkedIn)
- For data processing: Use Code tool
Provide a comprehensive research summary combining insights from all sources.
Step 4: Compare Results
Test your workflow with and without multiple tools:
- Single tool (Perplexity only): What quality of research?
- Multi-tool: How does the AI decide which tool to use? Is research richer?
Success Criteria
- ✅ Successfully connected 2+ additional tools to the research agent
- ✅ AI autonomously chooses appropriate tool for each subtask
- ✅ Research quality improves with multi-source data
- ✅ Can track which tools were invoked in execution logs
- ✅ Understand when to use each tool type
Stretch Goals
- Add 5+ tools: Web search, LinkedIn, company database, news API, sentiment analysis
- Tool chains: AI uses one tool’s output as input to another tool
- Conditional tool selection: Prompt AI to use certain tools only if specific conditions are met
- MCP integration: Research Model Context Protocol servers for enterprise tool connections
Real-World Applications
The multi-tool pattern you learn here scales to:
- Customer support bots: Search docs + Query CRM + Check order status + Generate response
- Financial analysis: Fetch stock data + Run calculations + Generate report + Send alerts
- Content research: Search web + Query database + Analyse sentiment + Draft article
- Sales intelligence: Research company + Check CRM + Find contact info + Generate pitch
Saving Your Work
After completing any challenge:
- Export Workflow: In n8n, click ⋯ menu → Download → Save as JSON
- Commit to GitHub: Save your workflow JSON files to your repository
- Document Changes: Add notes explaining what you built and why
Pro Tip: Each challenge teaches skills applicable far beyond cold email. Think about how these patterns apply to your specific use cases!