// workflows · payoff post · issue 01
You commented WORKFLOWS.
Here's the stack.
5 workflows. 1 operator. 0 hires. The shape, not the prompts.
The reason most creators don't make money with AI isn't tools.
It's tabs.
47 starred repos. 12 ChatGPT prompts saved in Notion. 6 free Claude templates from Twitter. Zero revenue.
Stop collecting. Start running workflows.
$ ls ~/repos | wc -l → 47 // $ stripe balance → $0.00
Below: the 5 workflows that replace a 5-person freelance stack. ~$5K of monthly invoices, gone. Replaced by one operator and a Claude account.
// how to read this post
You get
- The shape of each workflow
- The trigger and output
- Where to start today
- The mistake that kills it
- The build order
You don't get
- The full prompts
- The CLAUDE.md files
- The Claude skills
- The webhook configs
- The deploy scripts
That second list lives in the Starter Kit. Different reason for each, but it boils down to: the wiring is what makes them run. The wiring takes work to publish properly. Without it, you're back to copying prompts off Twitter.
The 5 workflows.
01 // LEAD CAPTURE
Comment-to-DM Automation
The shape
Someone comments a keyword on a reel or carousel. ManyChat catches it. Claude personalizes the DM based on the post they engaged with. The link logs the lead and triggers a Beehiiv sequence. You wake up to a populated list and zero manual outreach.
Trigger
Keyword comment on any post you run this on.
Output
A personalized DM with a single link, sent within 30 seconds. A new row in your CRM. The lead enrolled in the right email sequence based on which post they commented on.
Start here
Set up one ManyChat keyword trigger pointing to a free Beehiiv signup form. That's the spine. Get one keyword working before you scale to ten.
Don't do this
Don't ask the user multiple questions inside the DM. Conversion drops sharply with every extra question. One CTA. One link. One click.
What's not here
The Claude prompt that adapts the DM to the specific post they commented on. Generic DMs convert in the single digits. Personalized ones land 3-4x higher. The personalization is the lift. The prompt is in the Kit.
02 // CONTENT ENGINE
Carousel + Reel Generator
The shape
One topic in. Claude runs your demand-creation framework, drafts 8 slides, writes the caption, renders the PNGs at IG dimensions, and optionally generates a 5-second animated cover. You approve. You post.
Trigger
A topic plus your hook framework dropped into a Claude Project.
Output
Eight 1080x1350 PNGs. A caption. A cover animation. End-to-end in roughly 12 minutes vs the 4-hour designer round trip.
Start here
Write one SKILL.md for "carousel building" that contains your fonts, colors, spacing, slide framework, and brand voice. Claude loads it once. Every brief applies it.
Don't do this
Don't have Claude write the hook without a framework. Generic hooks die in the feed. Feed it your viral hook patterns and tell it to score the topic before drafting anything.
What's not here
The scoring matrix wired into the skill so every topic gets graded out of 25 before you spend a single token on slides. Anything under 18 gets killed at the gate. That's the demand filter. Kit.
03 // CLIENT OPS
Intake-to-Invoice Onboarding
The shape
Form submission triggers Claude. It writes the proposal, drafts the contract from your template, generates a Stripe invoice link, sets up the kickoff Notion doc, and ships one clean email to the client. You get a paid project. No follow-up admin.
Trigger
Tally form webhook on a new client inquiry.
Output
One email to the client containing the proposal, the payment link, and the kickoff doc. Total elapsed time: under 4 minutes. Manual version takes most VAs 90 minutes.
Start here
Map your current onboarding as a sequence of named artifacts (proposal, contract, kickoff doc). Claude can only generate what you've named. The naming IS the system.
Don't do this
Don't let Claude write contract clauses from scratch. Have it fill in your reviewed template. Hallucinated terms have killed deals. Templates only.
What's not here
The Tally to Claude webhook config and the Stripe invoice generator that writes line items dynamically based on the scope captured in the form. Kit.
04 // NEWSLETTER LOOP
Compound Audience Engine
The shape
Every post you publish feeds the next one. Claude pulls themes from your last 4 issues, picks the angle with the highest open and click rate, drafts the next issue in your voice, and proposes 3 subject lines for testing.
Trigger
A weekly scheduled task in your Claude Project or a manual "draft next" instruction.
Output
A 600-1,200 word draft in your voice with 3 subject line options and an editorial note about why this angle was chosen. Edit time drops to under 20 minutes.
Start here
Build one Claude Project per workflow (not one shared). Drop in your last 10 newsletters. Add a 3-paragraph voice profile describing how you sound. Now it writes like you, not GPT.
Don't do this
Don't let it generate topics from scratch. Topics that compound are picked from engagement data, not invented from thin air. Feed it your open rate history and let it surface patterns.
What's not here
The topic-mining query that pulls high-CTR themes from your Beehiiv archive plus the subject line tester skill. The compounding part is the mining. Kit.
05 // OFFER PAGES
Landing Page Builder
The shape
One brief in a markdown file. Claude Code scaffolds the page, writes the copy section by section, builds the components against your design tokens, renders the preview screenshot, and deploys to a staging URL. You iterate in chat, not in Figma.
Trigger
A brief.md file with: Promise. Mechanism. Proof. Stack. Price.
Output
A deployed landing page on a staging URL, a preview PNG for review, conversion tracking pre-wired. First version live in roughly 25 minutes.
Start here
Write your next offer brief as a single markdown file. Sections in this exact order: Promise. Mechanism. Proof. Stack. Price. That structure is what Claude Code scaffolds from.
Don't do this
Don't let Claude pick the layout from scratch. Pick a known-good template (long-form sales page or modular block layout). Constrain creativity. You want conversion, not a portfolio piece.
What's not here
The Claude Code starter repo with design tokens, component library, deploy script, and conversion tracking already wired. Skip the boilerplate. Kit.
Build order matters.
Don't try to build all 5 at once. You'll abandon all 5.
Pick the one that unblocks revenue fastest for where you are right now.
// pick the bottleneck
Audience but no leads
→ build 01
Leads but inconsistent posting
→ build 02
Clients but operations drain you
→ build 03
List but flat open rates
→ build 04
Offer but no page
→ build 05
One workflow per month. By month 5 you don't have employees. You have a system that runs the business.
// pre-flight checklist
Before you build any of these
- A Claude account (Pro at minimum, Max if you'll run all 5)
- One Claude Project per workflow, not one shared project
- A markdown file naming exactly what each workflow should produce
- An honest list of what a real human currently does in that role
- A spare 90 minutes to wire the first one end-to-end
Skip these and the workflows return slop. The setup IS the workflow.
Why this isn't taught.
Most AI content teaches prompts. Prompts are tactics. They expire.
Systems compound. The 5 workflows above are essentially the same in 2024 as they are in 2026. The prompts inside them change every quarter. The wiring doesn't.
If a creator is selling you "the 50 best ChatGPT prompts of 2026," they're selling tactics.
Buy systems instead.
Buy systems instead.
That's the difference between this newsletter and most of your saved AI bookmarks.
$ claude run workflow.stack --all
Loading 5 workflows...
Operator: 1
Hires required: 0
Monthly cost replaced: $5,000
✓ system ready
Want the actual builds?
CLAUDE.md files. Claude skills. ManyChat flows. Prompt libraries. Webhook configs. Deploy scripts. Every gap above, filled.
Get the Starter Kit →Or wait. The next 5 carousels each break down one workflow in public. Workflow 01 drops in 14 days. You'll have the system either way. Question is whether you want the wiring now or in 5 months.
— Mike
Mike Means Business · @mikemeansbusiness_ai · Issue 01 · 2026