Guide · August 12, 2026

Start here: the one-person AI business stack

Heads up: this post contains affiliate links. If you buy through them, we may earn a commission at no extra cost to you, it's how the site stays free. We only recommend tools we've actually run ourselves. Full disclosure.

Every workflow on this site follows one architecture. Learn it once and every tool decision you ever make gets easier, because you will stop asking “is this tool good” and start asking “which layer does this belong to, and is that layer my current bottleneck.” Those are different questions, and the second one is the one that makes money.

This guide is the whole map: what each layer does, what it costs at every stage, the order to buy in, the math behind the order, and the mistakes we either made ourselves or watched other operators make. It is long because it is the page everything else on this site hangs from.

The architecture: brain, machine, spokes

Picture your business as three layers.

The brain is Claude. Writing, planning, editing, strategy, code, analysis. Every workflow starts here because every business problem starts as a thinking problem. When we built the email machine, Claude designed the sequence arcs before a single email existed. When we built this site, Claude wrote the code. The brain is not one tool among many; it is the layer every other layer depends on. That is also why it comes first in the buying order, and why it has no serious substitute at its price.

The machine is your delivery system. Email, automation, funnels, checkout. The machine is what turns the brain’s output into revenue while you are asleep, at the gym, or building the next thing. Ours is GoHighLevel, running a 12,000-subscriber list that sends over 1,000 emails a day at a 39 percent open rate with a few hours of weekly maintenance. Before GHL it was simpler tools, and before that it was nothing, which meant every sale required us to be awake. The machine layer is the difference between having an audience and having a business.

The spokes are specialists. Voice (ElevenLabs), video (Pictory, HeyGen), SEO (Surfer), design (Canva), app building (Lovable). Each spoke does one thing the brain cannot: produce the finished asset in its medium. The rule that keeps spokes from bankrupting you: add a spoke only when a live workflow demands it, never because a launch video was impressive. Every spoke is a subscription, and subscriptions compound in the wrong direction.

Why this order and not another

People buy AI tools in the order they discover them, which is roughly the order of loudest marketing. That order is backwards, and here is the reasoning to fix it.

The brain comes first because it multiplies everything after it. A $20 Claude subscription makes your emails better, your scripts better, your product pages better, and your decisions better. No spoke can do that; ElevenLabs makes exactly one thing better. Highest leverage first.

The machine comes second because audiences decay without it. If you are producing content with the brain but have no capture and no nurture, you are renting attention and returning it every night. The machine converts attention into an owned asset, the list, and the list is the only asset in this whole architecture that compounds on its own.

Spokes come last because they are amplifiers, not engines. A voice tool amplifies a script that the brain wrote for an audience the machine captured. Spokes bought before the first two layers exist just produce prettier content that goes nowhere.

The four stages, with real math

Stage 1: Proving it. $20/mo. Claude Pro plus Systeme.io’s free tier, which genuinely covers 2,000 contacts. Your jobs at this stage: pick the niche, build a free hook worth an email address, publish weekly, and get the first 100 subscribers. Nothing else. If you cannot get 100 subscribers with the free stack, a paid stack would not have fixed it; the hook or the niche is wrong, and this stage exists to find that out cheaply.

Stage 2: First revenue. $30 to $50/mo. Add your first spoke, chosen by your next revenue project and nothing else. Launching an audio product: ElevenLabs at $22. Faceless channel: Pictory at $19. SEO site: hold off on Surfer until Stage 3, and use the free approximation in the writing hub. Your revenue goal here is modest and specific: one product or offer earning more than the stack costs. That flips the psychology permanently, because from then on the tools pay for themselves and upgrades are investments instead of expenses.

Stage 3: The machine upgrade. ~$117/mo. When your list clears roughly 1,000 engaged subscribers, or when you catch yourself wanting to treat two audience segments differently, move to GoHighLevel at $97 and rebuild the nurture with tag-based branching. This is the single biggest jump in the whole path, in both cost and capability. Budget a real weekend for the migration, or steal our architecture from the email machine post, which documents the exact build.

Stage 4: Compounding. $150 to $350/mo. More spokes as revenue justifies them, GHL’s $297 tier when limits bite, maybe Surfer and Semrush when SEO becomes a serious channel. By this stage the stack should be a single-digit percentage of the revenue it produces. If it is not, you skipped a stage somewhere, and the fix is almost always to go back and strengthen the machine layer, not to buy another spoke.

The five mistakes that burn the first $500

Buying spokes before the machine. The most common one. A creator with ElevenLabs, Pictory, Midjourney, and no email capture is spending $60/mo producing content that builds someone else’s platform. The audience lives on TikTok’s servers, not theirs.

Subscribing at discovery instead of at need. Every tool demo is impressive; that is what demos are for. The rule that saves roughly $200/mo: subscribe the week a live workflow needs the tool, not the week you discover it. Your bookmarks folder is free.

Tool-hopping the brain layer. Claude to ChatGPT to Gemini and back, chasing whichever model won this week’s benchmark. The switching cost is real: your prompts, workflows, and instincts are calibrated to one model’s behavior. Pick one, go deep, revisit twice a year at most. The differences between frontier models are smaller than the difference between a practiced operator and a tourist.

Automating before validating. Wiring Zapier chains and content calendars for a product nobody has bought yet. Automation multiplies whatever exists, including zero. Sell one first, manually, painfully. Then automate what worked.

Counting tools instead of workflows. The question is never “how many AI tools do I use.” Ours is: how many complete workflows run end to end without falling apart. Three tools in one working pipeline beat ten tools in none. Every stack on the stacks page is a complete pipeline for exactly this reason.

What this looks like when it works

A concrete week from the system this site documents: Monday, Claude drafts the week’s content from an outline and firsthand notes. Tuesday, the ElevenLabs pass turns the best piece into audio. Wednesday, screenshots and the human edit. Thursday, publish; the machine’s capture hooks are on every page. Friday, the newsletter goes to the list, drafted by Claude from the same notes in twenty minutes. Sunday, the sales campaign, also Claude-drafted, sells to the segments the machine tagged automatically. Total hands-on time: eight to ten hours. Everything else is the stack running.

That is the honest pitch for the whole architecture. Not “AI runs your business,” because it does not, and anyone selling that sentence is selling something. The real pitch: one person, working normal hours, operating at a scale that used to take a team, because each layer does its one job and the layers compound.

Start today, in order

  1. Claude, free tier, today. Prove to yourself you will use the brain daily before spending anything.
  2. Systeme.io free tier the same week. Capture from day one, even before you have anything to send, because the list you start today is the one compounding next year.
  3. Pick your build on the stacks page and follow its recipe. Each stack is a complete pipeline with the roles and costs laid out.
  4. Add the one spoke your build needs. One.
  5. Graduate to GoHighLevel when the list tells you to, around 1,000 engaged subscribers.

Questions the stacks page and hubs do not answer: ask the Stack Advisor, which knows every tool on this site and will hand you a stack instead of a lecture.