A pipeline that does not need you.
Most GTM systems are manual at their core - a founder or SDR doing the same prospecting tasks on a loop. This blueprint replaces that loop with a four-phase automated pipeline: signal detection, prospect enrichment, outreach sequencing, and call booking. The only step that requires you is closing.
This blueprint is the architecture. Deep Loom builds the implementation - so you have a running pipeline in 30 days, not 6 months.
The 4 Phases
The self-managing GTM pipeline operates in four sequential phases. Each phase has a defined input, a defined output, and a defined automation layer. Nothing crosses to the next phase without meeting a quality threshold.
| Phase | Input | Output | Automation Layer |
|---|---|---|---|
| Phase 1 - Signal | LinkedIn activity, job changes, company events | Qualified prospect list (ICP-matched, signal-tagged) | PhantomBuster + Sales Navigator |
| Phase 2 - Enrichment | Prospect LinkedIn URL | Enriched profile with pain points, company context, intent score | Clay + Claude API |
| Phase 3 - Sequence | Enriched profile | Personalized outreach messages sent on schedule | n8n + PhantomBuster + Claude |
| Phase 4 - Booking | Positive reply or call request | Booked call on Calendly + CRM entry created | Calendly + Notion + n8n |
The pipeline runs continuously. Prospects enter Phase 1 every day through automated signal monitoring. Prospects exit Phase 4 as booked calls, closed deals, or archived contacts. The founder only touches Phase 4 - the actual conversation.
The Signal Layer
The signal layer is the top of the pipeline. Its job is to identify prospects who are in-market right now - not just matching your ICP, but displaying a behavior that indicates buying intent or heightened receptiveness.
The Five Signal Types
- Post engagement signal: A prospect comments on or reacts to your LinkedIn post. This is the highest-intent signal - they have voluntarily engaged with your content. Conversion rate from signal to call: 18–25%.
- Profile view signal: A LinkedIn Premium member viewed your profile. Intent is lower than post engagement, but timing is excellent - they were just thinking about you. Conversion rate: 8–12%.
- Job-change signal: A prospect in your ICP just started a new role. New leaders are in maximum buying mode for the first 90 days. Conversion rate: 6–10%.
- Competitor engagement signal: A prospect comments on a competitor's content. They are actively researching your category. Timing-wise, this is excellent - they are in evaluation mode.
- Funding / hiring signal: A company in your ICP just raised a round or posted a role that matches a pain point your offer solves. Sourced via Crunchbase, LinkedIn Jobs, or the Clay funding signal integration.
Each signal type has a dedicated PhantomBuster phantom running on a daily schedule. The outputs from all five phantoms are merged into a single Clay table for Phase 2 enrichment.
The Enrichment Engine
Enrichment transforms a raw signal (a name and a LinkedIn URL) into a complete prospect profile - company context, likely pain points, recent activity, and a personalization seed for the outreach message. This is where Clay + Claude does work that used to require a full-time researcher.
The Clay Enrichment Table Schema
- LinkedIn URL → full profile data (title, company, tenure, recent posts)
- Company name → Clearbit / Apollo company data (size, revenue estimate, tech stack)
- Recent post activity → top 3 posts from last 30 days pulled via PhantomBuster scraper
- Job change date (if applicable) → days in new role, previous company, previous title
- Signal source → which of the five signal types triggered this row (used for message template selection)
The Claude Enrichment Prompt
Once the Clay table is populated, a Claude API call runs against each row to generate a personalization seed - a two-to-three sentence observation about the prospect that will be used in the outreach message.
You are writing a personalization seed for a B2B outreach message.
The prospect is: {{full_name}}, {{title}} at {{company}}.
Their recent LinkedIn activity: {{recent_posts_summary}}.
Their signal type: {{signal_source}}.
Company context: {{company_description}}.
Write 2–3 sentences that:
1. Reference something specific and real from their profile or recent activity
2. Connect it naturally to the problem we solve ({{offer_one_liner}})
3. Sound like something a thoughtful human noticed - not a bot template
Output only the 2–3 sentences. No preamble. No explanation.The Sequence System
The sequence system takes the enriched profile and the personalization seed and runs the five-step outreach sequence automatically. Every message is generated by Claude, personalized with the enrichment data, and sent on a time-delayed schedule via PhantomBuster.
The 5-Step Sequence Schedule
| Step | Message Type | Timing | Goal |
|---|---|---|---|
| Step 1 | Connection request + note | Day 0 | Get connected (target: 55–70% accept rate) |
| Step 2 | Value drop message | Day 2 post-connect | Deliver value before any ask |
| Step 3 | Soft question | Day 4 | Start a two-way conversation |
| Step 4 | Follow-up (if no reply) | Day 7 | Re-engage without pressure |
| Step 5 | Break-up message | Day 11 | Final message - generates 12–18% reply rate |
If a prospect replies at any step, they are immediately moved out of the automated sequence and flagged in Notion for founder review. The automation handles non-responders. Responders always get a human.
"The break-up message is the most counterintuitive part of this system. It converts at higher rates than any of the earlier messages - because it removes all pressure and gives people permission to ignore you. Ironically, that's what makes them reply."
The n8n Stack
n8n is the orchestration layer that connects every tool in the pipeline. Below is the complete workflow architecture - node by node.
| Node | Trigger / Action | Connected To |
|---|---|---|
| Signal Collector | Daily 6 AM - pulls PhantomBuster outputs from Google Sheets | Clay table |
| Deduplication Check | Checks Notion CRM for existing contacts with same LinkedIn URL | Clay table (filtered) |
| Clay Enrichment Trigger | Sends new rows to Clay waterfall via webhook | Claude API |
| Personalization Generator | Calls Claude API with enrichment prompt for each row | Notion CRM |
| Notion CRM Writer | Creates contact record with full enrichment + personalization seed | PhantomBuster queue |
| Sequence Launcher | Adds contact to PhantomBuster connection request queue | |
| Reply Monitor | Checks LinkedIn DM inbox every 2 hours for new replies | Notion (reply flag) |
| Booking Tracker | Monitors Calendly for new bookings and updates Notion CRM | Slack (notification) |
The 7-Day Launch Plan
This is the exact sequence Deep Loom follows when deploying the self-managing GTM pipeline for a new client. It is designed to have the pipeline live and processing its first prospects by day 7.
- Day 1 - ICP clarity: Define the ICP in precise, filterable terms: title, company size, industry, geography, and the signal event that indicates buying intent. Write the offer one-liner that will power all Claude prompts.
- Day 2 - Signal setup: Configure Sales Navigator saved search with ICP filters. Set up the first two PhantomBuster phantoms: post commenter scraper and profile viewer monitor. Test both and verify output format.
- Day 3 - Clay table build: Create the Clay enrichment table with the schema from Chapter 03. Connect PhantomBuster output to Clay input via webhook. Test with 10 manual rows to verify enrichment quality.
- Day 4 - Claude prompts: Write and test the personalization prompt from Chapter 07. Run 20 test rows through Clay + Claude. Review every output manually. Adjust the prompt until 80% of outputs are publishable without editing.
- Day 5 - Sequence setup: Configure the five-step PhantomBuster sequence with connection request note templates. Write all five message templates using the enrichment seed field. Test with 5 manual test contacts (use your own LinkedIn alt accounts or team members).
- Day 6 - n8n wiring: Build and test the full n8n workflow from signal collector to Notion CRM. Verify that every node handles errors gracefully. Run a dry-run with 10 real prospects from the Clay table.
- Day 7 - Launch at 20% volume: Enable the daily signal collectors and set PhantomBuster limits to 20 connection requests per day. Monitor every output for the first 5 days. Scale to full volume (50/day) after confirming quality.
The Prompt Library
These are the core prompts that power the self-managing GTM pipeline. Each is production-tested and designed to be customized with your specific ICP, offer, and voice.
Write a LinkedIn connection request note (under 200 characters) for:
Name: {{full_name}}
Title: {{title}}
Company: {{company}}
Signal: {{signal_source}}
Personalization seed: {{personalization_seed}}
Rules:
- Reference the signal naturally (do not say "I saw you commented on my post")
- Sound like a human, not a template
- End with a soft reason to connect - not a pitch
- 200 characters maximum (hard limit)Write a LinkedIn DM for someone we just connected with.
Name: {{full_name}}, {{title}} at {{company}}.
Personalization seed: {{personalization_seed}}
Our offer: {{offer_one_liner}}
Best relevant resource/insight: {{resource_description}}
Rules:
- Start with a reference to their situation, not ours
- Deliver one specific piece of value (framework, insight, or resource)
- Zero ask in this message - pure give
- Under 120 words
- Conversational, not polished corporateWrite the final follow-up message for a prospect who has not replied after 4 previous messages.
Name: {{full_name}}, {{title}} at {{company}}.
Original signal: {{signal_source}}
Rules:
- Explicitly say this is the last message
- Remove all pressure - give them full permission to ignore it
- Leave one door open (specific resource or invitation, not a pitch)
- Reference something specific about their work to show this isn't a mass message
- Under 80 wordsI am a B2B founder building an automated outreach pipeline. Help me write a precise ICP definition.
My offer: {{offer_description}}
My best current clients: {{client_examples}}
The problem I solve: {{problem_description}}
Generate:
1. Primary ICP (title, company type, company size, industry, geography)
2. The 3 buying signals that indicate this person is in-market right now
3. The 3 pain points I should reference in outreach (in their language, not mine)
4. The 2 ICPs that look similar but are wrong for my offer (anti-ICPs)
Be specific. Avoid generalizations."I used to think the hard part was building the pipeline. It turned out the hard part was the prompts. Once the prompts were right, everything else followed."
You now have the complete self-managing GTM blueprint: the four-phase pipeline architecture, the five signal types with conversion benchmarks, the Clay enrichment table schema, the five-step sequence schedule, the complete n8n workflow node map, a 7-day launch plan, and a prompt library covering every AI touchpoint in the system.
36 calls per week from a pipeline that runs without an SDR is not hypothetical. It is running right now for Deep Loom clients. The blueprint is yours. The build is where Deep Loom comes in.
The blueprint tells you what to build. Deep Loom builds it. 36 calls/week. Automated.
Every phase in this blueprint is real, deployable, and proven across Deep Loom client accounts. The challenge is the integration complexity - wiring signal sources to enrichment, enrichment to sequence triggers, and sequence outcomes back to your CRM in a way that does not break when a prospect changes jobs or an API rate-limits you.
Daniel builds this infrastructure for B2B founders who are done with manual prospecting. The average result in 90 days: a pipeline that runs five days a week without a SDR.
- Full signal-to-sequence pipeline deployed and tested in 30 days
- Clay + n8n + PhantomBuster wired into your existing CRM
- Prompt library customized to your voice, ICP, and offer