Your Best Customers Are a Blueprint – Here's How to Find More of Them
Every company has a handful of accounts that close faster, expand sooner, and churn less. Those customers are not random. They share patterns – industry, team size, tech stack, buying triggers – that predict who else will convert.
The problem is that most teams never extract those patterns. They eyeball a few logos, guess at titles, and blast a generic list.
Claude handles the entire pipeline – from mining customer intelligence to delivering a scored, contact-enriched prospect list ready for outreach.
How Claude Helps You
Claude Code runs a six-step lookalike prospecting workflow end to end. It analyzes your existing customers, extracts the ideal profile, searches for matching companies, scores and enriches each one, finds the right decision-makers, and formats everything for your outreach tool.
To do this, Claude uses these skills:
- customer-research – mines transcripts, reviews, support tickets, G2, Reddit, and forums for customer patterns (marketingskills)
- icp-identification – structures raw research into a scored Ideal Customer Profile with inclusion and exclusion filters (Gooseworks)
- tam-builder – runs Apollo Company Search, scores ICP fit 0–100, and assigns tiers (Gooseworks)
- lead-research-assistant – qualifies leads and generates personalized outreach strategies (ComposioHQ)
- company-contact-finder – finds decision-makers by name and title using Crustdata and SixtyFour people databases (Gooseworks)
You provide: customer transcripts or call recordings, CRM export of your best accounts, and your product URL. Claude does the rest.
The Workflow
Step 1: Analyze Your Best Customers
Start by pointing Claude at your raw customer data – interview transcripts, sales call recordings, support tickets, G2 reviews, even Reddit threads where your users talk about you. The customer-research skill extracts structured signal from every source.
What Claude pulls out:
- Jobs to be done (functional, emotional, social)
- Pain points mentioned unprompted and with emotional language
- Trigger events that started the buying process
- Exact vocabulary customers use to describe the problem
- Alternatives they evaluated before choosing you
This step typically processes 20–30 sources in under 10 minutes. The output is a research synthesis document – not a summary, but a structured extraction of patterns that feed directly into the next step.
Step 2: Extract the ICP Profile
Raw research becomes a usable targeting profile. The icp-identification skill takes the synthesis from Step 1, researches your company and competitors via the web, and proposes a structured ICP table.
The output includes both inclusion and exclusion criteria:
- Target titles: VP Operations, Head of Customer Success, Director of RevOps
- Company size: 51–500 employees
- Industries: SaaS, FinTech, MarTech
- Signals: recently hired for the role your product replaces, posted about the pain on LinkedIn
- Exclusions: agencies, government, companies under 20 employees
Claude presents the ICP for your approval before moving on. You can adjust any dimension – title seniority, headcount range, geography – and the downstream search adapts automatically.
Step 3: Build the Lookalike List
With the ICP locked, the tam-builder skill runs Apollo Company Search using your approved filters. Each company that comes back gets scored on a 0–100 ICP fit scale using weighted criteria:
- Employee count fit (30%)
- Industry match (25%)
- Funding stage alignment (20%)
- Geographic fit (15%)
- Keyword relevance (10%)
Companies scoring 75+ land in Tier 1. Those between 50–74 go to Tier 2. Everything below 50 gets deprioritized. A typical run surfaces 200–500 companies across 3–5 Apollo search queries, with 40–80 landing in Tier 1.
Step 4: Enrich and Qualify
Each Tier 1 and Tier 2 company goes through the lead-research-assistant skill for deeper qualification. This step adds context that a simple search cannot provide.
For every company, Claude identifies:
- How well the company's pain points match your value proposition
- Technology stack overlaps and gaps
- Growth signals (recent funding, hiring spree, product launches)
- Personalized outreach angles based on what the company is publicly doing
The result is a prioritized list where each account has a fit score and a recommended approach – not just "send a cold email," but a specific angle based on real intelligence.
Step 5: Find Decision-Makers
Qualified accounts need real contacts. The company-contact-finder skill queries Crustdata and SixtyFour people databases to locate decision-makers at each target company.
You specify the titles you care about – VP Engineering, Head of Product, CTO – and Claude returns:
- Full name and current title
- LinkedIn profile URL
- Location
- Seniority confirmation
A natural-language search runs first (e.g., "VP Engineering OR CTO at Acme Corp"). If fewer than three quality matches come back, a structured filter search runs as fallback. The skill typically returns 3–5 contacts per company, so a 50-company Tier 1 list yields 150–250 named contacts.
Step 6: Push to Your Outreach Tool
The final output is a structured CSV or JSON file containing companies, scores, contacts, and recommended outreach angles. Import it directly into Smartlead, HubSpot, Salesforce, or whatever CRM your team runs.
Every row includes the ICP score, tier assignment, company context, contact details, and the personalized angle from Step 4 – so your SDRs can start working the list immediately without additional research.
What You Walk Away With
- ICP document – structured profile with inclusion criteria, exclusion filters, and scoring weights derived from real customer data
- Scored company list – 200–500 companies with 0–100 ICP fit scores and tier assignments
- Qualified accounts – Tier 1 and Tier 2 companies enriched with pain-point alignment, tech stack, and growth signals
- Contact list – 150–250 named decision-makers with titles, LinkedIn URLs, and locations
- Outreach angles – personalized approach recommendations for every qualified account
- CRM-ready export – CSV or JSON formatted for direct import into your outreach tool
Why This Matters
Most prospecting fails because the list is wrong – built from guesswork instead of customer intelligence. This workflow starts with what actually makes your best customers tick and works outward from there.
The result is a pipeline that converts at higher rates because every company on the list resembles the accounts that already succeed with your product. That is what lookalike prospecting with AI looks like when Claude Code, Codex, and Goose handle the execution.
Get Started
Build your first lookalike prospect list today. Visit gooseworks.ai to get started.