You sent the connection request. They accept, and you send a message.

Then one of two things happens.

1. Nothing comes back at all. 😔

2. Or they do reply, and you freeze, because now what? 🤔

That second one is where almost all LinkedIn outreach dies.

Someone actually answered you. They’re paying attention for maybe 30 seconds.

And you either jump straight to pitching a call and sound like every other salesperson in their inbox.

Or you say something vague and friendly and watch the whole thing fade out over the next week.

I did that for years. And here’s what changed:

46 meetings from 1,938 connection requests. Six months, about 12 requests a day. 2.37% meeting rate.

These were all ‘cold’ prospects. Fresh connections, pulled from LinkedIn signals. No warm intros, referrals, or inbound.

DM people who already know you, past clients, dormant contacts, your existing network, and the rates go further up.

LinkedIn Outreach Benchmarks 2026 tracked 180,155 connection requests sent between January 2025 and August 2026.

Across that whole dataset, the average is 5.4 meetings booked per 1,000 requests sent. Just over half a percent.

At that rate, booking 46 meetings takes about 8,518 requests. I did it with 1,938, roughly a quarter of the volume.

Learn everything about how to get your connection requests accepted here:

Now, today is about knowing exactly what to say once someone replies, until booking the meeting.

Not just a script you copy.

It’s a conversational strategy, where every message has one job, and you always know which one you’re on.

Years of getting this wrong, until I got it right

It took me a few years to get here. Now you can apply it in 30 minutes.

Every rule in the system below exists because an earlier version of it cost me a conversation.

And it’s all packaged into the skill you can save further down.

But I’m not handing you just a button to press blind.

Because it’s a bad idea to automate something you don’t actually understand.

So before the quick fix, here’s the entire strategy laid out in full.

Read it first. Then let the skill run it for you:

Read more

The SaaS companies crushing it right now have one thing in common. And it’s not AI.

Replit, Lovable, Databricks, Supabase,.. all growing fast, and none is dying.

Meanwhile Salesforce cut 4,000 jobs last year.

Workday’s CEO spent three quarters changing his story from “customers renewing at lower headcount” to “we’re not focused on seats anymore.”

Jason Lemkin listed these survivors, but didn’t explain what they share that makes them survive.

I’ll do it here. NONE of them sell seats:

• Harvey sells legal work done.

• Sierra and Decagon sell resolved tickets.

• Replit and Lovable sell working software you describe in plain English.

• Databricks, Supabase, Vercel sell infrastructure the AI-native economy runs on.

Every single one is priced around outcomes or usage, instead of humans with logins.

‘Old’ SaaS assumed humans would be the primary operators: Clicking, configuring, updating dashboards.

$50 per seat, more seats = more revenue.

That worked for 15 years. Then AI started compressing headcount at the biggest SaaS company on the planet.

Where This Shows Up in Your GTM Stack

You’re not switching out Harvey or Databricks this week.

But somewhere in your own GTM tools, this same thing is probably already costing you, and there’s usually one clear fix.

Clay works the old way. You buy credits.

Even the newer stuff, like using Clay’s tools from inside Claude instead of opening Clay’s own dashboard, still needs a person there triggering it, one small batch at a time.

Every call burns credits whether it needed to happen right now or not.

Claude Code works the new way. You own it.

You pay for what actually gets done, research finished, lists built, messages written, not for logins or API calls.

That covers a big chunk of what GTM tools charge you for: finding leads, researching them, writing outreach.

It’s not a swap for your CRM, phone system, or support tools.

The teams I work with are cutting their Clay bill by 60%.

They move research to Claude Code, and Claude calls Clay directly when enrichment is the right job for it, instead of running everything through Clay by hand.

If you’re paying per seat or per credit for anything, that clock started ticking about a year ago.

Worth checking before your next renewal.

Jasper

PS: Want the Claude Code setup founders use to replace $800/month Clay workflows (MCP stack + CLAUDE. md template)?

Get it right here.

Before sending any LinkedIn DM, you need to be able to answer 3 questions.

Good outreach starts before you write the first line.

If you don’t filter for fit and context, no message will land.

Here’s the typical DM on LinkedIn everyone gets, and everyone ignores:

Here’s exactly why it doesn’t work:

❌ Zero targeting.

“I saw your content” tells you nothing, it’s the connection-note equivalent of “hope this finds you well.”

❌ It reads like an email.

Four dense paragraphs is way too long for a first message on LinkedIn. You already give up reading when you see a wall of text. Chat is not email.

❌ It’s only about them.

Every line pitches their company. Nothing about you or why you’d care.

❌ It’s a pitch, dressed as a compliment.

The “we’d love to have you” line is the ask, wearing a friendly hat.

❌ There’s no good offer.

Nothing compelling, nothing useful, nothing to say yes to before being asked for something.

3 Crucial Questions

These 3 checks save you from wasting effort on people who will NEVER engage.

1. Are they active on LinkedIn?

Plenty of people have profiles but hardly log in.

If they haven’t posted or reacted in weeks, they likely won’t accept your connection request.

You already filtered for this when building your lists.

The Sales Navigator “posted recently” check separates people who might engage from profiles that never will.

2. Are they open to connecting?

Get 50%+ of Your Connection Requests Accepted answered this one.

Of course, you can only DM someone if you are connected (unless you send InMail).

Your profile and targeting do most of the work here.

When your headline tells people exactly what you do, for whom, and with what outcome, acceptance rates climb well past 50%.

Those new contacts are already pre-qualified.

3. Are they open to talking?

Getting connected isn’t the same as starting a conversation.

You need messages that bring up a relevant business problem and eventually lead toward a call.

Most people freeze here and send something generic that goes nowhere.

Nobody’s sitting there hoping a stranger messages them.

Your job isn’t to get everyone talking. You just need to find the small number of people who actually want to.

If your connection request had a personal note, pick that topic back up.

If it was blank, wait at least 24 hours first. Some people will message you first in that window, either to sell you something or to ask what you do.

Either way, you’ll know instantly if they’re worth talking to.

After those 24 hours pass with nothing, it’s your move. That’s where the system starts.

Results from real campaigns

Here are the results of campaigns I ran with a client, using prosp.ai.

Reply rates ranged from 20.8% to 45%, depending on how ‘warm’ the leads are.

Every good LinkedIn outreach campaign should have a rate of at least 20%.

If you don’t reach that, it’s time to get back to the drawing board. You can do that right here, by adopting the 3-step system.

The 3-Step DM System

If you’ve read the Claude Code Email System, this will feel familiar.

Here’s what an effective LinkedIn sequence looks like:

Read more

Five pieces from this year got read more than anything else I’ve published, by a wide margin.

If you’ve only got 10 minutes to go through my content, spend it here.

They’re the ones that changed the most GTM setups.

1. Start with Claude Code for GTM: Where to Start if you haven’t built anything yet.

Many think building a prospect list means paying for a scraping tool or hiring someone to copy-paste for three hours.

This is the piece that proves neither is necessary.

2. Once that’s running, Clay vs Claude Code settles a question worth asking before you pay for either:

Clay is valued at $5B, and Claude Code can reach the same data directly through MCP.

Read it before your next renewal, not after.

3. The Only GTM Tools You Need in 2026.

Some of those twelve-step tool stacks run $3,000 a month and need a full-time engineer just to keep working.

Complexity is profitable, for the people selling it to you.

If your subscriptions have crept past what you can name off the top of your head, start here.

4. Catch Intent on LinkedIn with Claude

The one to read if you’ve been prospecting strangers while ignoring the network you already have.

Your next 10 leads are probably closer than you think.

5. And Context Engineering last, on purpose.

I have yet to meet a founder who said their AI outbound is compounding.

This is the piece that explains why, and what fixes it.

Your bonus: a real diagnostic (not a survey).

If you’ve got a CLAUDE.md or companies.csv already built, this prompt reads them directly instead of asking you to self-report.

If you don’t have those yet, it works from your answers instead.

Either way, it ends by naming the one piece above to read first, instead of a vague “area” to think about.

Help me figure out which of these five gaps matters most in my
GTM setup right now, and which piece I should read first
because of it.

## STEP 1: CHECK WHAT'S REAL
If a CLAUDE.md or companies.csv exists in this project, read
them and use what's actually there. If neither exists, tell me
so, and ask me the five questions below in one message, short
answers are fine for all five at once.

1. Is Claude Code actually running against my market, or still
   unused?
2. Could I name and justify every tool I'm currently paying
   for, in one sentence each?
3. Am I finding prospects mainly through obvious channels
   (ads, cold lists), or through real signals?
4. Have I actually looked at who's one or two steps away from
   me in my own network, or am I only reaching strangers?
5. Does every outbound session start from zero, or is there a
   file that remembers what worked last time?

## STEP 2: NAME THE ONE GAP
Tell me which single area is weakest, in one sentence, based
on real evidence where you have it and my answers where you
don't.

## STEP 3: SEND ME TO THE RIGHT PIECE
Match the weakest area to exactly one of these and give me the
title to go read next:
- Area 1 → "Claude Code for GTM: Where to Start"
- Area 2 or 3 → "The Only GTM Tools You Need in 2026" or "Catch Intent on LinkedIn with Claude," whichever fits my
  answer better, and say why
- Area 4 → "Catch Intent on LinkedIn with Claude"
- Area 5 → "Context Engineering"

Jasper

PS: If you want full access to all 5 pieces above + every guide, every Skill, every systemv —> get 35% off your first year below. Offer valid until Sunday.

Get 35% off for 1 year

The hottest trend in GTM right now is outsourcing your brain to an AI.

A few months ago, an AI startup (Gooseworks) wanted to run my GTM for me.

YC-backed, an AI coworker with its own computer, email, skills and memory. It would run outbound, manage the CRM, and I’d approve its work from Slack.

A client asked me last month whether he should just pay for one of these instead of learning Claude Code.

“Why would I build it when I can hire it?”

Fair question. Here’s what I told him:

The blind spot in every managed AI

These tools execute fast. What they can’t do is fix what you feed them:

• A vague Ideal Customer Profile gets automated into vague outbound

• Generic signals become generic campaigns, run faster

• Untested messaging gets scaled instead of tested

The strategic thinking still has to come from you. No tool changes that, whichever logo is on the invoice.

Which means the real question was never “AI or no AI.”

It’s which camp you join:

💳 The Renter

• Pays monthly for a managed AI coworker

• Fast start, polished interface, someone else’s roadmap

• Owns none of it: the prompts, the data, the logic and the off switch all belong to the vendor

The day the pricing changes or the startup pivots, his entire GTM walks out the door with it.

That’s not a hypothetical anymore.

Gooseworks still says “runs your GTM” on its homepage.

But three months on, what it’s actually shipping and pushing hard isn’t an autonomous coworker running your outbound.

Now it’s a tool that remixes ad templates for Meta.

🛠️ The Builder

• Creates the same system in Claude Code

• Slower first week, then compounding every week after

• Owns everything: the CLAUDE.md that encodes his customers, every prompt, every learning, at a fraction of the platform fees

Nobody can raise the rent on his own brain.

Even the AI startups know it

Here’s the detail that gives it away:

Gooseworks open-sourced every skill its AI coworker uses, both before the pivot and after.

That includes the exact skill that now makes the ad creative they charge you a subscription for.

Anyone can install it straight into Claude Code, for free.

Think about what that means. They’re charging monthly for something you could build yourself with the same free parts already handed out.

The subscription is the business, not the product.

The value was always the system underneath it, the one he open-sourced.

And when the big ambitious version of the company got hard to keep running, that free, owned system is the part that survived.

Lay the first brick today

The renter’s advantage is speed to start. So remove it.

The builder’s first brick is a CLAUDE.md, the file that teaches Claude Code who you sell to, and it takes one conversation to draft. Here’s the master pronpt:

You are my GTM system architect. I'm building the foundation
of my outbound system in Claude Code instead of renting a
managed AI tool: two files, CLAUDE.md and BATTLECARD.md, that
every future session will read before doing anything else.

Never smooth over a vague answer into something that sounds
specific. If my answer is soft, tell me it's soft and ask me
to sharpen it before moving on. A file built on vague answers
produces vague outreach, which is the exact failure I'm
building this system to avoid.

## STEP 1: INTERVIEW ME, ONE QUESTION AT A TIME
Wait for each answer before asking the next.

For CLAUDE.md:
- What I sell, in one sentence, and roughly what it costs
- My best customer of the last year: who, what was happening
  in their business right before they bought, what it was
  costing them
- The customer I regret taking: who, and the early warning
  sign I ignored
- Two or three job titles I actually sell to, not job titles
  I wish I sold to
- One phrase or tone I never want to sound like

For BATTLECARD.md:
- My single best proof point: a named customer (or
  anonymised, my choice) and a specific number. If I don't
  have one yet, say so rather than inventing one, and note
  it as a gap to fill before this file is trustworthy
- The one thing that makes a prospect say yes, described the
  way a customer would say it, not the way my website does

## STEP 2: DRAFT BOTH FILES
Write CLAUDE.md with these sections: ROLE, COMMUNICATION STYLE
(3 voice rules from my answers, not generic advice), IDEAL
CUSTOMER PROFILE (written as the condition my best customer
was in, not just firmographics), NO-GO LIST (built from my
regret customer).

Write BATTLECARD.md with: THE PROBLEM WE SOLVE, THE PROOF
POINT (exactly what I gave you, flagged clearly if it's
generic rather than a named result), OBJECTION HANDLING (ask
me for my single most common objection and draft one response
to it).

## STEP 3: STRESS-TEST BOTH FILES
Before finalising, tell me which answer I gave you was the
weakest, the one most likely to produce generic output if I
don't fix it now. Push back on that one specifically. Then
tell me what's still missing that a full system would need,
the proof point, the ICP, or the voice rules, so I know this
is a starting foundation, not the finished system.

Keep both files under a page each. Every future session
inherits them, so nothing generic gets in.

One conversation, and you own the file every managed tool keeps behind its login.

Takeaway

The renter and the builder end up with the same capability on day one.

They end up in different businesses by year one: one holds a subscription, the other holds a system.

Which camp are you in: paying for a managed AI coworker, or building the system yourself?

Reply “builder” and I’ll tell you what I’d add to your CLAUDE.md .

Jasper

PS: The prompt gets your first file drafted. The Claude Code GTM Starter is the full system: CLAUDE.md, five saved Skills, and the Rules that catch mistakes before they ship.

Get it here.

The other day someone congratulated me on my ‘fundraising’.

Warm little note attached to their connection request:

“Congrats on the fundraising! Super excited to see the Clay team making some serious waves these days. Would love to connect!”

I don’t work at Clay. I never have.

Some automation matched my profile to Clay and confidently personalised away.

That’s what personalisation at scale looks like from the receiving end.

And it’s why the entire approach is backwards.

The decision before the note

Here’s what happens when your request lands:

The person sees your name, your face, and your headline.

In about 2 seconds, they make one subconscious calculation:

“Do I recognise this person, and if not, could this connection be worth something?”

Two nos, and even the sharpest note disappears into the void.

A yes, and they click ‘Accept’ without reading a word you wrote.

The numbers

Across my client’s last campaigns, requests to well-targeted lists ran between 50% and 74% acceptance.

No clever notes or fake familiarity. Targeted lists, a headline that says something concrete, and mostly blank requests.

Hold onto that reply rate column on the right. We’ll come back to it at the end.

Those acceptance rates start with the list.

The targeting came from the plays in the last three editions:

• Live intent signals

• LinkedIn’s native hiring signals.

• Your competitor’s newest connections

If your list is full of people already showing interest in a solution like yours, the rest of this system turns them into connections.

Why connection notes backfire

Every note that opens with “Saw you liked...” or “We’re both in...” screams the same subtext: I’m about to sell you something.

Everyone has seen a hundred of these. The template is the tell, and more effort makes it worse, because effort spent proving familiarity signals exactly the opposite.

The senders rarely work out why their acceptance rate sits under 30%, because they keep fixing the copy instead of the system behind it.

The system, 3 steps:

Read more

The compound effect of AI has a dark side nobody talks about.

Build good habits with it and they compound into something real.

Build lazy habits and those compound too, into a dependence you don’t notice, until it’s too late.

I’ve had a version of this conversation with multiple founders this year, and it always splits the same way.

Same Claude Code, same MCP servers, same starting capability.

Six months apart, they’re describing two completely different relationships with the stack.

✅ The founder who compounds

• Reviews every AI-drafted email before it goes out

• Tests messaging manually before automating anything

• When something fails, updates the no-go list so the system never repeats it

• When something works, asks why and writes the pattern into CLAUDE.md

Each campaign lands a little sharper than the last, because the system remembers what he learned instead of him having to.

❌ The founder who gets dependent

• Automates everything from day one

• Hardly reviews the output, checks the dashboard instead of the emails

• Runs the same generic campaign month after month, because nothing feeds learning back into the system

Ask this founder six months in to describe his own ICP without opening a prompt, and watch him struggle.

AI didn’t make him better. It replaced a muscle he stopped using, slowly enough that he never noticed the trade.

Same stack. Opposite outcomes.

The gap isn’t which AI you use, since both founders above are usually running the exact same stack.

It’s where you’re treating it as a training partner or a shortcut.

• The founders who review, refine, and feed their learning back into the system are building something that compounds every week.

• The founders who automate and forget are building a dependency that gets harder to unwind the longer they wait.

Run this before automating anything else

The habit that separates the two doesn’t need to be complicated.

It’s writing down why something worked, every time, so the system inherits the lesson instead of losing it.

Here’s a Claude Code prompt that forces that habit instead of hoping you remember to do it:

You are my GTM system architect. I'm building the foundation
of my outbound system in Claude Code instead of renting a
managed AI tool: two files, CLAUDE.md and BATTLECARD.md, that
every future session will read before doing anything else.

Never smooth over a vague answer into something that sounds
specific. If my answer is soft, tell me it's soft and ask me
to sharpen it before moving on. A file built on vague answers
produces vague outreach, which is the exact failure I'm
building this system to avoid.

## STEP 1: INTERVIEW ME, ONE QUESTION AT A TIME
Wait for each answer before asking the next.

For CLAUDE.md:
- What I sell, in one sentence, and roughly what it costs
- My best customer of the last year: who, what was happening
  in their business right before they bought, what it was
  costing them
- The customer I regret taking: who, and the early warning
  sign I ignored
- Two or three job titles I actually sell to, not job titles
  I wish I sold to
- One phrase or tone I never want to sound like

For BATTLECARD.md:
- My single best proof point: a named customer (or
  anonymised, my choice) and a specific number. If I don't
  have one yet, say so rather than inventing one, and note
  it as a gap to fill before this file is trustworthy
- The one thing that makes a prospect say yes, described the
  way a customer would say it, not the way my website does

## STEP 2: DRAFT BOTH FILES
Write CLAUDE.md with these sections: ROLE, COMMUNICATION STYLE
(3 voice rules from my answers, not generic advice), IDEAL
CUSTOMER PROFILE (written as the condition my best customer
was in, not just firmographics), NO-GO LIST (built from my
regret customer).

Write BATTLECARD.md with: THE PROBLEM WE SOLVE, THE PROOF
POINT (exactly what I gave you, flagged clearly if it's
generic rather than a named result), OBJECTION HANDLING (ask
me for my single most common objection and draft one response
to it).

## STEP 3: STRESS-TEST BOTH FILES
Before finalising, tell me which answer I gave you was the
weakest, the one most likely to produce generic output if I
don't fix it now. Push back on that one specifically. Then
tell me what's still missing that a full system would need,
the proof point, the ICP, or the voice rules, so I know this
is a starting foundation, not the finished system.

Keep both files under a page each. Every future session
inherits them, so nothing generic gets in.

This is the kind of habit that separates the two founders above.

One does this every week without thinking about it.

The other has never run anything like it, which is why nothing he learns ever makes it into the next campaign.

Takeaway

The stack was never the variable. It’s identical in both cases above.

What decides the outcome is whether someone keeps doing the thinking, or hands it over the day the automation starts working well enough to ignore.

Run the prompt above after your next batch, and you’ll already know which side of this you’re on.

Jasper

PS: The prompt encodes one batch of lessons. The Context Engineering framework is the full system: the CLAUDE.md structure, the four components of an AI-ready GTM setup, and the method for making every session compound instead of repeat.

Get it here.

I was on a call with a CMO this week who runs a private group of growth leaders.

He sits with them every week and hears what they do.

He told me the market has not just slowed, but stopped moving.

The businesses that need solutions have quit buying. They’ve stopped bringing in outside help, and they still haven’t shrunk the teams they know are too big.

“They’ll run the old playbook,” until the board fires them.”

Why founders freeze

A new tool launches every day, and someone explains why it changes everything.

Another post insists that skipping AI outbound already means falling behind.

The feed fills with options before a single decision gets made.

What causes the freeze? Noise.

So you wait. Waiting feels responsible: research, “getting the setup right,” a quieter quarter to come.

None of it looks like a mistake, because standing still leaves no evidence behind.

A failed campaign leaves a mark, and so does wasted spend.

Doing nothing leaves nothing to explain on Monday, which is exactly why it feels safe.

A close rate that hid the problem

A founder I worked with last year had a 70% close rate. Every deal came through referrals and warm intros.

She hadn’t sent a cold email in five years, because why would she? The number said everything was working.

The number was also the trap. A 70% close rate on a shrinking pool of warm leads isn’t a strength but a ceiling with good lighting.

A 70% close rate sounds like skill. Hers was something else: she'd never tested herself against a stranger who didn't already trust her.

The unfreeze system

Four simple steps, each with one job.

1. Name one customer: Not your Ideal Customer Profile. The single real account you’d want ten more of.

2. Name the pattern: What was happening at that company right before they bought?

3. Find one match: One company showing that same trigger right now, with dated public evidence.

4. Send one message. One sentence, naming the evidence, sent today.

Here’s the system as a prompt, so it runs itself instead of living in your head:

You are my GTM strategist. I run a bootstrapped B2B company and I've stalled. Too many tools, too many tactics, too much noise. I keep researching and never start.

Remove every choice but one. Do not give me a system, a
stack, or a list of options. Give me exactly one company to
contact and one message to send. Nothing else.

## STEP 1: LEARN FROM ONE CUSTOMER
Ask me to name the single best customer I've ever had, and
tell you:
- what they do and roughly how big they are
- what was happening in their business when they bought
- the problem I solved, and what it was costing them
Wait for my answer before continuing.

## STEP 2: NAME THE PATTERN
From that, name the single situation that makes a company
likely to need me right now. Write it as one line:
"A company that has just [event], which usually means
[the problem I solve]."
Ask me to confirm or correct it. Do not search until I do.

## STEP 3: FIND THE ONE COMPANY
Search for companies showing that exact situation in the last
90 days, with dated public evidence: a hire, a launch, a
raise, a post. Weigh candidates internally, but show me only
the single strongest match.

Before showing it, check yourself:
- Is the evidence dated and specific, not inferred?
- Would you stake your own reputation on this match being real?
If nothing clears that bar, say so plainly and tell me what
would need to be true for a company to qualify, rather than
handing me a weak guess dressed up as a strong one.

## STEP 4: WRITE THE ONE MESSAGE
For the company you found, write the opening line I'd send:
one sentence, naming the evidence and what it likely means
for them. No pitch, no meeting request, no link.

Show me, in order:
- the company and the dated evidence, so I can verify it
  myself before sending anything
- the one message

Stop there. Tell me to send it before offering anything else.

Ten minutes gets you one company and one line, verified, ready to send.

What this fixes, and what it doesn’t

This fixes today’s problem: getting the first message out before the noise talks you out of it.

It doesn’t fix next week.

It won’t tell you which signal to build the next list around, or stop that list from going stale, or turn one reply into a habit that runs every week without you remembering to trigger it.

I built that part as a weekly routine, for when this week’s start needs to become a system.

That wall shows up for almost everyone within a week of unfreezing.

The first message works, and then you stall again because nothing is running underneath it.

Takeaway

The noise got loud enough that standing still started to feel like the safe choice.

But standing still is the most expensive move on the board, paid every month spent waiting to feel ready.

Three things you can do today:

• Name the one customer you’d want ten more of.

• Run the prompt above and get one company, one line, verified.

• Send the message

Jasper

PS: This article gets you unstuck once. Premium keeps you unstuck:

• Full access to the Claude GTM Systems Library
• 30+ Claude Code Skills for list building, lead enrichment, and outreach that books meetings with your ideal clients.
• Every prompt, skill, and framework ready to apply and run. New skills added regularly. You keep everything permanently.

Get Premium

Every day, your future customers tell LinkedIn they’re trying to solve a problem. They just don’t tell you.

• Someone in your market clicks “attend” on an event about the exact problem you solve.

• Someone else comments on a post about it.

• A third adds a “hiring” badge that says they’ve got a gap and a budget.

Each one is a person looking for a solution to their problem, in public, where you could see it. Most people walk straight past.

In case you’ve walked past them you’ve sent the messages. Built the list, right titles, the people you’d actually want as customers.

But almost nothing comes back. So you assume the copy’s wrong and rewrite it.

The problem was never your message, but reaching the right people at a random moment.

So even a perfect message goes nowhere, because the person reading it isn’t thinking about the problem you solve.

Why timing beats volume

Many chase the loud signals. A company raises money, a company posts a job, and everyone piles in at once.

The quiet ones are better, and barely anyone watches them. They’re the small things people do on LinkedIn every day without thinking twice.

Less obvious, less crowded, and just as much a sign that someone’s in the market.

Too many names, too little time

Collecting these signals is easy. The list just comes back huge and mostly useless.

Pull 500 event attendees and maybe 40 are real buyers. Grab everyone who commented on a post and most are fans and job seekers.

Then you’d have to read all 500 by hand. That’s hours nobody sustains, which is why these signals stay theoretical for almost everyone.

That’s the part Claude takes off your hands.

Two steps: grab the list, then qualify

Two jobs, two helpers.

1. Grab the names. A LinkedIn automation tool collects everyone who showed a signal, in bulk. You can use HeyReach for this. It’s quick at gathering, which is all you need it for. Check my previous walkthrough of how to do this exactly.

You don’t have to pay for one, though. Claude in Chrome reads the same lists straight off the page, pulling the names into a clean list for you.

2. Sort and rank. Claude reads that messy list, drops everyone who isn’t really your buyer, and ranks the rest by how interested they look, telling you why each made the cut.

Where to start

Four signals, and they’re not equal. Work them in this order:

Read more

A few years ago I almost ran my business into the ground, on a good week.

I was drowning in client work and got complacent. Booked solid, delivering, busy in the best way. And because the pipeline was full, I did the laziest thing a founder can do.

I stopped feeding it.

No outbound, no prospecting. Why would I, with more work than I could handle?

For months, nothing happened. Then the market shifted.

A few projects wrapped, a couple didn’t renew, and I looked up to find the top of my funnel gone empty.

I’d been so sure of the work that I’d let the one thing that brings in more work go quiet.

So I rebuilt outbound from zero. Building pipeline when you’re calm is work. Building it when you can feel your cash running out is a different kind of pressure.

I got out of it. But I never forgot how avoidable it was.

You might be making the same mistake

Right now, I see a whole market of teams making this exact mistake. Trusting a channel because it feels safe, while it slowly stops working.

Inbound is changing underneath everyone, and AI is the reason.

People used to search, scroll a page of links, and click yours. Now they ask an AI, get one answer, and never see the list your SEO fought for years to climb.

Your ranking holds. The clicks it used to earn quietly go elsewhere.

And the founders best at the old game are the most exposed. They trust the channel that built them, and keep waiting for it to recover.

The dip you can always explain away

I watched this land on another founder that sells to retailers, names you’d recognise. A business built entirely on his network and inbound, deals that turn into five-year clients.

He’d message me every few months and agree it was time to start outbound. But every time, nothing moved. The channel still felt warm enough to wait on, and starting felt big.

So he waited. Another quarter, another dip he could explain away.

The founders who get out of this have one thing in common

I take on a handful of founding teams at a time, by choice, and the ones I say yes to share one thing: They’ve stopped waiting.

I can’t make that call for anyone, and I’ve stopped trying. Nobody changes someone who isn’t ready.

The most I can do is say it plainly: every month you wait, someone who started is reaching the clients you assumed were yours.

The way out is the thing I stopped doing

The way out? Reach the right companies directly, before they go looking for what you sell. The ones showing in public, right now, that they have the exact problem you fix.

Most teams never start, because the internet handed them ten tools and a hundred tactics and they froze.

So here’s the first step, small enough that you can’t talk yourself out of it. Point this at your best customer, and it builds a starting list of companies in the same situation right now.

You are my outbound starter. I've leaned on inbound and my
network for too long, and it's drying up. I need to start
reaching the right companies directly. Don't hand me a system
to master. Hand me one sharp list I can act on this week.

## STEP 1: LEARN FROM MY BEST CUSTOMER
Ask me to name the single customer I most wish I had ten more
of, and tell you in plain words:
- What they do and roughly how big they are
- What was happening in their business when they bought
- The specific problem I solved, and what it was costing them
- How they found me, or how I found them
Wait for my answers before doing anything else.

## STEP 2: FIND THE REAL TRIGGER
Work out the underlying situation that made them a buyer. Not
the industry. The condition: a recent change, a pressure, a
moment. Write it back to me in this exact shape:
"A company that has just [specific event], which usually means
[the problem it creates that I solve]."
Ask me to confirm or fix it. Do not search until I confirm.

## STEP 3: BUILD THE LIST, SIGNAL-FIRST
Search public sources for ten companies showing that trigger
right now. Rules:
- The trigger must be visible from something real and dated in
  the last 90 days: a hire, a launch, a funding round, an
  expansion, a public post, a job ad.
- The company must plausibly fit the size and type of my best
  customer.
- No guesses. If you can't point to dated evidence, leave it out.
- If you can only verify seven, give me seven and say why.

## STEP 4: NAME THE BUYER, NOT THE OBVIOUS NAME
For each company, the person to reach owns the problem the
trigger points to, not always the most senior name. Small
company: the founder. Bigger one: the function head one level
above the pain, not the CEO who'll never reply. Give me the
exact role at each.

## STEP 5: HAND ME THE STARTING BATCH
Return a table, strongest first:
- company
- domain
- the dated evidence they share my customer's trigger
- the person and role to reach
- one line on the angle I'd open with, naming the evidence
Then tell me the single company to email first, and why it's
the best use of my first hour. Keep it to ten.

Ten companies, one afternoon. That’s all the first step is.

I had my good week years ago. It’s the most expensive thing I never noticed.

Don’t wait for yours to end before you feed the funnel again.

Best,

Jasper

PS: That prompt gets you one list. The full system, every Claude Code Skill, the Systems Library, and every guide, is what turns one list into pipeline you build every week, without waiting for inbound to save you.

Become Premium to Build

I’ve been sitting on this one for a few years.

Every time I thought about publishing it, I hesitated. It’s the kind of play that stops working the moment too many people know about it.

But it’s the most direct way to source leads I’ve come across, and it belongs with the readers who’ll actually use it more than it belongs in a drawer. So here it is.

One caution before we start. The less your competitors know this is possible, the longer it keeps working for you. So keep it to yourself.

The first time I ran this, I’d set the filters, pulled a list, and started reaching out. It worked faster than anything I’d run before.

Then a competitor messaged me on LinkedIn:

“hey man! are you trying to poach my clients?”

Whoops. I hadn’t touched his CRM or broken into anything. The filters were sharp enough that I’d pulled his leads, and some of his freshest actual clients with them.

The highest-intent signal in B2B

Think about what has to be true for someone to connect with your competitor, this week.

Often, they’re not just browsing. People don’t usually connect with a vendor in your category for nothing.

Many are researching the space, sizing up a provider, or already mid-conversation about buying the kind of thing you sell.

The connection is the visible residue of an active buying process.

A fresh connection to your direct competitor isn’t an inference. It’s often a person who has taken a deliberate step toward buying what you sell.

There’s almost no warmer public signal, because there’s no closer proxy for intent than someone walking toward your competitor’s front door.

No list was ‘stolen’ really. The buyer showed their hand, and you read it before anyone else did.

What you need

• LinkedIn Sales Navigator. Required. The filters that make this work don’t exist on a free or Premium account.

• A way to export the leads. Either Claude in Chrome, which reads the list and pulls it for you, or a lead-enrichment tool with a Chrome extension like Apollo, which is faster and can attach emails as it goes. Pick one.

• Claude (Chat or Code). For the cleanup and enrichment pass after export. Code only if you want the weekly routine on top. For a one-off, chat is fine.

The system

Step 1: Build the search and save it.

In Sales Navigator, run a lead search with these filters:

Read more

Every LinkedIn guide tells you the same thing: Post consistently, comment daily. Build an audience. Feed the algorithm and wait for the inbound to come. That’s a full-time job, and likes don’t pay the burn rate.

There’s a different way to pull pipeline off LinkedIn, and it needs zero posting or ads.

No content calendar, engagement farming, or showing up every morning to perform in public. You read the signals already sitting in the platform, and you act on them directly.

The signals are public. They’re dated. And almost nobody acts on them properly, because everyone’s too busy chasing reach.

Here’s how to turn them into a scored list of leads, using three tools:

• Claude Code

• Claude in Chrome

• LinkedIn Premium Account, ideally Sales Navigator (free trial via the link)

A job post is a signal

When a company hires a role related to your solution, they’ve just told you they have the problem you fix. For instance:

• A company hiring a Head of Sales has a sales-scaling problem, right now.

• A company hiring three customer success managers has an onboarding or retention problem, right now.

• A company posting for a RevOps lead has a pipeline-visibility problem, right now.

The hire is the company saying, out loud: “This is the thing we’ve decided to spend money on solving.”

That’s a better trigger than any firmographic filter. “Companies with 50 to 200 employees in fintech” tells you who might fit.

“Companies that posted a Head of Sales role last week” tells you who’s in pain today.

One is a guess about fit. The other is a dated signal of intent.

The setup

Here’s the constraint that shapes everything, and the thing teams often get wrong on their first try.

Claude Code cannot read LinkedIn.

LinkedIn sits behind your login. It’s session-gated, which means the page only exists inside your authenticated browser. Claude Code works off-page, on files and the open web, so it never sees what’s on your LinkedIn screen. Point it at a LinkedIn URL and it gets a login wall, not your jobs feed.

So this runs two tools doing two different jobs.

1. Claude in Chrome reads what’s on the screen. It lives in your browser, inside your logged-in session, so it sees exactly what you see: the jobs feed, the search results, the profiles. It’s the eyes.

2. Claude Code does the thinking off-page. It takes the raw data Chrome pulls and qualifies it, scores it, and works out who to actually contact. It’s the brain.

Code thinks, Chrome reads, and the four steps below move between them. Each step ends with a clear output the next one picks up.

You need both tools on a paid Claude plan. Claude Code runs on your machine. If you’ve followed my earlier guides you already have it set up; if not, the Claude GTM Starter walks you through it.

Claude in Chrome is the part that reads LinkedIn. It’s a browser extension. Open Chrome, find Claude in the Chrome Web Store, click Add to Chrome, sign in with your account, and pin it.

A Claude icon lands in your toolbar, and clicking it opens a side panel that stays open while you browse. That panel is where the Chrome prompts will run.

Once one of the Chrome prompts below works well for you, save it as a shortcut. Type “/” in the panel, pick it, and it runs again instantly, no pasting.

The two tools are built to work together by design, Code in the terminal, Chrome in the browser, so you’re not bolting together a workaround. You’re using them exactly as they’re meant to pair. That’s the whole kit.

And Claude Cowork? It could stand in for Claude Code, and for the off-page thinking it works just as well. We focus on Code because it’s on every paid plan, it remembers your setup week to week, and the steps are identical either way.

Step 1: Code builds the search string

Off-page text work, so this is Claude Code.

Before you touch LinkedIn, you need a precise search that pulls only the roles that signal your buyer. LinkedIn’s job search runs on boolean logic (AND, OR, NOT, quotes, parentheses), and getting it right is the difference between 25 clean signals and 300 rows of junk.

The biggest noise source is staffing agencies, the firms that post these exact roles constantly because recruiting is their business, and this prompt strips them out from the start.

Run this in a Claude Code session:

Read more

A bootstrapped fintech founder I met last week has closed deals most companies never see in a decade.

No outside money and no sales team. He sells through a few senior people with thirty years of relationships each, himself included.

On paper, sales is handled. Then he said the thing that gave it away:

“We have either no business outreach, or a very senior business outreach. Nothing scalable in the middle.”

A brilliant top end, an empty bottom, and nothing in between.

When a senior relationship closes, it’s enormous. But when it doesn’t, there’s nothing underneath to catch the quarter.

No steady flow of smaller, reachable companies keeping the pipeline warm between the big wins.

His product is built. It works. And the entire middle of his market has never heard of it, because reaching them was never anyone’s job.

A strong network feels like a sales engine, right up until the month it goes quiet. Then you find out it was never a system.

It was a few brilliant people and their address books, and address books run out.

This can’t be fixed with another senior hire. You need to build the repeatable layer underneath the relationships, so the pipeline stops living or dying on whether this quarter’s introduction came through.

So here’s a tool that builds the first piece of it:

Give Claude one deal you won through your network, a warm intro, a relationship, anything that didn’t come from cold outreach.

It works out what really made that company a buyer, then finds ten more in the same situation that you can reach directly, no introduction needed.

That’s the missing middle, found in a prompt:

You are my lookalike account finder.

I will describe one customer I won through a personal
relationship or a warm introduction, not through cold
outreach. Your job is to find ten more companies in the
same situation that I can reach directly, without an
introduction.

## STEP 1: UNDERSTAND THE WIN
Ask me to describe the one customer:
- What they do and roughly how big they are
- What was going on in their business when they bought
- The specific problem my product solved for them
- Why they were a good fit, in one sentence

Then ask me two things about who I can serve:
- The size and geography of company I sell to
- Any companies or types I should exclude (competitors,
  industries that never buy, anyone already on my list)

Wait for my answers before going further.

## STEP 2: FIND THE PATTERN
Work out the underlying situation that made this company
a buyer. Not the industry label. The condition they were
in: a stage, a recent change, a specific pressure that
created the need for what I sell.

Write it back to me as a one-sentence trigger, in this
shape: "A company that has just [specific event], which
usually means [the problem it creates]."

Ask me to confirm or correct it. Do not search until I
confirm. This trigger is the thing every company on the
list must share. Industry and size are secondary.

## STEP 3: FIND TEN LOOKALIKES
Search public sources for ten companies showing that
exact trigger right now, not companies that are merely
similar to my customer.

Rules:
- The trigger must be visible from something real and
  recent (last 90 days): a hire, a launch, a funding
  round, an expansion, a public post, a job ad.
- The company must fit the size and geography I gave you.
- Skip anything on my exclude list, my original customer,
  and obvious competitors of mine.
- Do not include a company unless you can point to the
  dated evidence. No guesses, no "likely", no inference.

If a company fits the trigger but not my ICP, leave it
out. A perfect trigger at the wrong-size company is a
wasted row.

## STEP 4: OUTPUT
Return a table, sorted strongest trigger first:
- company
- domain
- the dated evidence they share the trigger (with the
  source)
- the buyer role I should reach
- opening angle: one sentence I could send, naming the
  specific evidence and what it likely means for them.
  Not a generic line. It must reference the actual thing
  you found.

Below the table, tell me:
- Which one I should reach first, and why
- Any company you were unsure about, and what would
  confirm it

Real companies only. If you can only verify seven, give
me seven and tell me why, rather than padding to ten.

A great network hides an empty pipeline the same way a great close rate hides a thin one.

Both feel like winning until the source dries up.

Jasper

PS: Run that prompt on your best network win and reply with what comes back. I’ll tell you which of the ten I’d email first, and why.

This is the free edition of Founder’s GTM 🦊. Premium gets you every Claude Code skill, the full Systems Library, and every guide. The working parts that actually build pipeline.

Go Premium

Last Thursday, I built an outreach system live at the Half Baked webinar. No slides, just Claude Code, a connection to Apollo, and a real client to demo-prospect for.

The client was Half Baked themselves. I asked the system to build a targeted account list, find the right people inside those companies with their details, and write an email to each one. All in one session.

Two moments stuck with me. When the list came back, their founder (Darragh Lucey) spotted two companies on it that were already their clients.

The system had never spoken to him (it was a demo), it just read their ICP and found companies that fit.

Here’s the recording of the full build:

The question Darragh asked me live is worth answering before you watch: “Why Claude Code, and not just normal Claude chat?”

Because Claude Code doesn’t just tell you things. It runs the operation. It connects to Apollo through an MCP server, takes on larger multi-step processes, and fixes its own errors mid-run.

For a twelve-minute demo I built a list of ten leads. If you want to build a list of a hundred, or a thousand, you need Claude Code, because that takes real brain power.

It’s an agentic way of working, not just a conversation with a few basic actions. That’s the difference between a chatbot that helps you think and an agent that runs a go-to-market operation.

If you read the my previous article, you already have the prompt for step one. This is the rest of the system it belonged to.

The whole thing runs in three steps:

1. Build the account list. Claude Code searches the web for companies showing a real signal, then falls back to Apollo to fill any gaps.

2. Enrich the list. Through the Apollo MCP connection, it finds the right buyer inside each company and pulls their verified email.

3. Write the outreach. It writes one personalised email per contact using my YYY framework, the same structure I’ve tested across thousands of cold emails.

I ran all of that live with three prompts, on top of one CLAUDE.md file holding the context. Below is the full walkthrough: the three exact prompts, how each one ran, and the output.

The full build, with all three prompts

Read more

Everyone’s racing to plug AI into their GTM. New tool, new integration, new automation, every week.

But almost none of it fixes the thing that actually breaks it.

I've spent over a decade watching outbound campaigns fail, and the cause is almost always the same. The people on the list were rarely in a position where the email mattered to them when it landed.

No tool fixes that, because the tool runs after the targeting decision’s already made. Point a faster engine at the wrong accounts and you reach the wrong people quicker.

The tool’s the last 20% of the job. The first 80% is deciding who belongs on the list and why now.

Get that right and a plain text email beats a polished sequence sent to a list built from a job title and a headcount filter.

That’s what I’m showing this Thursday, live.

The session

I’m co-presenting a session with Andy McCotter-Bicknell from Apollo, hosted by the Half Baked newsletter.

To save your spot, use the link below 👇🏼

One thing to be clear about before you sign up: This is the lean build. No Clay tables, fifty-step automations, or GTM-engineer setup that takes a weekend to wire together.

Andy and I are both building from inside Claude, with nothing you need a technical background to follow.

If you’ve ever read an outbound thread and thought half of it was written in a language you don’t speak, this session’s the opposite of that.

Andy takes the product path. He builds a list, creates a sequence, and hits go, all from inside Claude using Apollo’s new integration. He’s debuting their build-sequences-from-Claude feature, and it’s impressive to watch.

I take the foundations: the targeting and the messaging. In 15 minutes, I build a signal-driven account list, find the right person at each company, then write the copy for every contact using the YYY messaging framework.

Together it’s the full picture. Andy shows you how easy the tools have become. I show you how to think about who you point them at, so it produces replies instead of faster rejection.

What you’ll walk away with

The three-step workflow I run, start to finish:

1. Claude builds an account list from signals.

2. Apollo finds the right person to contact at each account.

3. Claude writes a YYY email for every contact.

Step one’s the part that decides everything. A perfectly enriched contact and a beautifully written email are worthless if the company wasn't living the problem you're describing when it landed.

One thing worth saying: this prompt runs best on top of a properly set-up Claude project. If you haven’t built one yet, the Account List Builder walks you through the whole setup, the project folder, the CLAUDE.md, the signal scoring, in about an hour.

The prompt slots straight into that system. You can run it without the full setup, but you’ll get more out of it with the foundation in place.

Power Prompt: The Signal-Driven Account List Build

On Thursday I’ll show you how to build the list when you already know the problem to look for.

This prompt does the harder thing. It works out the problem for you, by reading your own best customers first, then finds the companies living that exact problem right now.

It’s the half most teamsget wrong, and it’s the half that decides whether any of this works.

Two stages in one prompt. Find the pattern, then find the companies.

Run it inside a Claude Code session. Before you start, drop two or three of your closed-won deal notes, call summaries, or even a short paragraph on each of your best customers into a folder called /customers/ inside your project.

If you don’t have written notes, the prompt will ask you questions instead. It works either way.

Here’s how it runs:

Read more

The founder of a US B2B fintech told me this back in March. Eight people, healthy ARR, decent retention:

“Time is our enemy. We know what we should be doing. We just never get to do it properly.”

They had already done what most teams are still trying to do:

• Claude Code installed

• A clean ICP document

• A Clay subscription running for over a year

• Apollo credits sitting unused at the end of every month

This founder is ahead of 90% of bootstrapped founders. The stack is in place, and the posts have been read.

There is just no time to use any of it properly.

The bottleneck has often nothing to do with knowing about sales tools. It is simply finding 90 minutes a week to actually use them.

And 90 minutes a week, every week, is harder to defend than a single Saturday afternoon of setup.

So I created a Claude prompt to fix the time problem instead of the knowledge problem. It runs every Monday, and takes about 10 minutes.

Last Monday it produced this:

THIS WEEK’S FIVE

Row 1: A Series C in DACH. Hired a new VP Finance announced on LinkedIn 17 May. The funding round three months ago combined with this hire confirms the breaking point window is open. Strongest move today.

Row 2: A French spend management platform. Posted three Controller roles in the last five days, two in Paris and one in London. Multi-entity scaling pattern confirmed. Reach out before the roles get filled.

Row 3: ...

Across the last few weeks of these Monday lists, he sent 60emails. Five turned into replies, two into calls.

That’s the output. Below is how to build it: the prompt, the CLAUDE.md rules that keep it consistent every week, and the trick that makes it sharper the longer you run it.

The Monday Refresh Prompt

Below is the prompt that you can run every Monday. Paste it into your Claude session inside your gtm-list-builder project folder.

It reads your existing companies.csv, refreshes everything rated HOT or WARM in the last 30 days, and produces the Monday hit list you saw above.

You are my Monday morning list refresh GTM agent.

Your job is to find the five companies on my list
who are the strongest move this week, and tell me
in plain language why.

## STEP 1: LOAD CONTEXT
Read CLAUDE.md in full before doing anything.
Open companies.csv from the project folder.
Identify every row rated HOT or WARM with a
date_added inside the last 30 days.
These are the rows you refresh this week.
Skip everything older than 30 days.

## STEP 2: SEARCH IN THIS ORDER
For each company, run these searches in this exact
sequence. Stop the moment you find a trigger inside
the last seven days, mark the company HOT THIS WEEK,
and move to the next company. Do not keep searching
once a fresh trigger is confirmed.

  1. LinkedIn company page — new senior hires posted
     in the buying centre in the last 7 days
  2. LinkedIn job posts — new postings matching the
     BUYING TRIGGERS in CLAUDE.md, posted in the last
     7 days
  3. Recent news search — funding, expansion,
     acquisition, product launch, regulatory change,
     anything in the last 7 days
  4. Website — product pages, pricing changes,
     case studies posted in the last 7 days

If none of the four return a trigger inside the last
seven days, check whether the original trigger from
date_added is still active. If yes, mark REFRESHED.
If the original trigger is now more than 60 days old
and nothing fresh exists, mark STALE.

## STEP 3: UPDATE THE CSV
Only update these three columns:
  buying_signal       (HOT, WARM, COLD)
  buying_trigger      (the freshest trigger you found,
                       with the date it appeared)
  date_added          (today's date if a new trigger
                       was found, otherwise leave it)

Save after every 10 rows. Do not overwrite other columns.
Do not add new rows.

## STEP 4: PRODUCE THE MONDAY REPORT
Output exactly this format. No preamble. No closing.

THIS WEEK'S FIVE
Row 1: [Company name] — [trigger event, dated] —
       [one sentence on why this is the strongest
        move you can make today]
Row 2: [...]
Row 3: [...]
Row 4: [...]
Row 5: [...]

If fewer than five companies qualify as HOT THIS WEEK,
do not fill the gap with weaker rows. Say:
"Only [N] companies are hot this week. Want me to add
new accounts to the list, or focus on warm follow-up?"

PATTERN OF THE WEEK
One sentence on any cluster you noticed across the
fresh triggers. Example: "Three companies on
the list hired new CFOs in the past five days."
If no cluster is visible, write: "No pattern this week."

DROPPED FROM HOT
List any company that moved from HOT to WARM or STALE
this week and why, one line each.

## STEP 5: FEED LEARNINGS BACK
At the very end, append a single line to CLAUDE.md
under a section called ## REFRESH LEARNINGS.
Format: [date] — [observation worth remembering].
Only add a line if you noticed something new about
the buying pattern, not just the week's activity.
Example: "20 May — Companies in DACH region showing
stronger trigger density than UK over last four weeks."
If nothing new emerged, do not add a line.

The CLAUDE.md addition

For the prompt to behave consistently every week, add this block to the end of your CLAUDE.md. Without these rules, Claude rates differently from session to session.

## WEEKLY REFRESH RULES
- Save changes to companies.csv after every 10 rows
- Mark HOT THIS WEEK only if the trigger is from
  the last 7 days
- Drop a company from HOT to WARM if the original
  trigger is more than 60 days old and nothing
  fresh appears
- For the priority five, prioritise multi-trigger
  companies over single-trigger companies
- If fewer than five qualify, name the gap honestly
  instead of filling it with weak rows
- Append new pattern observations to ## REFRESH LEARNINGS
  at the end of this file

How to run it

Open your terminal:

cd ~/Desktop/gtm-list-builder
claude

Paste the prompt. The first run takes 20 to 30 minutes because Claude is doing the research from scratch. From the second week onward it is an incremental check, and 10 minutes is realistic for a list of 50 companies.

What you get back every Monday

• A five-company hit list with a documented reason next to each one.

• A pattern observation across the fresh triggers.

• A note on which companies dropped off and why.

• One line appended to your CLAUDE.md so next week’s refresh is sharper than this week’s.

That is what makes the system compound. Three months in, your operating manual knows your buying patterns better than you do.

Claude Routines

Ten minutes is short. But you still have to open your terminal every Monday and remember to paste the prompt.

Routines closes that gap. A Routine is a saved Claude Code job that runs by itself on a schedule. Anthropic runs it on their cloud while you are still asleep.

To turn the Monday Refresh into a Routine, open Claude Code inside your project folder and type:

/schedule

> Run the Monday Refresh Prompt every Monday at 9am
  my local time. Email the report to me when it
  completes.

Claude Code translates the natural language into a schedule, confirms the timing, and saves it.

What changes once it is running:

• You open your inbox on Monday morning

• The Monday report is already there

• You read it over coffee and decide who to message first

Ten minutes drops to close to zero. The Routine refreshes the list and drops the report in your inbox. You still read it, decide who is worth an email, and adjust. But the digging is done before you sit down.

Takeaway

The problem was never knowing about the tools, but finding the time to use them every week.

The Monday Refresh Prompt fixes that. You build your list once, then every Monday the prompt checks it for fresh signals and hands you five companies worth emailing, with a reason next to each one. 90 minutes of research becomes 10.

Three things make it work:

• The prompt searches in a fixed order and stops the moment it finds a fresh trigger, so it stays fast

• The CLAUDE.md rules keep the scoring consistent week to week, so Monday’s list is built the same way every time

• Each run adds one learning back to your CLAUDE.md, so the system gets sharper the longer you use it

Then Routines removes even the ten minutes. Schedule the prompt once and the report lands in your inbox every Monday morning.

Build the list once, and let the prompt keep it warm. Then let Routines run the prompt.

Best,

Jasper

PS: You can now get 30% off the annual plan. Offer valid until June 6th.

Get 30% off forever

The majority of founding teams running their Go-To-Market through Claude are using maybe 10% of what it can do.

It’s not because a lack of effort. They just haven’t used Claude’s Skills properly.

Skills have been available since October 2025 and most teams are still missing them, which is why every Monday morning still feels like rebuilding the system from memory.

Once I created my first one, building a fresh prospect list took ten minutes instead of ninety.

Writing a world-class cold email took two minutes instead of fifteen.

Reviewing six months of closed deals to find what my best clients had in common happened in a single command, instead of an afternoon with a spreadsheet.

The thinking happened once, when I wrote the Skill. Every run after that was simply execution.

Quick Note On What This Covers

Skills work in both Claude and Claude Code in your terminal. The file format is identical, but what each one can do is different.

Skills in Claude run inside the chat: research, drafting, single-company briefs.

Skills in Claude Code run on your PC they read your CSVs, update your CLAUDE.md, call MCP servers like Apollo and Brave Search, and save files back to your project folder.

This article and every GTM Skill in Founder’s GTM focuses on Claude Code. If you have not set it up yet, the Account List Builder walks you through it first.

What a Skill actually is

A Skill is a saved workflow that lives in a folder inside your Claude Code project.
It tells Claude exactly how to do one specific job: which files to read, which steps to follow, what to output, what to skip.

You write it once. Every time you want Claude to run that job again, you type one command and Claude follows the instructions.

The clearest way to understand a Skill is to compare it to the two things you are probably already using if you’ve played with Claude Code:

• A prompt is a one-off. You type it, Claude does the work, the session ends, and the instructions are gone.

• A CLAUDE.md is your project’s permanent operating manual that loads automatically in every session: your ICP, your no-go list, your rules, the background context that applies to everything you do in that folder.

• A Skill sits between them. It loads only when you ask for it, or when Claude decides it is relevant to the task at hand. One specific job, one specific output, available on demand.

Without Skills, every workflow lives in your head or in a notes file you copy and paste at the start of each session. With Skills, every workflow lives in a file Claude reads automatically when you call it.

The Account List Builder uses a Skill called /list-builder. Every system in the Founder’s GTM System Library ships at least one. Each one is a folder, each one is a file, and each one runs identically every time.

The Anatomy Of A Skill

Here’s what a Skill folder actually looks like on your PC:

.claude/skills/quick-brief/
├── SKILL.md              ← the instructions
└── references/
    └── good-briefs.md    ← optional supporting material

Only SKILL.md is required. Everything else is optional. Reference files, examples, templates, anything you want Claude to read when running this specific Skill lives in subfolders and loads only when needed, which keeps your session lean.

The SKILL.md itself has two parts:

• A short YAML frontmatter block at the top (description).

• Plain markdown instructions below.

The frontmatter tells Claude what the Skill is called and what it does:

---
name: quick-brief
description: Produce a one-page outreach brief for a single company. Pulls recent news, hiring signals, and the most likely person to contact.
---

The description matters more than it looks. Claude reads it at the start of every session to decide whether to load this Skill automatically when relevant.

A weak description means the Skill never gets used. A sharp description means Claude finds it for you the moment a related task comes up.

Everything after the frontmatter is your instructions in plain English: the steps Claude follows, the inputs it needs, the outputs it produces, the edge cases it should handle.

Why This Matters For GTM

Three reasons Skills change how founders run outbound: repeatability, compounding and speed.

The first is repeatability. Every prospect list scores the same way, every email follows the same logic. The output last Monday and the output this Monday are produced by the same instructions, which means your system stops being something you “do well when you have time” and becomes something that just happens.

The second is compounding. Each Skill you write makes the next one cheaper. Reference files get reused across multiple Skills. The signals you document for the list builder feed into the hook writer. Three months in, you have a library that works for your business specifically, not a generic GTM stack lifted from someone else’s playbook.

The third is speed. A workflow that took ninety minutes the first time runs in ten once the Skill exists. The system survives a busy week, which is the only kind of week founders have.

Build Your First GTM Skill In 15 Minutes

This is a working Skill you can paste into your project folder today, called Quick Brief.

Read more

A few months ago I cancelled my Clay subscription. When I wrote about it, the pushback came fast.

“Sure, fine. But what about the email enrichment waterfall?”

“What about running thousands of rows?”

“What about the table view?”

Fair points, and here’s what they have in common: None of them are about enrichment.

All of them are about the same thing: workflo…

Read more

Two agentic coding tools launched within two months of each other last year.

Both run locally in your terminal. Both support MCP servers and maintain persistent project context. The difference:

OpenAI’s Codex CLI is open-source under Apache-2.0.

Claude Code is not. At the foundation layer, they are otherwise functionally similar.

But here is where it gets interesting for B2B founders: Your GTM motion defines which one you pick.

The short version: Codex builds your product. Claude Code sells it.

Let’s dive in.

Tribalism

, founder of Trigify, recently moved toward Codex and away from Claude Code, calling the community loyalty around the two tools irrational.

Codex CLI has over 700 releases on GitHub, around 74,000 stars, and a release cadence most open-source projects would be embarrassed to match.

For context, Claude Code sits at over 112,000 GitHub stars, and Gemini CLI has entered the race too.

All three are moving fast. In my read of the market, a comparable Claude Code release produces significantly more noise outside engineering circles.

But Max is running a heavy engineering operation where parallel agent workflows and complex codebases matter. His context is not yours if you are a B2B founder trying to build pipeline.

The tribalism exists because the two tools serve different work, and the communities that formed around each one reflect that split.

Claude Code went viral in the 2025 winter holidays when non-developers discovered it, including many non-programmers who used it for vibe coding.

By January 2026 the GTM community was fully inside it. Founders, operators, and GTM teams adopted it to build outbound systems, research ICPs, and draft sequences without writing code.

Codex stayed with engineers. Different work for different people, and different patterns.

Claude Code for GTM. Codex for the plumbing.

Eric Nowoslawski shared a session that made the split clearer than anything I have seen.

He used Codex’s /goal command to take a 500,000-row CSV, upload it to Supabase, create a Clay table via webhook, map LinkedIn profiles to personal email enrichment, configure a multi-provider waterfall from cheapest to most expensive source, and update Supabase with the results. All in one session.

His reaction: “I couldn’t believe it was working so well.”

That’s Codex doing what Codex does: complex, multi-step, production-grade infrastructure that would have taken a developer a full day to wire by hand.

This isn’t a threat to Claude Code, but proof of exactly why the two tools belong in different parts of your stack.

Where Claude Code wins for GTM:

• Writing cold emails that sound like a person who understands their buyer’s situation

• Analysing call transcripts and catching the moment a prospect went quiet because their champion left

• Defining a sharp ICP by pattern-matching across customer interviews and win/loss notes

• Prioritising accounts based on judgment rather than just a score

• Skills ecosystem for outbound

• MCP patterns for lead enrichment tools

• The list-building and enrichment workflows solving the same GTM problems

Where Codex wins:

• CRM integrations

• Lead scoring models

• Webhook setup

• Data pipelines, automation infrastructure, and design-to-code workflows from screenshots or wireframes.

Claude Code wins where the work requires more judgment, language, and an understanding of human context.

Codex wins where the work requires more precision, structure, and production-grade reliability. Shipping the infrastructure that will eventually run the motion Claude Code is designed to execute.

The founder who stopped guessing

A startup running an HR tech platform had twelve closed deals in the previous year, all from referrals and one conference. No outbound or system, and no way to repeat it.

I asked them what those twelve customers had in common. They all had HR teams and were mid-size companies. But beyond that they couldn’t tell why they bought and the others didn’t.

That’s the most expensive kind of not knowing.

1. We spent the first session loading closed-won notes into Claude Code. Call transcripts, CRM fields, email threads. Every one of their twelve customers had posted a People Operations or HR Systems role in the sixty days before first contact. Not a Head of HR. Specifically an operational or systems-focused hire, which meant they were trying to build process.

2. The second session built a scored list of thirty-eight companies showing the same signal right now.

3. The third session wrote the first three-email sequence using his exact proof points and a specific hook for each account based on what the job posting said.

They didn’t need Codex for any of that. Just a system that could read their own data, find the pattern, and turn it into outreach.

Takeaway

If you’re a B2B founder doing outbound, building prospect lists, writing sequences, and managing a lean GTM motion without a technical co-founder, stick with Claude Code.

The ecosystem was built for you. The Skills, the CLAUDE.md patterns, the MCP integrations for enrichment and list building, all of it compounds the longer you use it.

If you’re a GTM engineer or technical founder shipping production infrastructure alongside your sales motion, running complex data pipelines, or managing parallel agent workflows across a real codebase, Codex earns its place.

Pick the one that matches your motion, and don’t switch because of hype.

The best GTM system is the one you actually understand well enough to improve every week.

Sincerely,

Jasper

PS: Every week you run outbound without a signal-based list is a week you are reaching the wrong companies.

Join the founders who built their first signal-based list in a single Claude Code session.

—> The Claude GTM Starter Guide shows you exactly how: the setup, the CLAUDE.md, and the first session from scratch. One hour. Your first scored list by end of week.

Four founders running outreach on LinkedIn sent me their numbers. Different offers, different markets, but the same pattern.

The only thing they had in common was this: before they used a system, their DMs were getting ignored. After the system, conversations started happening.

This edition breaks down the friction that kills conversations and shows the 3…

Read more