The 10-AI Agent GTM Playbook

The exact 10-agent AI system that books 18+ qualified meetings a week — from signal detection to closed-loop pipeline analytics.

A Quick Introduction

I'm Shabbir, co-founder of Coolie and a 2x founder with 11 years in sales. I've sent, received, and analysed over 628,000 cold emails and LinkedIn DMs — enough to know that the way most teams still do outbound stopped working around 2024.

Outbound itself is alive. 75% of my qualified meetings this quarter came from email and LinkedIn — using the exact system in this playbook. What's dead is the spray-and-pray version of it: cold lists, templated personalisation, and "Congrats on the funding!" openers.

B2B Founders and GTM teams don't need another tool. They need a system where AI finds prospects, researches them, writes the outreach, and learns what's actually generating pipeline.

The unlock isn't more AI skills. It's getting specialised agents to work together.

This playbook gives you the exact 10-agent AI org chart that replaced what used to take a team of 5-10 people. Your AI (Claude, ChatGPT, Grok, or whichever LLM you prefer) does the thinking. Coolie does the execution. You approve and close.

Read this end to end and you'll have the same setup my team uses to book 18+ qualified conversations a week.

→ If you'd rather skip ahead and try the tool that automates it: trycoolie.ai (7-day free trial)

A quick video that shows everything in less than 2 minutes.

https://youtu.be/aJRoYiJf5YM

The System Architecture

Before anything else, understand how the pieces fit. This is the difference between a system that runs and a hacky mess.

Your LLM = the intelligence layer. Reasons about your ICP, reads signals, writes messages, adjusts follow-ups. Use whichever model you prefer — Claude, ChatGPT, Grok, or any LLM that supports MCP or API connections.

MCP = the connection layer. The protocol that lets your AI talk directly to other tools, no copy/paste.

Coolie = the execution layer. Your LLM can't connect to LinkedIn directly. Coolie bridges the gap, manages LinkedIn limits, runs campaigns across email, LinkedIn, WhatsApp and calls.

You = the validation layer. You set the strategy. You approve qualified leads, review messages before sending, take the calls. You focus on building relationships and closing deals.

Each block stays in its lane. That's what makes the system robust.

The 10-Agent AI GTM Org Chart

Here are the 10 specialised agents and how they chain together — from finding a prospect to closing the loop on what's working.

No massive sales team. No 15 disconnected tools. No spending all day researching prospects. One coordinated AI GTM system.

The flow looks like this:

→ Signal Agent finds prospects and monitors accounts
→ Enrichment Agent gets contact details
→ Scoring Agent qualifies them
→ Research Agent uncovers why they'd buy now
→ Copywriting Agent writes the outreach
→ Sequence Agent orchestrates multi-channel delivery
→ Inbox Agent handles replies
→ CRM Agent manages the pipeline and reactivates warm leads
→ Content Agent turns winning angles into inbound content
→ GTM Brain analyses what's working and feeds it back into the system.

That's the loop. Each agent below gets its own section with the framework, prompts and workflows to run it.

Agent 1 — Signal Agent: Find and Monitor Prospects Who Are Ready to Buy

This is where the whole system starts. Most teams waste 90% of their time contacting people who aren't ready to buy. A message sent at the right moment converts 3-5x better than a perfect message sent at the wrong time.

The Signal Agent does two jobs: it finds new ICP-fit prospects using 40+ buying signals via Coolie, and it continuously monitors every account — active prospects, dead sequences, even disqualified leads — for new signals that create a fresh reason to reach out.

Here are the 7 most important signals to build your prospecting motion around.

1. Technology Used (Competitor)

Companies already using a competitor's product understand the problem and have budget allocated. Show them how your solution is better, faster, or more cost-effective.

2. Competitor Churn or Trial Ended

Companies that recently cancelled a competitor's subscription or finished a trial. The problem is fresh in their mind. Reach out while they're still evaluating alternatives.

3. Competitor Renewal

Companies approaching a competitor's renewal date. Reach out before they commit to another contract.

4. Recent Funding

New capital means new budget. Priorities and tool stacks get re-evaluated in the first 90 days post-announcement.

5. Recent Job Change (< 90 days)

Decision-makers who just joined or got promoted are reviewing existing vendors and looking for quick wins. Help them hit their first-90-days priorities.

6. Buying Intent Topics

Companies engaging with content, product comparisons, or searches related to your solution are telling you what they're working on.

7. Hiring for Key Roles

Hiring multiple SDRs = scaling outbound. Expanding CS = needs customer management tools. Connect the hire to the business challenge it signals.

Key takeaway: Don't just target companies that fit your ICP. Target ones showing signs they need your solution right now.

🧠 Full Signal Library → Click Here To Read 🧠

How to set it up

Open a chat with the Coolie connector enabled (setup is in the appendix) and tell your AI:

Here is my ICP: [paste the ICP]
Set up continuous monitoring in Coolie to detect these 7 signals:
[list the 7 signals]
Prioritize signals X, Y, and Z (strong active intent).
Notify me as soon as a prospect has these signals.

From there, the Signal Agent surfaces hot leads automatically — and keeps watching every account for new signals even after first outreach. When a prospect who went cold suddenly shows a new signal (new funding, leadership change, competitor churn), you get notified with a fresh reason to re-engage.

The re-engagement prompt

When the Signal Agent surfaces a new signal on a dead prospect:

Prospect: [name + role + company]
Previous outreach: [summary of what was sent and when]
Previous tension anchor: [what the old sequence was based on]
New signal detected: [the new signal]

Write a completely fresh opening message based on the NEW signal.
Do not reference the previous outreach.
Follow the 4-block framework.
Max 60 words. Human tone.

This is how you turn a "no" into a "not yet" — and "not yet" into a meeting when the timing is right.

Agent 2 — Enrichment Agent: Get Verified Contact Details

Finding a prospect is useless without a way to reach them.

The Enrichment Agent waterfalls across 15+ data providers to find verified emails and phone numbers for every prospect the Signal Agent surfaces. It runs automatically inside Coolie — no manual enrichment needed.

What it does:

• Tries multiple data sources in sequence until it finds a verified email
• Cross-references phone numbers across providers
• Flags contacts where verification confidence is low
• Updates contact data if it goes stale

You never need to buy separate enrichment tools or manually copy/paste from LinkedIn. By the time a prospect hits your review queue, they already have contact details attached.

Agent 3 — Scoring Agent: Qualify Before You Spend Time

Before you ask your AI to find prospects, it needs to know exactly who you want to reach.

Define your ICP first

The step 90% of people rush. And the #1 reason "AI campaigns" produce mediocre results.

Here is my offer: [describe your offer in 3 lines max]
Here is my average deal size: [amount]
Here are my last 5 signed customers: [list with role + industry + company size]

Build my ideal ICP from this data.
Format: exact role, company size, industry, likely buying signals,
common objections, decision triggers.

Be extremely specific. No generic answers.

Your LLM outputs a strong ICP in under 30 seconds. Save it — you'll reuse it in every conversation.

Score every prospect

Once your ICP is defined, the Scoring Agent qualifies any prospect in one message:

Here is a prospect: [LinkedIn profile or name + company]
Here is my ICP: [paste from above]

Score this prospect 0-10 on:
- ICP fit (role, company, industry)
- Buying maturity (intent signals detected)
- Accessibility (likelihood of reply)

Final verdict: contact now, nurture, or discard.
Justify in 2 lines max.

Run this on dozens of profiles in one session. It's your first filter.

Finding prospects at scale

Once ICP + signals are set, ask your AI to find matching prospects via the Coolie connector. They show up automatically in your Coolie account, enriched with the intent signals from Agent 1 and the contact details from Agent 2.

Agent 4 — Research Agent: Reference Tension, Not Events

You've got high-intent, qualified leads with contact details. Now the harder question: what do you actually say?

Most reps stop at "Congrats on the funding!" That's why most reps live at ~1% replies.

The funding round isn't the message. The pressure that funding round creates for your specific buyer — that's the message.

I call this 3rd-Level Research.

The 3 Levels of Insight

Level 1 — The Observable Event. What happened. The funding round, the hire, the launch. Every seller sees it. Every seller sends the same congratulatory line. Nobody responds.

Level 2 — The Strategic Intent. What they're doing with it. Where the money is going, what the hire signals about direction. You have context — but your prospect already knows all this. You haven't earned a reply.

Level 3 — The Tension It Creates. What's keeping them up at night right now. The predictable internal friction this move creates for this role at this company. The conversation they're already having with their boss. Walk straight into that conversation and you stop pitching. You start contributing.

A real example: Hireflow raises $12M

Level 1 (what everyone sends):

"Congrats on the round! Would love to chat." → ignored.

Level 2 (slightly better):

"Saw you're going upmarket — let's chat." → still ignored.

Level 3 (books the meeting):

"Hi John, read the Forbes piece on your enterprise push. As you move into Fortune 1000, are you seeing more resistance from buyers already locked into Workday or SuccessFactors than from competing point solutions? It's becoming the harder unlock for the teams we work with."

You're not referencing the funding. You're referencing the friction.

Two ways to run 3rd-level research

Same framework either way. Pick the path that matches your volume.

Path 1 — Manually with your LLM

Prospect: [name + role + company]
Signal detected: [paste signal from Coolie]
My offer: [2-line summary]
My ICP: [paste ICP]

Step 1 — Identify the Level 3 tension this signal creates for
THIS role at THIS company. Not what happened. Not what they're
doing with it. The internal friction it creates for this persona
this week. Answer why me, who now in the sequence

Give me 3-5 specific tensions, ranked by how much it would
dominate this person's thinking right now.

Expect ~15 minutes per prospect. Great for your top strategic accounts.

Path 2 — Automated with Coolie's Research Agent

For every prospect, the Research Agent runs the same framework in the background: 300+ contextual data points, scored against your value prop, top 3-5 insights surfaced with a "why this matters right now" explanation per persona.

Click "Research" → "View" on any prospect, or ask your AI in an MCP chat:

Run deep research on the leads in my current campaign.
Surface the top 3 tensions for each one, ranked by relevance
to my value prop.

Same framework. Same quality of thinking. Just fast enough to run across your full pipeline.

My own week: Research Agent on the full pipeline, 20 minutes a day reviewing top tensions, manual prompt on 2-3 strategic accounts where I want to think hard myself.

Agent 5 — Copywriting Agent: Turn Research Into Messages

Now that research is done, the writing is almost mechanical.

The 4-block message framework

A great LinkedIn message in 2026 doesn't look like SDR copywriting. It looks like something a thoughtful human would write to another thoughtful human.

Block 1: Observed trigger (1 line)

The precise signal that made you reach out. Not "I saw your profile." Something concrete.

"Saw your post about technical sourcing pain at Series A startups."

Block 2: Connection to the tension (1 line)

Show you understand the friction, not just the event.

"The teams I work with hit the exact same wall around month 4."

Block 3: Open question (1 line)

No pitch. A real opening.

"Curious how you're handling it internally right now."

Block 4: Soft exit (1 line)

Remove pressure. Keep dignity.

"No reply = no problem, I'll assume it's not a fit right now."

The generation prompt

Prospect: [name + role + company]
Level 3 tension (from research): [paste tension]
My offer: [2-line summary]

Generate 3 opening message variants following this structure:
Block 1: Concrete observed trigger
Block 2: Connection to the tension
Block 3: Open question (no pitch)
Block 4: Soft exit

Max 60 words per message. Human tone. No corporate jargon.
Reference the tension, not just the event.

Pick the one that fits best. Send via Coolie.

The follow-up sequence

Most people send one message, wait, drop a generic "just following up" five days later. Result: 2% reply rate. Here's what actually works.

Day 0 — Opening message. The 4-block structure above.

Day 2 — Contextual follow-up. Not "bumping this." Bring something new tied to the same tension: an article, a stat, a sharper question.

"Just found this study on exactly the Workday lock-in problem [link]. Curious what you think."

Day 5 — Breakup. Signal it's your last attempt without guilt-tripping. Often triggers a reply out of politeness or curiosity.

"I won't push further. If this comes back on your desk, my door's open."

Day 30 — Signal-based re-entry. If the Signal Agent detects a new intent signal from that prospect weeks later, your AI re-engages with a completely new hook. Not a continuation of the dead thread.

The sequence prompt

Prospect currently in sequence: [name + message history]
New signals detected since last message: [signals]

Generate the next step based on timing:
- > 9 days with no reply → breakup message
- New signal detected → full reset with a new hook
- Otherwise → contextual follow-up on the same tension

Your AI manages timing. You approve.

Agent 6 — Sequence Agent: Multi-Channel Orchestration

A stack that only runs on email misses 30-40% of opportunities. Many decision-makers check LinkedIn & read email daily.

The point isn't to spam across channels. It's to follow the same conversation on the channel where the prospect actually is.

The orchestration principle

→ First touch on LinkedIn (where the signal was detected)

→ If no reply after 2 messages, switch to email

→ Cold call in between a full email + LinkedIn cycle

The key: every channel references the same Level 3 tension. Different angle, same conversation. One voice across four surfaces, not four campaigns in one inbox.

The 7-day multi-channel cadence that works

• Day 1 AM — LinkedIn connection request, no message
• Day 1 PM — Email #1: tension + question, ~60 words, no pitch
• Day 2 — LinkedIn DM (if accepted): conversational, same tension
• Day 3 — Cold call. Opener references prior touches and the same tension. Objection handlers tied to the same friction.
• Day 4 — Email #2: new angle on the same tension (not "bumping this")
• Day 5 — Follow-up on LinkedIn #2 new angle on the same tension (not "bumping this")

Hand-writing this for one prospect takes ~20 minutes. For 30 prospects a day it's not happening manually.

Coolie's Sequence Agent takes the top-ranked tension from Agent 4 and writes the full 7-day multi-channel sequence (email + LinkedIn) — all anchored to the same tension. You edit and approve.

The orchestration prompt

Prospect: [name + email if available + LinkedIn]
Messages already sent and status: [history]
Prospect's recent activity: [LinkedIn activity + email opens]
Level 3 tension anchor: [paste tension]

Recommend the next channel and the next message:
- LinkedIn inactive > 7 days → switch to email
- Email opened 2x no reply → back to LinkedIn, fresh angle on same tension
- During the email & LI follow-up → cold call script with same anchor
- Otherwise → continue on current channel

Coolie handles sending across email and LinkedIn. Your AI orchestrates the decision.

Agent 7 — Content Agent: Turn Outbound Into Inbound

Your outreach generates data that most teams ignore. Which tensions get replies? Which angles book meetings? Which objections come up most? The Content Agent turns those insights into content.

The flywheel

→ Outbound teaches you what your market cares about
→ Content Agent turns that into posts
→ Posts attract inbound
→ Inbound feeds your pipeline
→ Pipeline data makes outbound smarter.

Most founders treat outbound and content as separate activities. They're the same system.

What the Content Agent produces

• LinkedIn posts based on the tensions and objections you're hearing in outbound replies
• X/Twitter threads breaking down the problems your prospects actually describe

The content generation prompt

Here are the top 5 outbound messages that generated replies this week:
[paste messages + the tensions they referenced]

Here are the top 3 objections prospects raised:
[paste objections]

Generate 3 LinkedIn post ideas that:
- Reference the tensions without naming specific prospects
- Position my expertise around these problems
- End with a question or observation that invites engagement
- Match this tone of voice: [paste 2-3 of your existing posts]

Max 150 words per post. No hashtag spam. No corporate jargon.

The rhythm

Run this weekly. 20 minutes reviewing what worked in outbound → 3 posts for the week. Your best outbound angles become your best content. Your best content attracts prospects who already understand the problem.

Agent 8 — CRM Agent: Manage Your Pipeline

Sequences generate interest. The CRM Agent makes sure that interest turns into pipeline — and that warm leads don't die from neglect.

What it does

Pipeline management:

• Tracks every prospect's stage: new lead → contacted → replied → meeting booked → opportunity → closed
• Flags deals that stall — prospects who booked a call but went quiet, or opportunities stuck at the same stage for 2+ weeks
• Surfaces the next best action for every deal in your pipeline

Warm lead reactivation:

• Identifies prospects who showed interest but went quiet (replied positively, opened emails 3+ times, clicked links but never responded)
• Pairs with the Signal Agent — when a warm lead shows a new signal, the CRM Agent bumps them to the top of your review queue
• Drafts reactivation messages based on new context, not generic "checking in" messages

The pipeline review prompt

Run this daily (5 minutes):

Pull my current pipeline from Coolie.

For each stage, show me:
1. How many prospects are at each stage
2. Which deals have been stuck at the same stage for 7+ days
3. Which warm leads have new signals since they went quiet

For stuck deals: recommend the next action (follow up, new angle,
breakup, or move to nurture).
For warm leads with new signals: draft a reactivation message
using the new signal. Max 50 words. No guilt.

The reactivation prompt

Here are prospects who replied positively 2-4 weeks ago
but haven't scheduled a call:
[list from Coolie]

For each one:
- What was the last thing they said?
- Any new signals detected since?
- Draft a reactivation message that references their original
  interest + any new context. Max 50 words. No guilt.

The CRM Agent keeps your pipeline honest. No vanity metrics, no deals sitting in "interested" for 6 weeks. Every prospect either moves forward or gets an honest next step.

Agent 9 — Inbox Agent: No Reply Falls Through the Cracks

When sequences are running across email and LinkedIn, replies come in from multiple channels at different times. The Inbox Agent makes sure you respond to the right people, fast.

What it does

Reply categorisation:

• Scans every reply across email and LinkedIn in real time
• Categorises each one: hot lead / warm / objection / not interested / auto-reply / out of office
• Flags hot replies (interest, questions, meeting requests) for immediate attention
• Alerts you within minutes when a high-value prospect replies

Response drafting:

• For hot leads: drafts a response that moves toward a call — short, direct, no over-selling
• For objections: identifies the objection type (budget, timing, authority, need) and drafts a response that addresses it without being pushy
• For "not now" replies: tags the prospect for the CRM Agent to track and reactivate later

The reply handling prompt

Here are today's replies across all channels:
[paste replies or let Coolie pull them]

For each reply:
1. Categorise: hot lead / warm / objection / not interested / auto-reply
2. For hot leads: draft a response that moves toward a call. Max 40 words.
3. For objections: identify the objection type and draft a response
   that addresses it without being pushy. Max 60 words.
4. For warm leads: recommend next action and timing.
5. For "not now": tag for CRM Agent reactivation with the reason
   they gave and the best re-entry window.

The daily rhythm

Morning: Inbox Agent pulls overnight replies → you review hot leads first → approve or edit responses → send via Coolie. Takes 10 minutes.

The difference between the Inbox Agent and the CRM Agent: Inbox handles the immediate response. CRM handles what happens after.

Agent 10 — GTM Brain: Learn What's Actually Working

This is the agent that makes the whole system smarter over time. Without it, you're running campaigns. With it, you're running a learning machine.

What it analyses

Signal effectiveness:

• Which of the 7 signals produce the highest reply rates?
• Which signals lead to meetings vs just replies?
• Which signal combinations predict closed deals?

Message performance:

• Which tension angles get the most responses?
• Which 4-block variants outperform?
• What message length works best per channel?

ICP accuracy:

• Which ICP segments convert at the highest rate?
• Where are you getting replies but no meetings? (ICP is too broad)
• Where are you getting meetings but no closes? (messaging is off, or wrong persona)

Channel performance:

• Email vs LinkedIn vs cold call: which books more meetings per ICP segment?
• What's the optimal channel sequence for each persona?

The analysis prompt

Run this bi-weekly:

Pull my outbound performance data from Coolie for the last 14 days.

Analyse:
1. Reply rate by signal type — which signals drive the most engagement?
2. Meeting rate by tension angle — which research insights convert?
3. Reply rate by ICP segment — where am I wasting effort?
4. Channel performance — where should I spend more/less time?
5. Message performance — which variants win?

Give me:
- Top 3 things to double down on
- Top 3 things to stop or change
- One ICP adjustment recommendation based on the data

Be specific. Numbers, not generalities.

The feedback loop

The GTM Brain's output feeds directly back into the system:

• Signal Agent gets re-prioritised based on which signals actually convert
• Scoring Agent gets recalibrated based on which ICP segments close
• Copywriting Agent gets updated based on which angles work
• Sequence Agent adjusts channel mix based on what's booking meetings
• Content Agent gets the winning angles to turn into posts

This is the loop that turns a campaign into a compounding system.

Your First Week

Follow this order. No "9 calls in 2 days" promises — that comes later, after the system has data to learn from.

Day 1 (30 min)

• Run the MCP + Coolie setup (appendix)
• Build your ICP with Agent 3

Day 2 (45 min)

• Turn on the 7 intent signals with Agent 1
• Let it run 24 hours so first leads come in

Day 3 (1 hour)

• First qualification session: your AI outputs 20-30 leads
• Keep 10-15 priority leads
• Run 3rd-level research manually on your top 5 (Agent 4)
• Launch the sequence for 7-10 of them (Agent 5 + 6)

Days 4-5

• Watch the first replies (Agent 9 flags them)
• Approve the follow-ups your AI proposes
• Review your pipeline in the CRM Agent (Agent 8)
• Adjust your ICP if needed

Days 6-7

• Scale to 20 new prospects in sequence
• Turn on email + LinkedIn integration for non-responders
• Run your first content session — turn winning angles into 2-3 LinkedIn posts (Agent 7)

After one week — a system running.
After one month — 18+ qualified meetings a week, consistently.

At the end of week 2, run your first GTM Brain analysis (Agent 10). That's when the system starts learning.

Appendix — MCP + Coolie Setup (5 minutes)

The one-time technical setup. Do this when you're ready to run the system — not before you've decided it's for you.

You need:

• An AI account (Claude, ChatGPT, Grok, or any LLM that supports MCP connectors)
• A LinkedIn account
• A professional email account
• A Coolie account (free trial available)

The setup below uses Claude as an example. The process is similar for other LLMs that support MCP.

Step 1 — Open your AI

Go to claude.ai or open Claude Desktop. Click "Customize" or go directly to claude.ai/customize/connectors.

Step 2 — Add a custom connector

Click "Connectors" → the "+" → "Add custom connector."

Step 3 — Add Coolie

Name it "Coolie AI" and paste this URL:

https://trycoolie.ai/api/mcp

Click Continue.

In your Coolie account go to Settings → API Key → Create New Key. Copy and store it.

Back in your AI: in request headers, select x-api-key and paste your API key. Click Add,

then click on Allow.

Step 4 — Enable the connector

In any chat, click the "+" at bottom-left and enable Coolie AI under connectors. Or go here and click "Connect" next to Coolie.

Done. Your AI now has read/write access to your Coolie account. Head back to Agent 1 and start running the system.

# Get Started

→ Create a free Coolie account: trycoolie.ai (7-day trial)

Most teams have signals running through their pipeline within 10 minutes of signing up. If you want a hand setting it up, message us inside the product — our team replies in under an hour and walks you through your first campaign live.

Built for B2B SaaS founders prospecting solo, GTM and Sales teams running outbound at scale, and sales leaders who want their reps spending time on conversations instead of research tabs.

Happy prospecting. 🚀

— Shabbir, Co-founder, Coolie

Real Results From Teams Running This System

1. How Recruiterflow's Sales Team Tripled Reply Rates in One Week

→ Click here to read the full case-study

2. How Datatracks 6x Email Reply Rates and Gave ~208 Hours

→ Click here to read the full case-study