Claude Code + n8n Just Changed Cold Email Campaign Building Forever
If you’ve ever built a cold email campaign from scratch, you know the pain. Days—sometimes weeks—of tedious work just to get a single campaign live. You’re hunting for signals, qualifying leads, enriching data, personalizing messages, and loading everything into sequences. It’s exhausting.
But what if I told you that entire process could be condensed into 30 minutes of mostly hands-off work?
That’s exactly what I’m going to show you in this breakdown of how I built a signal waterfall end-to-end campaign building workflow using Claude Code and n8n. This isn’t theory—this is a real system that autonomously builds outbound campaigns while you’re out living your life.
The Setup: Claude Code Meets Conversational Automation
The workflow starts with Claude Code loaded up in my terminal. I paste in opening instructions that tell it what files to read, and from there, it starts asking questions. But here’s where it gets interesting: I use Whisper Flow to answer those questions verbally. No more typing out lengthy specifications or clicking through endless menus.
The system immediately starts with practical checks—like verifying how many credits are available in AI Arc (the lead database). In this case, it showed 14,335 credits available. But the real magic is in the watchdog feature.
The Watchdog: Your Safety Net
The watchdog monitors the entire workflow. If n8n crashes halfway through (because let’s be honest, sometimes tools glitch), the watchdog notices within 2 minutes and resumes everything from exactly where it left off. No data loss. No starting over. Just seamless recovery.
13 Questions to Campaign Launch
Claude Code walks you through 13 questions to gather everything needed to build your campaign. Here’s what that looks like in practice:
Question 1: ICP and Product/Service Filter
I told it: “Let’s target companies in the UK. We want SaaS companies with 3-8 people in the London area, and we’re excluding recruitment agencies and staffing agencies.”
It confirmed my inputs and asked for clarification on additional exclusions. I added consulting and agency firms to the exclude list. Simple.
Question 2: Location Filter
It suggested using London, United Kingdom in the proper JSON format. I confirmed and added an additional filter: companies founded after 2014.
Question 5: The Offer and Messaging
This is where you define your value proposition. I explained: “I specialize in engagement-driven outbound for B2B companies looking for a consistent flow of leads. The problem I solve is unpredictable lead flow. The value proposition is qualified leads with predictable, steady flow.”
Question 7: Signals
Signal-based outreach is where campaigns go from generic to powerful. The available signals include:
- Recent LinkedIn posts
- New in role (someone recently hired)
- Job postings (hiring signals)
For this demo, I chose “new in role” only. Why? Because when someone starts a new leadership position, they often have a fresh budget and a mandate to make changes. That’s prime timing.
Question 8: Snippet Prompts
This feature is brilliant. You can create predefined personalization snippets that adapt to each prospect. For example, different industries use different terms—some want “demos,” others want “consultations” or “sign-ups.” The system can auto-generate these contextual variations.
There’s also an “observation” snippet that gets generated when a signal is found. For example: “I see that you just started as VP of Sales at Acme Corp.” This becomes your opening line.
Question 9: Campaign IDs
For each channel (cold email or LinkedIn), two campaign types are created:
- Signal campaigns: For leads where a signal was found
- No signal campaigns: For leads without signals
This ensures your messaging is always contextually appropriate.
Let It Run: The Autonomous Build Process
After answering all questions, I simply said “Go.” What happened next was pure automation magic.
The system determined this was campaign number 33 (based on previous campaigns) and created a Google Sheet with proper naming conventions. It then ran an AI Arc probe and found 441 companies matching the criteria. At 0.1 credits per company, that’s 44.1 credits total.
After I approved, the workflow kicked into gear:
Step 2: Company Search and ICP Qualification
The system pulled 440 companies and ran them through a GPT qualification prompt. Result: 270 qualified companies, 170 rejected. All automatically logged in the Google Sheet and reported via Slack notification.
Step 3: People Search
Here’s where the intelligence really shines. The system only pulled people at qualified companies, and it capped the number of people per company based on employee headcount:
- Up to 20 employees: 2 decision-makers
- 21-40 employees: 3 people
- 41-80 employees: 4 people
- 81+ employees: 5 people
Result: 218 people from 270 companies.
One person was disqualified because their title was “Co-founder and Board Member”—the system correctly identified this as someone likely not actively involved in day-to-day operations.
Step 4-7: Signal Enrichment and Email Writing
The system checked for “new in role” signals and found one person who had recently joined as Chief Revenue Officer. It automatically generated an observation: “Noticed you recently joined Newmark as Chief Revenue Officer.”
But here’s the sophisticated part: before generating any personalization, it ran a pre-step prompt to clarify:
- What product/service does this company offer?
- Who are their buyers?
- What problem do buyers have?
- What solution does the company provide?
- What outcome do buyers expect?
- What are the implied challenges in selling this?
Why? Because AI often confuses who’s the seller and who’s the buyer. This pre-step ensures downstream prompts generate accurate, high-quality outputs.
Step 8: Sequencer Load
The final step loaded leads into the appropriate campaigns:
- 143 loaded to the SmartLead “No Signal” campaign
- 74 loaded to the HeyReach LinkedIn campaign
- 1 person with a signal loaded to the email “Signal” campaign
Total time: 30 minutes. All autonomous. All logged. All tracked.
The Real Game-Changer: Complete Autonomy
During those 30 minutes, I didn’t babysit the process. I could have gone for a walk, hit the gym, or grabbed lunch. The system sent Slack notifications at every step:
- Step 2 complete: Companies found and qualified
- Step 3 complete: People search finished
- Step 7 complete: Emails written
- Step 8 complete: Sequences loaded
Every notification included key metrics—how many leads were found, qualified, written, and loaded.
Why This Matters for Your Business
Building outbound campaigns traditionally requires:
- Hours of manual work in Clay tables
- Constant context-switching between tools
- Risk of errors in data transfer
- Days or weeks before a campaign goes live
This system eliminates all of that. You have a conversation with Claude Code, approve the plan, and let it run. The result? A fully built, signal-enriched, personalized outbound campaign ready to generate meetings.
This is the future of outbound. Not more tools. Not more complexity. Just intelligent automation that actually works.
Want help automating your lead generation? Book a call