Quick framing before I get into anything tactical: most operators are using Claude wrong.
They treat it like ChatGPT. They open a chat, ask a question, get an answer, close the chat. Maybe they save the prompt for later. That's it. That's their entire AI workflow.
That's not what Claude is. Claude is an operating layer. It's a department. It's the thing that connects every other tool you use, automates every grunt task you'd otherwise do yourself, and lets you run multiple brand-level workflows in parallel that would have required a team of 5-10 people two years ago.
This module is how I actually use it. Not the surface-level "ask Claude to write you a tweet" version. The full operating-layer version that lets me run multiple brands, ship case studies in days instead of months, and create more content + ads + SOPs in a week than most operators ship in a quarter.
The setup most operators are missing
First thing — stop using Claude in the browser. Download the desktop app. The desktop app is faster, has better connector integrations, supports more parallel chats, and feels different to use. The browser version is a fallback for when you're on someone else's computer.
Only exception: Claude in Chrome. This is a separate Chrome extension that lets Claude actually control your browser — navigate sites, fill out forms, do the manual click-work you'd otherwise do. To set this up: in your Claude desktop app, go to Customize → Connectors → Claude in Chrome. Enable it. It installs as a Chrome extension. (I personally use Dia browser, but it works the same way in any Chromium-based browser.)
When the extension is installed, you can ask Claude to do anything in your browser. Need to pull data from a SaaS dashboard that doesn't have an API? Ask Claude. Need to run through 20 product pages and pull pricing into a spreadsheet? Ask Claude. The extension prompts you for approval each step (or you can set "always allow" once you trust the workflow).
Most of mine are on always-allow. If you're cautious, leave it on "needs approval" while you build trust. Either way, make sure your Claude account has 2FA enabled. This is the access layer to your entire business — protect it.
Connectors are the real unlock
Once you have the desktop app + Chrome extension running, the leverage point becomes connectors. Connectors plug Claude into the tools your business actually runs on.
What I run connected:
- Google Drive — Docs, Sheets, Slides, PDFs. This is the one you absolutely need on day one. When I'm building a brief or SOP, Claude pulls source docs from Drive directly. When I want to spit something into a Google Doc, Claude does it and gives me the link. The friction loss vs copy-pasting back and forth is enormous.
- Notion — this is how I shipped this entire case study so fast. I write in Claude, Claude pushes the content into Notion as proper pages with formatting, sub-pages, page mentions, the works. The same workflow runs my brief dashboard, my SOP library, and the media buyer pipeline (covered in Module 2's Media Buyer sub-page).
- Google Calendar — scheduling, conflict detection, meeting prep
- Fathom — every recorded call I have on Fathom is now searchable + summarizable through Claude. I can ask Claude "what did the consulting client I had Tuesday say about their hooks performance?" and it pulls the answer. Massive for anyone doing calls.
- Gmail — I don't run this myself yet, but the play is to integrate it with my customer support VA so she can have Claude triage tickets, draft replies, and surface escalations. That's coming.
- Calendly — for any consulting, scheduling, lead capture
- Custom connectors — you can write your own if you have the engineering chops or hire someone who does. I haven't gone deep here yet but it's where the next level of leverage lives.
There are a bunch more — Uber, Uber Eats, even Cash App. Most of these aren't relevant to running a DTC brand, but the principle is: anything you do manually that has an app, has a connector somewhere. Find the ones that match your bottlenecks.
The principle: remove constraints, don't add workflows
Quick philosophical layer because this is where most operators fuck up.
Don't use AI just because you can. That's the equivalent of buying a shiny new tool because it's shiny. You end up with five overlapping workflows, none of which work, and your day got slower instead of faster.
The right framing: AI exists to remove constraints. Find the bottleneck. Find the manual step. Find the hire you couldn't afford. Find the workflow that's eating 2 hours of your day. Then ask Claude how to remove it.
If a workflow is already working fine, leave it alone. Adding AI to a working system for the sake of "using AI" makes it worse. Subtracting work is the play. Adding work is not.
Same way Module 2 covered the water-hose analogy — you're looking for kinks to unkink, not new pipes to add. Claude is one of the highest-leverage tools you have for unkinking. Use it that way.
Skills — where the real compounding lives
Connectors plug Claude into your stack. Skills are how Claude executes repeatable workflows on top of that stack.
A prompt is a one-time instruction. A skill is a reusable program. This is the distinction most operators never make, and it's the difference between using Claude like a chatbot vs using it like an operating system.
Where you build them: in the Claude app, go to Customize → Skills → upload. Skills come as .zip files. Drop the zip in, Claude unpacks it, the skill is now active. If you're confused on the setup, literally just ask Claude "how do I upload this skill?" and it'll walk you through it.
Skills I run on the daily:
- Image generation skill — produces high-volume static ad images via the Claude API at pennies per day. This is the one you'll get free in email 2 of the lead magnet (along with the setup walkthrough).
- Seedance — text-to-video and image-to-video for AI UGC ads. Storyboard in, MP4 clips out, with last-frame chaining for seamless transitions.
- Sora 2 Pro — premium video generation, locked to the highest tier. Same storyboard input, different model.
- DTC advertorial — the Resilia-style template. Pre-written copy in, deployable HTML out. GemPages-ready, all theme overrides solved.
- DTC ad batch generator — full ad production pipeline. Brief in, 3 hook variations + matched images + Drive uploads + tracking sheet row, all in one autonomous run.
- First-person women copy — high-emotional-valence copy in a woman's voice for diary-format advertorials, founder stories, testimonial-style pages.
- Skill creator — the meta one. Teaches Claude how to build new skills for me on demand. When I notice a workflow that's repeatable, I tell the skill creator to package it, and now I have a new skill in my library.
The compounding effect: every skill I build is permanent leverage. Once it exists, the workflow it covers is solved forever. Image generation isn't something I think about anymore. Advertorial deployment isn't something I think about anymore. Both are pre-solved by skills, freeing my brain for the part of the work that actually requires my attention.
I'm giving away most of my skills (with my API keys removed) inside the masterclass + via the email sequence you signed up for. The skill files themselves are the compressed form of dozens of hours of operating experience — the kind of thing that would normally cost $1k+ on a course platform and isn't worth gatekeeping at the case study level.
Don't be precious about your prompts and skills
Quick aside that's saved me a lot of mental energy.
Most operators treat their prompts and skills like sacred IP. They hoard them. They password-protect them. They worry about losing them.
Don't. If you lose a skill, a prompt, a workflow doc, an SOP — Claude rebuilds it in 10 minutes. Open a new chat, explain what the workflow needs to do, paste any reference materials you have left, and ask Claude to rebuild it. Often the rebuild is better than the original because you've learned what worked and what didn't in the meantime.
The leverage isn't the prompt itself. The leverage is knowing what to ask Claude to build for you. That meta-skill — being able to articulate a workflow well enough that Claude can build a skill for it — is what compounds. The actual file contents are replaceable.
This frees you from the operator's curse of trying to safeguard everything you've ever built. Build it, use it, share it where it makes sense, rebuild it when needed. Move fast.
The killer move: parallel chats across multiple projects
Here's where Claude becomes a genuine force multiplier instead of a faster Google.
Most operators run one chat at a time. They open a chat, work the problem, close the chat, open a new one. Single-threaded.
I run multiple chats in parallel. Right now, while I'm working on this case study with one chat, I usually have:
- Another chat researching the avatar for one of my brands
- Another chat drafting native ad copy for a different brand
- Another chat generating images for a campaign
- Another chat reviewing a cold-email reply from a prospect
- Another chat debugging an SOP a hire just asked about
All running in parallel. All making progress on different problems at the same time. I jump between them.
This isn't multitasking — multitasking means one brain switching between tasks. This is parallel processing. Each Claude chat is its own brain working its own problem with its own context. I'm just the supervisor checking in on each one.
What this does to throughput: you stop being a single-threaded operator. You become 4-5 brains operating concurrently. The work that would have taken a week done sequentially gets done in a day.
This is the thing most operators don't realize is possible. It's also the single biggest reason I can run multiple brands while shipping a 50-page case study in days.
A few rules to make parallel chats work without breaking your brain:
- One project per chat. Don't blend "research for brand A" with "copy for brand B" in the same chat. Context bleeds across, outputs get worse.
- Name your chats clearly. I rename every chat in the sidebar to its actual purpose ("CRWN avatar research" not "untitled chat"). Without this, you lose track within a day.
- Use Projects for ongoing work. Claude Projects let you carry context + files + custom instructions across multiple chats on the same long-term project. This case study lives inside its own project.
- Don't try to run more than ~5 in parallel. Beyond 5, the supervision overhead exceeds the throughput gain. Find your number.
Every operator I know who uses Claude at the level I use it runs parallel chats. Every operator who's stuck on Claude as "a faster Google" doesn't. Make the switch and you'll feel the leverage immediately.
Bottom line — stop being a VA
The hard line on all of this:
Stop using Claude like an assistant. Use it like a department.
Multiple workflows in parallel. Skills handling the grunt work. Connectors integrating every tool you run. You're the one whose brain matters — every other layer of the org chart is automatable now, and Claude is how you automate it.
Operators who don't make this shift are doing $200/hr work at $20/hr speed. Operators who do make it are running multi-brand portfolios from a laptop with a team that fits in one Slack workspace.
Pick which version you want to be.
What's in this module
This module breaks down into 4 sub-pages, each covering a specific layer of how I actually run Claude as an operating layer.
- Catalyst — the single most important psychological framework I use for hooks. Julius's Catalyst Framework: external events (regulatory, discovery, exposure) as the trigger, blame-shift architecture to avoid forcing negative self-admittance, perspective-shift multiplier for ad multiplication. Why this beats desire-based hooks every time.
- Videos — Seedance + Sora pipelines for AI UGC video. Storyboard in, MP4 clips out, with last-frame chaining for seamless transitions and character consistency across scenes. The storyboard format that works.
- Native — full walkthrough of setting up the imagen skill + writing a native ad end-to-end. The Native Ad Copywriter prompt structure. Eugene Schwartz frameworks inside prompts. Hook variations via perspective-shift. Voice authenticity.
- Advertorials — the DTC advertorial skill (Resilia-style template). Pre-written copy in, deployable HTML out. GemPages-ready, all theme overrides solved. Anti-AI voice principles.
Dropping next inside the masterclass: SOP Building + The Notion Integration — the full Claude → Notion → Google Apps Script pipeline that runs my brief dashboard, image editor queue, and media buyer Master Tracker (cross-references Module 2 Media Buyer). Voice memo → Whisper → Claude → V1 SOP. The stress-test loop. Drift management. The whole system, walked through start to finish.
This drops into the masterclass on the next biweekly update cycle. The masterclass updates biweekly with whatever I'm actively running, iterating, or learning. The case study you're reading is the foundation — the masterclass is the live, evolving extension that stays current as my brands and systems evolve.
Want the full, video version? Every SOP, notes, extra information, plus calls & a chat with me inside our 1,000+ member community — here inside the RunItUp Masterclass → Join here
Join now to lock in your rate, prices are increasing quickly.