Our 7-Step Process for Building Cold Campaigns Inside Claude Code

Creating cold outreach campaigns used to take hours. List building, enrichment, copywriting, and campaign setup each lived in a separate tool, and every handoff between them cost time and introduced errors.

Claude Code collapses that work into minutes. It handles list building, data enrichment, personalized copywriting, and launching new campaigns without you ever leaving the terminal.

We run outbound for 70+ B2B clients, and this is the exact seven-step process we use to go from a raw company list to a ready-to-send campaign inside a single session.

1. List Building

There are several ways to build the source list, and Claude Code handles all of them from one place.

Lookalike targeting is the first. Using PredictLeads, you find companies that resemble the ones already paying you, then expand outward from accounts that have a proven fit.

You can also pull leads from standard B2B databases. Apollo.io, CompanyEnrich, and Sales Navigator each cover different slices of the market, and Claude Code processes 50,000+ rows from large CSV files without choking.

Lookalike targeting is the fastest way to clone accounts that already convert, and we built a mini tool that runs the same logic on your customer list.

You can find companies similar to your best customers here, for free:

Lookalike Finder Tool

2. Account Scoring and Tiering

A raw list is not a campaign. Before any outreach goes out, every account gets scored against our ICP.

We built custom GTM skills, including a lead-scoring skill, that run our ICP criteria against the full list. Claude Code reads each account, weighs the criteria the way a senior rep would, and assigns a score, autonomously, across the entire list at once.

From there we segment into three tiers. Tier 1 accounts get manual outreach. Tier 2 accounts go into multi-channel sequences. Tier 3 accounts run through email outreach. The tier decides how much personalization effort each account earns downstream, so the highest-fit accounts get the most attention.

3. Intent Signals

Scoring tells you who fits. Signals tell you when to reach out.

APIs such as PredictLeads and Trigify.io detect hiring activity, press releases, funding announcements, and social engagement. Claude Code calls the right API for each signal type and layers the results on top of the existing targeting, so a fresh trigger moves an account up the priority list automatically.

Based on these signal APIs, we built a mini tool that surfaces the same buying triggers without writing a line of code.

You can see which companies are showing intent in your space right now, for free:

Intent Signals Tool

4. Decision-Maker Filtering

Targeting the right company is only half the job. You still need the right people inside it.

APIs such as LeadsFactory return multiple decision makers per company, filtered by the ICP definition: job title, seniority, and department. Claude Code pulls the relevant contacts for every account on the list, so a single company can surface a full buying committee rather than one stray name.

We packaged the same capability into a free tool that maps the decision makers at any target account.

You can identify the right contacts at your target companies here, for free:

People Finder Tool

5. Data Enrichment and Validation

With the right contacts identified, the next step is finding verified ways to reach them.

Claude Code runs a waterfall across several providers to find phone and email data, starting with the highest-confidence source and cascading through the rest for the gaps. Apollo.io, FullEnrich, Prospeo, and CompanyEnrich each cover contacts the others miss, which pushes the match rate higher than any single provider could on its own.

Before the list reaches a sequencer, Claude Code cleans the risky addresses. Validation through BounceBan strips catch-all and high-risk emails so the campaign launches on a clean list and the sending domains stay protected.

6. Copy Generation and Deployment

Clean data is ready for copy, and this is where the process stops feeling like data work.

Claude Code pulls our best-performing campaign copy straight from lemlist or Instantly.ai through their APIs, then drafts custom copy built on the signals, current initiatives, and target persona behind each segment.

It works in batches of 100 to 200 leads at a time. Smaller batches keep quality control tight, so a weak angle gets caught and corrected before it goes out to thousands of contacts rather than after.

7. Feedback and Auto-Improvement

The process does not end at launch. This is the step that separates a one-off campaign from a system.

Claude Code analyzes campaign metrics after launch, gathers insights on the segments that perform best, and finds lookalikes of your top-converting leads to feed the next batch.

Campaigns improve on their own over time, in the same way a Meta Ads pixel sharpens as it gathers data. Each launch teaches the next one, so the targeting compounds instead of resetting every time you start fresh.

8. The Setup That Makes This Work

Going from prompt to ready-to-send campaign in minutes is real, but it sits on top of groundwork you build once.

Three pieces make up that foundation. A claude.md file holds your scoring rules, tool-calling preferences, and built-in workflow logic. Custom outbound skills encode the repeatable actions, like lead-scoring and tiering. API keys connect Claude Code to the data and sending layers it orchestrates.

The API stack is worth choosing deliberately, by layer. For data, we lean on Clay, Apollo.io, CompanyEnrich, and Openmart. For enrichment, Prospeo, FullEnrich, CompanyEnrich, and LeadsFactory. For signals, PredictLeads and Trigify.io. For orchestration, Clay, Relevance AI, and n8n. For action, Expandi, Instantly.ai, and lemlist. For validation, BounceBan.

Once that foundation is in place, building a campaign becomes a single prompt. If you want to understand where your current GTM motion stands before wiring this together, see how your approach compares below:

GTM Report Tool

---