clay-automation

Files (7)

• SKILL.md
• automation-playbook-index.md · references/
• eric-nowoslawski-outbound.md · references/
• framework-notes.md · references/
• gtm-automation-expert-playbook.md · references/
• check-output.py · scripts/
• output-template.md · templates/

SKILL.md (first lines)

# Clay Automation

## Overview

Clay is the orchestration layer where enrichment, scoring, and routing converge.
Used correctly, it's a GTM multiplier. Used incorrectly, it's a credit-burning
machine producing data nobody trusts.

This skill covers **when and how to roll out** Clay automation — data quality,
table separation, rollout phases, and n8n handoff.

**Playbook index:** `references/automation-playbook-index.md` — all 38 automation + tool + gtm-ops playbooks.

**Automation strategy (before tool config):** Jen Igartua (Go Nimbly) — data before AI,
maturity levels 0–4, human+machine division. Canonical → `references/gtm-automation-expert-playbook.md`
(Pattern 30). Outbound copy/infra → Eric Nowoslawski / Pat Spielmann / Justin Michael — not this skill.

**Tool implementation**
(table columns, LeadMagic waterfalls, loops) lives in `tools/`:
- `clay-toolkit` (`tools/clay-toolkit`) — tables and waterfalls
- `clay-loops-toolkit` (`tools/clay-loops-toolkit`) — signal loops

## When to Use

- "Build a Clay enrichment workflow"
- "Set up a Clay table for our prospecting"
- "Configure a waterfall in Clay"
- "Use Claygent for research"
- "Push enriched data from Clay to HubSpot"
- "Optimize our Clay credit usage"
- "Design a Clay automation pipeline"

## Authoritative Foundations

Clay workflow design follows patterns from DAMA-DMBOK data-quality dimensions,
Ziellab (3 separate waterfalls: company, email, phone), and GTME Pulse
(10 production templates tested at $5M-$100M ARR companies).

The core principle: Clay is a routing engine, not a CRM. Enriched data
lives in your CRM; Clay processes it en route.

- **Eric Nowoslawski — Crawl Walk Run (Growth Engine X).** Roll out Clay
  automation in phases: **Crawl** — manual campaigns for 5–10 companies (no AI);
  **Walk** — system prompt with business context + manual examples, review first
  50 outputs; **Run** — automate in Clay → Smartlead/SEP with Supabase for block
  lists. Default campaign type: **Creative Ideas** (3 constrained capabilities per
  prospect). Playbook → `references/eric-nowoslawski-outbound.md`.

## Prerequisites

- Clay account (Pro plan or above for waterfall features and conditional columns)
- API keys for enrichment providers (Apollo, ZoomInfo, PDL, etc.)
- CRM connected (HubSpot, Salesforce, or Attio)
- ICP defined and documented

## Step-by-Step Process

### Phase 1: Table Architecture

**Rule: separate company and person tables.**

Company table: one row per domain. Enriches firmographics, tech stack,
company-level qualification once.

Person table: one row per contact. References company data via domain
lookup. Enriches email, phone, LinkedIn, role-level qualification.

This separation prevents credit waste from re-enriching the same company
data across every contact row.

### Phase 2: Waterfall Configuration

For each data field, configure a conditional waterfall:

| Field | Primary | Fallback 1 | Fallback 2 |
|---|---|---|---|
| Email | LeadMagic Email Finder | Apollo | Hunter |
| Phone | Apollo | Cognism | ContactOut |
| Company data | Clay native | Apollo Company | Clearbit |

Set conditions: each fallback only fires when the previous step returns
empty or error. Use Clay's conditional logic: "Only run if [previous
column] is blank."

### Phase 3: Claygent Configuration

Claygent is AI-powered web research. Configure prompts explicitly:

**Good prompt:** "Find the work email for [name] at [company]. Search
the company's team page, LinkedIn profile, and press releases. Return
the email AND the source URL. Do NOT guess or construct emails from
patterns. If no verified source, return empty."

**Bad prompt (don't use):** "Find me their email."

Critical rules:
- Always require source URL citation
- Explicitly prohibit pattern-guessing
- Set credit cap per Claygent call (5-10 credits)
- Use only for the 10-15% that structured providers miss

### Phase 4: CRM Push

Push enriched data to CRM with a clay_status property:

| clay_status | Meaning | Action |
|---|---|---|
| pending | In enrichment | Hold — not ready |
| enriched | Enrichment complete | Ready for verification |
| verified | Verified and safe | Can enter sequences |
| exported | Pushed to CRM | Done, archive in Clay |

Only contacts with clay_status = verified enter sequences.

### Phase 5: Credit Optimization

1. **Qualify first, enrich deep later.** Run ICP filters before expensive
   contact enrichment. Cuts costs 30-40%.

From LeadMagic/gtm-skills, MIT.

More from LeadMagic/gtm-skills