Big Idea

The argument in one line.

Cold outreach fails not because of bad tools but because of a broken daily process — and an agent that scrapes, sequences, and self-diagnoses collapses three hours of manual work into a 15-minute weekly review.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

A service-based founder who knows cold email works but has burned out on the daily grind of writing, tracking, and following up manually.
Someone already paying for Instantly or Apollo but still spending two hours a day managing the process by hand.
A developer or builder comfortable with Claude Code who wants a working outreach agent as a starting point, not a SaaS subscription.
Anyone who has tried cold email, gotten inconsistent results, and assumed the tool was the problem rather than the targeting or copy.

SKIP IF…

You are looking for a fully managed outreach service — this requires you to build and run the agent yourself inside Claude Code.
Your business model does not involve B2B sales or high-ticket services where cold outreach economics make sense.
You are not yet clear on your offer, target audience, or value proposition — the agent amplifies clarity, it does not create it.

TL;DR

The full version, fast.

The thesis is that manual cold email is not a discipline problem but a systems problem, and the fix is an agent that removes the daily execution burden while adding a feedback loop that manual senders never have. The agent lives at /admin/outreach, built by a single Claude Code skill invocation: it scrapes 100 leads from a 300M-person B2B database, loads them into Instantly, and launches a five-step sequence. After each batch runs for a week, an AI Insight panel generates four diagnostic cards — subject line analysis, body analysis, audience fit, list quality — and a specific next-batch recommendation. The benchmark: 32% open rate and 3.4% reply rate on a live campaign, reaching 11 paying clients in 14 days at 15 minutes of human time per week.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 01:57

01 · The problem with manual cold email

Opens with the 11-client result, then empathy section: 93% of students knew cold email was the answer but could not sustain it. Calculus: every day on manual = $2,000/week in lost clients.

01:58 – 02:44

02 · Why 5 clients, not 1

Five is the validation floor. One = fluke. Two = friends doing favors. Five = proof of pattern. Five clients unlock case study material, real pricing data, and a repeatable objection pattern.

02:45 – 04:35

03 · The dashboard build

Claude Code reads a skill file and business.md, scaffolds a Next.js dashboard at /admin/outreach. Screen demo: Scrape 100 Leads (Apify, Apollo, personal emails prioritized), Load to Instantly, Launch Campaign.

04:35 – 05:30

04 · The AI Insight panel

Four diagnostic cards fire after each batch: subject line analysis, email body analysis, audience fit, list quality. Plus a specific next-batch recommendation block.

05:30 – 06:44

05 · Campaign history and weekly workflow

Campaign history table tracks every batch. Time commitment: 5 min daily, 15 min weekly, 30 min monthly. Benchmark thresholds for week 1 and week 2+.

06:44 – 07:31

06 · The two traps that kill Mission 3

Trap 1: tool shopping when reply rate is low. Trap 2: retreating to content marketing. Content marketing before mission four is procrastination dressed as strategy.

07:32 – 09:39

07 · Social proof and CTA

Student results: Refa (2 customers from 90 leads), Maria ($4,500 client in 7 days), Karen (3 clients), Kooky AI ($46K MRR). Personal confession: manual version got same 11 clients at 3 hrs/day. CTA to agentfounders.com/claude.

Atomic Insights

Lines worth screenshotting.

The AI Insight panel is what separates an agent from an email tool — it tells you what to fix after each batch so you never have to guess.
Five clients is the validation floor: one is a fluke, two could be friends doing favors, five is where a pattern becomes proof.
Switching to a new cold email tool when your first batch underperforms is almost always the wrong move — your copy or targeting is the problem, not the platform.
Content marketing before you have five closed clients is procrastination dressed as strategy.
Personal emails have higher deliverability than company emails — the scraper prioritizes them automatically.
Subject lines under six words and emails under 100 words are the non-negotiable specs for this sequence.
The manual version got the same 11 clients in the same timeframe — the agent reduced the time cost from three hours a day to 15 minutes a week.
A domain warming period means week-one daily send limits should be 10-20, not 50 — pushing too hard too early damages deliverability before the system proves itself.
An open rate below 40% means your subject lines are broken; a reply rate below 3% means your offer-to-lead match is broken — these are different problems with different fixes.
The AI insight after the first batch caught three issues the creator never would have spotted manually: subject lines too long, list too senior, sequence too aggressive.

Takeaway

Replace the daily outreach grind with a feedback loop.

WHAT TO LEARN

Most cold email failures are process failures, not tool failures — and the fix is a system that executes automatically and tells you what broke after each batch.

Manual cold email dies not from laziness but from the compounding cost of tracking, following up, and context-switching — removing those tasks is the actual leverage.
Five paying clients is the minimum threshold to trust that your offer works; anything below that is anecdote, not signal.
When a batch underperforms, the AI diagnostic cards tell you whether the problem is the subject line, the body copy, the audience targeting, or the list quality — treat these as distinct levers with distinct fixes.
Tool-switching in response to bad metrics is almost always a misdirection; the copy and targeting are the variable, not the platform.
Retreating to content marketing before you have consistent inbound closes is a pattern-break, not a pivot — it delays proof of product-market fit by months.
Knowing your own numbers (open rate, reply rate, close rate) is not vanity; it is the only way to identify which stage of the funnel is the actual bottleneck.
An AI feedback loop that runs automatically between human review cycles compounds faster than any improvement a human manually decides to make after each batch.

Glossary

Terms worth knowing.

Apify: A web scraping and automation platform used here to call a 300-million-person B2B database filtered by job title via an Apollo actor.
Instantly: A cold email sending platform that handles deliverability, domain warming, and sequence execution. The agent loads leads into it via API rather than through its web UI.
AI Insight panel: The diagnostic layer of the dashboard that runs after each batch completes its first week, generating four cards with specific recommendations for the next batch.
Domain warming: The practice of gradually increasing daily email send volume from a new domain to build sender reputation and avoid spam filters.
Humanizer: A post-processing step applied to AI-generated email copy that checks for and removes phrasing patterns typical of language model output before sending.
A/B test (sequence): Running two versions of a subject line or email body simultaneously across the lead list to compare open or reply rates and identify the stronger variant.
MCP server: Model Context Protocol server — a connection layer that lets Claude Code interact directly with external services like Supabase and Instantly without manual API wiring.

Resources

Things they pointed at.

02:50toolApify ↗

02:55toolApollo.io ↗

03:05toolInstantly ↗

03:10toolSupabase ↗

09:23linkagentfounders.com/claude ↗

Quotables

Lines you could clip.

02:37

“Every day you spend on manual cold email instead of running an agent is two clients going to someone else.”

Converts time cost to concrete revenue loss — works as a standalone hook→ TikTok hook↗ Tweet quote

07:32

“Content marketing before mission four is procrastination dressed as strategy.”

Punchy, contrarian, quotable with zero setup needed→ IG reel cold open↗ Tweet quote

08:34

“I changed all three. Next batch, double the reply rate. I never would have seen that myself.”

Concrete outcome from following the AI diagnostic — validates the product claim→ newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

Built a cold email agent inside Cloak Code that got me 11 paying clients in fourteen days. Not a cold email template, an agent. A web dashboard with three buttons.

Button one scraps 100 leads from a 300,000,000 person database. Button two loads them into instantly.

Button three launches a five step email sequence with a b testing that runs for fourteen days automatically. I checked the AI insights once a week. The agent handles everything else.

And the part that surprised me most was not the 11 clients. It was what happened when I read the AI insight panel after the first batch. I will show you that at the end.

Hi. My name is Chris. I built a $60,000 a month AI education business.

Forty three percent of students never got result. I killed it. I started 200 founders.

The ones who made it has one thing in common. They did cold outreach every day. The problem is most people cannot sustain cold outreach manually.

So I built an agent that does it for you. First system, you approve, you send, let's go. Tell me if this sounds like you.

You know, you need to do called email. You open Gmail, you write one email. Fifteen minutes, you send it, you write another one.

By 10AM, you sent four emails and you are exhausted.

By Thursday, you lost track of who you emailed. Your Google seat has 200 rows and 14 columns and you do not know which ones are wrong.

Or you signed off for instantly, but you still spend two hours a day writing each email by hand, checking who replied, managing follow ups manually. The tool is there, but the process is still manual. 93% of my old students knew cold email was the answer.

They could not sustain it. Not because they were lazy, because the process was broken. Right?

Send, track, follow-up, repeat every day. For weeks, the manual process kills momentum before the outreach gets a chance to work. Every day you spend on manual cold email instead of running.

An agent is two clients going to someone else. $2,000 a week, not money you're missing out on, money you are losing.

I was that person. Google Draps, sticky notes, Google Seats with 14 columns, Then I built the agent. The target is a five clients, not one.

One client is a fluke. Two could be friends doing flavors. Five is the floor where I might have something becomes this works.

So five client unlock three things, case study materials, five different stories instead of one. Real pricing data, what price got accepted, what got pushed back, what made them lost, and a pattern, the same objections, the same words that made them lean in.

That pattern becomes your script, your landing page, your YouTube hooks. And the cold outreach agent, one skill, just you're gonna type slash the cold outreach agent inside your folder.

Cold reads the skill file, reads your business dot markdown file and builds a web dashboard at admin and outreach. And you're instantly EpiFi and super base MCP servers are already connected from mission one, so there is nothing new to wire up. You review approve and press the first button, same rules as the mission.

The top of the dashboard, scrap 100 leads button. Behind it, FFI is calling a 300,000,000 plus b two b database, Apollo actor filtered by job title.

And here's my log, 1,000 lead, $1, and I scrap YouTube and Google Map, email extractor, everything inside here, and LinkedIn to email too.

Personal emails prioritized because they have higher delivery than company emails. Result auto saved to super base.

And then I click it and thirty seconds later, 100 leads in my table Like this, personal email, company email, personal email.

And here is a five step email sequence below the scrap button, the sequence viewer, five clickable steps showing full email contents. This is my first email sequence a b test. First name, tired of AI treadmill, and second, the tool treadmill is killing your progress.

Every email is under 100 words, subject lines under six words. Written from your vizis. Markdown file voice, humanizer checked, no AI slope.

In my email sequence right now, 32% open rates and 3.4% reply rate. Three buttons that is hold, to flow.

Select your daily sending limits, 10 or 20 or 30, account per day, load to instantly, button transfers leads from super base to instantly via API. And then finally, after finish this, launch campaign button activates the instantly campaign.

Scrap, load, launch, three clicks, your cold outreach is live. And here is AI insight panel. This is a part that makes this an agent, not just an email tool.

After each batch completes its first week, AI generates four diagnostic cards, subject line analysis, checks open rate. If below 40%, it suggests three alternate subject lines.

An email body analysis check. Text reply rate identifies what feels too salesy, too long or too generic. Target audience fit and list quality.

And then it gives a next batch recommendation. Keep subject line a. Rewrite body paragraph two.

Filter for companies under 50 employees. Increase daily limits to 30, like having a media buyer reviewing your campaign every week and telling you exactly what to fix.

And here's a campaign history every batch recorded, date, search query, lead account, open rate, reply rate, bounce rate, what you changed in status.

You look back at 10 batches and see exactly what improved and why. Below that, the lead list, air table style table, an email name company type, current step, step status, and date. And this is a fifteen minutes a week.

Morning, five minutes open slash admin slash outreach. Check if any replies came in. If someone wants a call, respond.

That is it. And weekly, fifteen minutes, read an AI insight from the last batch. Check performance, scrap next 100 leads with adjusted filters.

Load, learned, done. Monthly, review all batches. Identify your winning subject line and sequence.

Update the sequence in your config. Scale daily, send limits if metrics are Korean. So week one, your validation sprint 100 leads scrapped.

Daily send limits 10 to 20 per account because your domain is still warming up. Open rate above 40% of your subject lines are broken. Reply rate above 3% of your offer to lead match is broken.

At week two and beyond, 100 to 200 leads per batch. Daily limit 30 to 50.

Target two to four course booked per week from positive replies. One to two closes per week at 30 to 50% close rate on held call. And the most common bottleneck is copy.

Your cold email is too long to generate or too polish it. The AI Insight Panner catches this. Rewrite it shorter and more human.

Run the next batch, check the delta. And there is the two traps that kill mission three. Trap one, tool shopping.

Your first batch gets one percent reply rate and you start googling best cold email tool 2,026. You sign up for two new services. Stop.

The tool is not the problem. You're copying or you're targeting is. The AI insight panel already told you what to fix.

Fix that. Not the tool. Step two, the fallback to content marketing.

You run two batches, low replies. You decide to call the email instead and start posting on LinkedIn instead. Three months later, 40 followers, zero clients, content marketing before mission four is procrastination dressed as strategy.

Fix your cold email, run another batch, graduate to mission four when you have five closes. Finally, before moving to mission four, five things, five closed deals, real names, real dollar amounts, payment received, one detailed case study from your high ticket close, Know your own number.

My open rate is x percent, my reply rate is y percent, my close rate is z percent. Your dashboard is live at slash admin at outreach. And you can describe the system in one sentence.

I scrapped 100 leads, loaded two instantly, and the agent sends a five step sequence. I check AI insights weekly. Hit all five and mission three is done.

Here's what happened to me. I run the manual version first, Google, sticky notes, three hours a day for two weeks, 11 clients. Then I built the automated dashboard.

Same results, fifteen minutes a week instead of three hours a day. And the AI insight I promised you at the start, after my first batch, the pander said, your subject lines are too long, your target list is too senior, Your sequence is ascending too aggressive. I changed it all three.

Next batch, double the reply rate. I never would have seen that myself. The agent is solving.

Rifa spent $10,000 on courses. Zero clients built the agent. 90 leads scrapped.

Two paying customers. Maria built an agent, loaded her warm network first. $4,500 client in seven days.

Karen, thousand emails through the agent in six weeks. Three clients. The AI inside panel turned her failed first batch into feedback that made batch three work.

Cookie AI, same system at scale, $46,000 MRR. That is the cold outreach agent, one dashboard, three buttons, a five step sequence that runs for fourteen days.

AI insights that tell you what to change. You scrap, you load, you learned, the agent sends. If you want the full five mission road map and the three core skills that build this agent, Go to the agentfounders.com/cloud.

Free. Link below. I built this for the person I used to be, the one with the Gmail drafts and the sticky notes.

See you in the next one.

The Hook

The bait, then the rug-pull.

Eleven paying clients from fourteen days of running an agent. Not a template, not a sequence — an agent that scrapes, sequences, and tells you what to fix, while you spend fifteen minutes a week reviewing the output instead of three hours a day grinding through Gmail drafts.

Frameworks

Named ideas worth stealing.

01:58concept

The Five-Client Floor

One client = fluke
Two = friends doing favors
Five = proof of pattern

Do not declare product-market fit until you have five paying clients from outreach. Five unlocks case study material, real pricing data, and a repeatable objection pattern.

Steal forany founder deciding when to scale outreach or build a sales page

04:19list

Scrape / Load / Launch

Scrape 100 leads (Apify to Apollo, personal emails prioritized)
Load to Instantly (Supabase to API)
Launch campaign (five-step sequence, A/B subject lines)

The three-click cold outreach workflow. Each step is a button in the dashboard; the agent handles execution, the human reviews approval.

Steal forany outreach system where the manual loop is killing momentum

04:35list

Four AI Diagnostic Cards

Subject line analysis (open rate < 40% triggers 3 alternate suggestions)
Email body analysis (too salesy / too long / too generic)
Target audience fit
List quality + next-batch recommendation

The feedback loop that makes this an agent rather than a tool. Fires automatically after each batch's first week.

Steal forany AI-assisted workflow where you need automated diagnostic output between human review cycles

05:30model

15 Minutes a Week

5 min daily — check replies, book calls
15 min weekly — read AI insight, scrape next 100 leads with adjusted filters
30 min monthly — review all batches, update sequence, scale send limits

The target time budget once the agent is live. Compared to 3 hours/day manual as the baseline.

Steal forany automated workflow where communicating the ongoing maintenance burden is part of the value proposition

CTA Breakdown

How they asked for the click.

VERBAL ASK

09:23link

“If you want the full five mission road map and the three core skills that build this agent, go to agentfounders.com/claude. Free.”

Soft and non-pushy — framed as a gift to the person he used to be. URL shown in captions. Effective because it directly extends the video content rather than pivoting to a sale.

MENTIONED ON CAMERA

02:50toolApify ↗

02:55toolApollo.io ↗

03:05toolInstantly ↗

03:10toolSupabase ↗

09:23linkagentfounders.com/claude ↗

FROM THE DESCRIPTION

PRIMARY CTAWhere the creator wants you to go next.