How Shane Hummus Uses Claude to Do 90% of His Content Work

Shane Hummus lets Claude do about 90% of his content work. Not 100%.

I like that honesty. Most "I automated everything with AI" setups quietly skip the part they still do by hand. Shane's doesn't.

He has 1.73 million subscribers on YouTube, says he's made well over $10 million on the platform, and has used AI tools for content since 2022, before ChatGPT even came out.

So his system is worth a close look.

Here's the whole thing. Every step, every tool, every setting. Then what I changed when I ran the same model on my podcast.

#The big idea: one video feeds everything

Before any tool, Shane fixes the strategy.

You make one long video a week. That's it. That's the hard part, and you only do it once.

Then that one video becomes everything else:

• Shorts
• Instagram posts
• LinkedIn posts
• X, Facebook, all of it

He calls YouTube the "mother platform".

And here's the part most people get backwards: you can't run it the other way. You can't turn a LinkedIn post into a video. There's nothing to cut.

→ Long video in. Everything else out.

Shane tells the story of a former CMO from a big ad agency. Small channel, about 2 hours a week. One video a week, then she cut it into LinkedIn posts and clips. In her words, YouTube was the "hero piece". Shane says that weekly video is what got her first 37 clients.

Not the subscriber count. The discipline of one thing a week feeding everything else.

It's also how I run my own podcast. More on that below.

#Step 0: Give Claude eyes

Shane puts it nicely: "Claude out of the box is a genie still in the lamp."

It's powerful. But it can't see anything. It doesn't know what's working on YouTube right now.

So he plugs two research tools into Claude. Claude calls these plug-ins "connectors": once one is connected, Claude can use that app for you.

vidIQ. In Claude, click Customize, then Connectors. Search "vidIQ". Click Connect. A tab opens and asks you to approve your account. Approve it.

Then go back to Connectors and search vidIQ again, to make sure it's really connected. You'll see a list of permissions. Shane's advice: leave everything on default. Don't touch it.

TubeLab. Same thing. Connectors, search TubeLab, Connect, approve, check it's there. Default permissions.

(He'd also use 1of10, but it can't plug into Claude yet. He says he's asked them many times.)

Why both? They look like they do the same thing. They don't.

• TubeLab tells you about the channel. Is it making money? Roughly how much? Face on camera or faceless? How big? How often does it post? It's basically a list of businesses you can filter.
• vidIQ tells you about the video. How many people watched it. How it did against the channel's normal numbers. How fast it's climbing right now.

Channel data plus video data. Those are Claude's eyes.

#Step 1: Switch from Chat to Cowork

In the box where you type, next to the plus button, you'll see two modes: Chat and Cowork.

Chat is fine. But Shane uses Cowork for all of this.

Cowork is the one that actually does the work. It runs multi-step jobs, holds onto your files, and gives you better output for this kind of task.

And here's the part that makes it a real automation instead of a faster way to type: Shane says Cowork runs remotely. You can set something to run every morning, close your laptop, and it still runs.

One limit he flags: if a task needs a folder on your actual computer, that one still needs your machine awake. Everything in this workflow runs through connectors, so a closed laptop is fine.

#Step 2: Install the skills

A skill is a saved set of instructions that teaches Claude one job, so you don't have to explain it every time. Shane runs four:

1. A niche validator
2. A video idea finder
3. The "holy trifecta" (title, thumbnail, intro)
4. A script writer

He gives away "light" versions of them in his video description. The full versions are for his paid community.

To install one: click Add, then Upload skill, then pick the file. That's it.

Once installed, you call any skill by typing / in the chat box.

Each one hands its result to the next.

#Step 3: Your niche, in one sentence

This is the step everyone wants to skip. Don't.

Shane mentions a Reddit post where someone said it took them 20 years to find their niche. And they were still looking.

Why does it take so long? Because the old way is a loop: make something, wait, see if it works, adjust, wait again. Every turn of that loop takes months.

Shane's shortcut is one sentence, in one of three shapes:

• I help X do Y.
• I help X overcome Y.
• I help X do Y by Z.

That's it. Every automated step after this inherits that sentence. And it's the one thing no tool can write for you.

How he runs it: type /, pick the niche validator, and describe yourself. Out loud.

He uses Claude's voice button instead of typing. His example, said messy, like talking to a friend: he was a burned-out pharmacist who started making content on the side, and now he wants to help other burned-out professionals do the same.

Claude interviews him first. How far along is he? How does he want to make money? How many hours a week can he put in?

He answers. Claude hands back a few niche sentences. He picks one and copies it.

That sentence now rides through every step that follows.

Why it matters: content aimed at one specific person with one specific problem is much easier to grow and to sell from. Shane shares a few examples from his community where small channels found clients this way.

The sentence comes first. Everything else follows.

#Step 4: Find ideas that already won (the Icahn method)

This is the part I liked most.

Shane pastes his niche sentence into the chat and runs the video idea finder.

The skill uses what he calls the Icahn method, named after investor Carl Icahn, who made his money buying businesses that were doing well despite being run badly, then fixing them.

Same logic, for YouTube. You look for videos that:

• have 100,000+ views
• come from a channel with 100,000 subscribers or fewer
• get at least 5 views per subscriber
• and are mediocre or, ideally, bad

Why bad? Think about what that combination tells you.

The video didn't win because of the audience. It didn't win because it was well made. It won because of the idea.

That gap is your opening.

His example, in the table above: a doctor in his community who helps people age well found a video for people aged 60 to 80. Shane says she made her own version, a real expert this time, with a very similar title, and it got 978,000 views in a month.

Two rules here:

• Don't copy excellent videos. That's a terrible idea.
• Don't copy big channels. Also terrible.

You're looking for proof that a topic pulls, even from a small channel, even done badly.

His run came back with 12 ideas that passed all four filters, with ratios like 30 to 1, 17 to 1 and 15 to 1. Plus 3 near-misses it flagged anyway. It even wrote the hooks.

Twelve ideas, found rather than brainstormed.

#Step 5: Package it (the holy trifecta)

Outside of the niche and the idea, three things decide if anyone clicks: the title, the thumbnail and the intro.

Shane compares it to a product. Best product in the world, terrible packaging? Nobody tries it.

But here's the rule that matters more than any of the three on its own:

All three have to tell the same story.

In his words: "Congruence actually beats quality every single time."

A 10/10 thumbnail and a 10/10 title pointing in different directions usually lose to a 7/10 thumbnail, a 7/10 title and a 7/10 intro all pointing the same way.

Think about how people actually decide. Nobody reads your title while scrolling. They see the thumbnail first. Then the title. Then they click, and the intro has to deliver what the thumbnail promised.

So if the idea is what makes the video work, the idea has to be visible in the picture.

He runs the trifecta skill on the winning idea. It hands back:

• a thumbnail mockup: where his face goes, what expression, what words, what prop
• 10 title options
• 5 thumbnail text options
• a what-not-to-do list (he loves this part)
• 5 intro options

Notice something: the titles stay about 90% the same as the original winning video. On purpose. Change it too much and you stop conveying the idea that was already proven.

Shane's example here: a hunting and outdoors channel that had been at it for 12 years and couldn't repeat an old hit. They ran the method on their own back catalogue, rebuilt an old idea with a matching thumbnail, title and hook, and he says it got 340,000 views and 11,000 new subscribers. Then they repeated the formula on purpose.

#Step 6: Write the script from your own voice

Now the script writer skill. Type /, pick it, paste the title and angle. It asks what call to action you want at the end. Shane takes the recommended one.

And here's the anti-slop ingredient: the yap session.

Before Claude writes anything, Shane hits the voice button and just talks for 5 to 15 minutes. His stories. His numbers. His mistakes. What he'd tell a friend who's starting today.

Then Claude writes from that.

Same tool, completely different output. The one thing that changes is whether your voice went into it.

What comes back:

1. An outline of everything in the video, in order. Some people record straight from this.
2. A word-for-word script underneath.

In his community it's about 50/50: half read the script, half talk from the outline.

One rule either way: read the intro word for word, and practise until it sounds natural. Read badly, everyone sounds like an airport announcer.

#Step 7: Record it yourself (with your phone)

This is where Claude stops.

If you have a personal brand, you record yourself. Shane is clear: no AI avatar of yourself. He says YouTube is demonetising channels that try it, and people can tell. Once they can tell, they don't trust you, and they don't buy.

(Business owners can have someone on the team record instead. Fine. But for most people, the personal brand is the point.)

Then a story I really enjoyed.

Shane wanted a fancy mobile studio with his apartment view behind him. He says he spent north of $30,000. Hired a lighting pro, hired a camera person. Four hours of setup later, his face was still too dark. More light. A $3,000 light here, a $10,000 light there. So much light he couldn't read his own screen. Then the sun moved. Then the camera overheated.

Next day, someone said: why not just try an iPhone?

It looked better than the $30,000 setup.

→ Gear was never the problem.

Phone or webcam. That's enough.

#Step 8: Edit it yourself (Claude is too slow, for now)

Shane uses Descript. There's even a Descript connector inside Claude: Customize, Connectors, search Descript.

So he tested it.

Claude took 28 minutes to edit a 5-minute video. He did the same edit himself in about 3 minutes. And his was better.

His verdict: it'll get there in a couple of years. Not today. Edit it yourself.

His Descript routine, in order:

1. Studio Sound. Cleans the audio: room echo, background noise, that thin webcam sound.
2. Shorten word gaps. Find all, then "shorten all". Every silence across the video, gone.
3. Remove filler words. All the ums and uhs, removed in one click.
4. Delete sentences in the transcript. Highlight a sentence, hit delete. It's gone from the video. Like editing a Google Doc.
5. Zoom for emphasis. Highlight the words, Change layout, pick zoom.

You're editing the transcript, not the video. That's why it's fast.

#Step 9: Upload once, the smart way

One minute that saves you forever: in YouTube Studio, go to Settings, then Upload defaults. Add your description template, your standard tags, your default visibility.

Every upload from then on arrives pre-filled.

You can also publish straight from Descript to YouTube once it's set up.

#Step 10: Let your winners pick the next idea

You'll never run out of ideas with the Icahn method. But if you feel stuck, Shane does this:

1. Download your YouTube analytics data.
2. Give it to Claude with a prompt that asks for a rating with decimals and a confidence level for each idea.

Why decimals? It forces Claude to actually rank, instead of telling you all ten ideas are great.

Why confidence? Because a 9/10 idea at 40% confidence is a guess. A 9/10 at 90% confidence is a plan.

And notice what he asks about: the winners, not the losers.

Everybody opens analytics to find out what went wrong. Shane opens them to find out what went right, and does more of that. There are a million ways to fail and only a few ways to win.

→ The Icahn method tells you what works out there. Your own analytics tell you what works for you, with your face and your voice. Keep both.

#Step 11: Post it everywhere

The last step. And the one almost everybody skips.

Shane says most of his highest-viewed Instagram posts were just YouTube content, posted again. Same on X, LinkedIn and Facebook.

He keeps the platform-specific tuning for his paid program. Fair enough.

The problem with this step isn't that it's hard. It's that it's tedious.

Five apps, each with its own captions, sizes and best times. So for most people it stops after a few weeks.

#This is the step we built Pozy for

I run the exact same "mother platform" model on my Profitable Founder podcast.

One long episode goes to YouTube. Then Claude Code cuts it into 9:16 clips, and schedules them through Pozy: two a day on Instagram, X and YouTube Shorts. Under each one, a comment keyword that sends the full episode by DM.

Nobody opens five apps. I pick the hooks, and the week lands in the content calendar.

Here's what that did on Instagram:

Two things I do differently from Shane.

1. My edit is code, not clicks.

Shane edits by hand in Descript, and for one video a week he's right: it's faster.

But I make 8 to 10 clips per episode. So Claude Code built my editing for me. It writes down every word with its exact timing, keeps whoever is talking in frame, and adds the title and subtitles. Every clip gets checked before anything posts. The whole thing, step by step: how I turn one episode into two weeks of clips.

At ten clips a week, that pays off. At one video a week, do what Shane does.

2. Reading the numbers doesn't need a spreadsheet.

Shane downloads his analytics and drops them into Claude. With Pozy connected, Claude reads your posts' numbers directly: views, likes and comments per video, plus reach, saves, shares and watch time on Instagram Reels.

One honest limit: for YouTube, impressions and watch time still come from a YouTube Studio export. So for Shane's exact prompt, you still download that one file.

#What I'd steal from Shane, today

If you only take five things from this:

→ One long video a week. Everything else gets cut from it.

→ Find ideas, don't brainstorm them. 100K+ views, small channel, 5x views per subscriber, bad execution.

→ Title, thumbnail and intro tell one story. Congruence beats quality.

→ Talk first, then let Claude write. Your voice notes are the difference between content and slop.

→ Post everywhere, without opening five apps. That last step is where the easy reach is.

#Questions

Does Claude really do 90% of it? For Shane, yes: research, ideas, packaging and scripts. Recording and editing are still him. In his test, Claude needed 28 minutes for an edit he did in about 3.

Do I need a fancy camera? No. Shane spent over $30,000 on a studio setup and an iPhone looked better. Phone or webcam is enough.

Do I need to be on YouTube if my audience is on LinkedIn or Instagram? That's Shane's whole point: a long video can feed every other platform, and nothing else can. You can't cut a LinkedIn post into a video.

Will Claude post without asking me? With Pozy, Claude shows you the plan in the chat first. You can also hold every post for approval. See Approvals.

Does Pozy replace vidIQ or TubeLab? No. They do the idea research. Pozy posts, runs your inbox and reads your own accounts' numbers.