End-to-End Workflow

​

Campaign Phases

​

Phase 1: Foundation

​

Build Company Context

Build a company context file for www.extruct.ai

1. Claude reads your website to extract product info and value prop
2. Asks for voice rules, ICP profiles, win cases
3. Creates /claude-code-gtm/context/extruct_context.md

• Single context file with ICP, voice, win cases, proof library
• This file is referenced by every downstream skill

​

Phase 2: Research

​

Option A: Fast Hypothesis Building (No API Required)

Build a hypothesis set for enterprise SaaS platforms
targeting companies in North America

1. Reads context file for ICP and win cases
2. Asks what you know about the vertical
3. Drafts 3-7 hypotheses with search angles
4. Saves to /claude-code-gtm/context/enterprise-saas/hypothesis_set.md

​

Option B: Deep Market Research (API Required)

Research the enterprise SaaS vertical—what are their
data enrichment pain points?

1. Reads context file and runs research queries
2. Extracts data points, tool comparisons, pain mechanisms
3. Saves research + hypothesis set

• /claude-code-gtm/context/enterprise-saas/sourcing_research.md
• /claude-code-gtm/context/enterprise-saas/hypothesis_set.md

​

Phase 3: List Building & Enrichment

​

Step 1: Find Companies

Find 500 enterprise SaaS platforms similar to salesforce.com

1. Reads hypothesis set for search angles
2. Runs lookalike/semantic/discovery searches
3. Deduplicates and removes DNC domains
4. Uploads to Extruct table for enrichment

• Extruct table with company profiles
• Optional local CSV at /claude-code-gtm/csv/input/{campaign-slug}/companies.csv

​

Step 2: Design Enrichment Columns

Design enrichment columns to qualify these companies
for our data enrichment hypothesis

1. Reads hypothesis set
2. Proposes 3-5 columns (segmentation + personalization)
3. Refines with you interactively
4. Outputs column_configs JSON

• Segmentation: Columns that score hypothesis fit (e.g., “Data Infrastructure Maturity”)
• Personalization: Columns for email hooks (e.g., “Recent Product Launch”)

​

Step 3: Run Enrichment

Add these enrichment columns to table abc123

1. Creates agent columns in Extruct table
2. Triggers enrichment run (research agents execute per row)
3. Polls for progress and shows % complete
4. Spot-checks quality with sample rows

​

Step 4: Segment by Fit

Segment this list into Tier 1/2/3 based on enrichment data

1. Reads hypothesis set and enrichment columns
2. Scores each company on hypothesis match + data quality
3. Assigns tier (1 = best fit, 3 = no match)
4. Exports segmented CSV

• /claude-code-gtm/csv/input/{campaign-slug}/companies_segmented.csv

• Tier 1: Strong hypothesis match + rich enrichment data → personalized emails
• Tier 2: Moderate match → hypothesis-based templates
• Tier 3: Weak match → exclude or re-research

​

Phase 4: Email Generation & Sending

​

Step 1: Find People

Find VPs of Engineering at these Tier 1 companies

1. Reads company CSV
2. Runs LinkedIn search via Extruct API
3. Returns people with titles, LinkedIn URLs

​

Step 2: Get Contact Info

Get emails for these contacts using Prospeo

1. Reads people CSV
2. Calls contact enrichment provider (Prospeo, FullEnrich)
3. Returns verified emails, phones, social profiles

​

Step 3: Build Email Prompt Template

Build an email prompt for enterprise SaaS VPs of Engineering

1. Reads context file (voice, proof library, value prop)
2. Reads research + hypothesis set
3. Synthesizes everything into a self-contained prompt
4. Saves to /claude-code-gtm/prompts/enterprise-saas/en_first_email.md

• Voice rules and banned words (from context file)
• Hypothesis-based P1 angles (from research)
• P2 value angles (from context → What We Do)
• P4 proof points (from context → Proof Library)
• Structural variants (by role and seniority)

​

Step 4: Generate Emails

Generate emails using the prompt template and contacts CSV

1. Reads prompt template + contact CSV
2. Applies prompt per row with enrichment data
3. Outputs email JSON per contact
4. Saves to /claude-code-gtm/csv/output/{campaign-slug}/emails.csv

• Tier 1: Individual attention, routed to simulation for review
• Tier 2: Batched by hypothesis group
• Tier 3: Skipped

​

Step 5: Review Tier 1 Emails (Optional)

Simulate responses for these Tier 1 emails

1. Reads generated emails
2. Simulates persona reading the email
3. Flags issues: buzzwords, weak hooks, generic copy
4. Suggests rewrites constrained by voice rules

​

Step 6: Upload to Sequencer

Upload these contacts to Instantly campaign 'enterprise-saas-q1'

1. Reads context file for DNC list
2. Filters contacts against DNC
3. Uploads to sequencer via API
4. Returns upload summary

​

Feedback Loop

Update context file with campaign results from Instantly

1. Imports campaign metrics (reply rate, positive replies)
2. Updates Campaign History table in context file
3. Promotes/retires hypotheses based on performance
4. Adds new proof points from positive replies

​

Example: Full Campaign in One Prompt

I'm building www.extruct.ai.
One customer is www.salesforce.com—they use us for market intelligence.
Find 200 similar enterprise SaaS platforms in North America,
enrich them with relevant data points,
and generate personalized emails to VPs of Product.

1. Build context file from extruct.ai
2. Generate hypothesis set for “enterprise SaaS platforms”
3. Run lookalike search from salesforce.com
4. Design + run enrichment columns
5. Segment into tiers
6. Find VPs of Product
7. Get emails
8. Build prompt template
9. Generate emails
10. Export for sending

​

What Gets Created

claude-code-gtm/
├── context/
│   ├── extruct_context.md
│   └── enterprise-saas/
│       ├── hypothesis_set.md
│       └── sourcing_research.md (if using market-research)
├── prompts/
│   └── enterprise-saas/
│       └── en_first_email.md
└── csv/
    ├── input/enterprise-saas-q1/
    │   ├── companies.csv
    │   ├── companies_segmented.csv
    │   ├── people.csv
    │   └── contacts.csv
    └── output/enterprise-saas-q1/
        └── emails.csv

​

Next Steps