How I Built a $10K/mo Business with 6 AI Employees

Last month, my business made $10,372. I didn't send a single invoice, write a single line of customer support code, or manually fulfill a single order. The entire operation is run by six AI agents and a series of interconnected automations I built. This isn't a VC-funded startup; it's a one-person, AI business income machine built on tools like n8n, Supabase, and the Claude API. Here’s how the money actually flows in.

The 4 revenue engines (and how much each makes)

The $10K isn't from one magical product. It's the sum of four distinct, automated systems that run concurrently. Critically, each engine leverages the same core infrastructure and AI "employees," keeping my active management time to under 5 hours a week. The breakdown is precise: FlowStore marketplace brings in $2,500/month, digital products on autopilot generate $3,200/month, content-driven revenue adds $1,800/month, and a fourth engine—micro-automation services—contributes the remaining $2,872/month. This diversification is key; if one stream dips, the others compensate. The foundation is a central Supabase database that tracks customers, licenses, and content across all engines, with n8n as the orchestration layer. My first step was mapping these engines out as separate but connected workflows. Your next step should be the same: audit your skills and assets to see what could be productized into at least two of these models. Don't try to build all four at once.

Engine 1: FlowStore marketplace ($2,500/month)

This is a curated marketplace for pre-built n8n workflows, like "Twitter to Newsletter Automator" or "AI-Powered Lead Qualifier." I sell each workflow for a one-time fee of $147-$297. The key is that purchase is just the beginning of the AI business income cycle. The automation handles everything: When a purchase is made via Paddle, the webhook triggers an n8n workflow. This workflow: 1) Creates a customer record in Supabase, 2) Generates a unique license key using a simple Node.js function (crypto.randomBytes(12).toString('hex')), 3) Sends a personalized onboarding email via Resend with the download link and license key, and 4) Adds the customer to a specific "workflow buyers" segment in my CRM. The downloadable workflow file itself is hosted on Vercel Blob Storage, with access gated by the license key. This entire process, from payment to delivery, runs without me. For a deeper dive on this model, see my complete guide to selling automation workflows online. Actionable next step: Package one of your most-used automations into a templated format. Use n8n's n8n.nodes to replace your specific API keys with credential inputs and document it for a novice.

Engine 2: Digital products on autopilot ($3,200/month)

This engine sells niche, evergreen digital guides (e.g., "The Supabase & n8n Cookbook") for $47. The "autopilot" part is crucial. The product is created once by an AI employee—I use the Claude API with a detailed prompt and my own markdown templates. The sales page is a static Vercel page. When someone buys, the Paddle webhook triggers a different n8n workflow than the marketplace. This one: 1) Charges the card, 2) Fetches the latest product file from Blob Storage, 3) Uses Resend's dynamic template to send the file attached, and 4) Logs the sale. The real magic is the follow-up sequence. Based on the product_id in Supabase, another n8n workflow (scheduled daily) checks for customers 3 days post-purchase and sends a helpful "implementation tip" email, and at 7 days, asks for a review. This simple sequence boosts engagement and reduces refund requests. Actionable next step: Identify one piece of "tacit knowledge" you have. Use the Claude API to expand it into a 20-page PDF. Set up a simple Paddle product and a two-email sequence in Resend or your ESP.

Engine 3: Content-driven affiliate and ad revenue ($1,800/month)

This engine fuels the others. I run a technical blog where I detail my automation builds. Each article is an organic lead generator. The content is largely researched and drafted by an AI agent (Claude + custom instructions), but I add the crucial technical specifics and code snippets. The monetization is dual: display ads via Carbon Ads ($400/month) and strategic affiliate links ($1,400/month). The affiliate links aren't generic "buy this laptop"; they're for the exact tools I use and recommend in the tutorials, like n8n Cloud plans, Supabase Pro, and specific VPS providers. A key piece of infrastructure here is a small Node.js script that runs weekly, checking the health of all affiliate links in my Supabase content table to avoid 404s. The blog itself is a Next.js site on Vercel, with ISR for fast loading. This content engine directly supports my AI business income by building authority and driving targeted traffic to my FlowStore and digital products. For more on the agent setup, read how I replaced 6 employees with AI automations. Actionable next step: Write one detailed "how-to" post for an automation you've built. Include the n8n JSON export. Place affiliate links only for tools you explicitly use in the tutorial.

Engine 4: AI Employee Kit sales ($2,500/month)

This is the most hands-off engine. I sell pre-configured "AI Employee Kits"—essentially n8n workflow templates with clear instructions—that automate specific business functions. Each kit sells for a one-time fee of $97-$297. I built the sales process itself with n8n. Here’s how it works:

1. A purchase on Paddle triggers my n8n webhook.
2. The workflow creates a customer record in Supabase, generates a unique license key, and sends a Resend email with the download link and access instructions.
3. A separate daily n8n workflow checks the Supabase purchases table for new sales in the last 24 hours and posts a summary to a private Discord channel. That’s my only daily touchpoint.

The kits are just ZIP files containing an n8n.json workflow file and a README.md. My top sellers are the "SEO Content Reviewer" ($297) and the "Weekly Analytics Digest" ($97). I don't do custom support; the documentation has to be bulletproof. I use a simple Vercel-hosted page with a Stripe-like checkout powered by Paddle. Gross margin here is near 100% after payment processing fees.

The tech stack and monthly costs ($120 total)

My rule is: if a cost scales with revenue (like payment processing), it's fine. If it's a fixed cost, it must be minimal. Here’s the breakdown:

• n8n Cloud (Pro Plan): $20/month. Non-negotiable. I self-hosted for months but the $20 is worth the uptime and not having to manage a server.
• Supabase (Pro Plan): $25/month. My single source of truth for customers, licenses, and content metadata. The pgvector extension is key for some AI workflows.
• Vercel Hobby Plan: $0. My marketing site and documentation are static Next.js sites. Never hit the limits.
• Resend: $20/month. For all transactional emails (downloads, receipts). Their React Email library is perfect for crafting good-looking automated emails.
• Paddle: ~$55/month in fees. This is a percentage of revenue, not fixed. They handle VAT/tax globally, which is worth the fee alone.
• Misc. AI Credits: ~$0. I use the Claude API via n8n's native node, but costs are sub-$1/month as most workflows are lightweight.

Total fixed costs: ~$65. The rest is variable. I avoid subscriptions for "maybe" tools. Every tool must justify its existence in a live workflow.

Month-by-month growth timeline

This wasn't linear. Here’s the real timeline:

• Months 1-2 ($0 revenue): Building in silence. I created the first 3 n8n workflows for myself: one to generate LinkedIn posts, one to format blog drafts, and one to curate news. Total time invested: ~80 hours.
• Month 3 ($400): Launched the cheapest AI Employee Kit ($47). Sold 8 copies mostly to Twitter followers. Realized people needed more documentation.
• Months 4-5 ($1.5k/mo): Launched FlowStore. Listed my own workflows. The first marketplace sale from a stranger was a huge motivator. Started writing about the "AI employees" concept.
• Month 6 ($3.8k): Digital product engine kicked in. A product about building AI automations sold 31 copies in a week, driven by a detailed case study blog post.
• Months 7-9 ($6k-$8.5k): Compound effect. FlowStore had 50+ listed workflows. Affiliate revenue from tool reviews started. Upped the price of my top-tier kit to $297.
• Month 10+ ($10k+): Stability. The system runs. I now spend about 10 hours a week on maintenance, content, and building one new thing.

The key was that revenue from one engine fed content for another. A sale from a kit would lead to a case study blog post, which drove affiliate clicks and marketplace visibility.

3 mistakes I made and what I'd do differently

1. Building custom admin panels too early. I spent two weeks building a fancy React dashboard to manage customer licenses. It was buggy and I dreaded maintaining it. The fix: I moved everything to Airtable initially, then later to Supabase with simple, secure direct table access. I manage everything through filtered views now. Never build internal tools if you can use a spreadsheet or a database GUI.
2. Under-pricing the first kits. I sold the first 50 "SEO Reviewer" kits for $47. The support burden was identical to the $297 price point, and the customers were more demanding. The fix: I doubled the price to $97, then later to $297, adding better documentation and example outputs with each jump. Higher price = better customers.
3. Not isolating workflow APIs. In one early n8n workflow, I called the OpenAI API, the GitHub API, and the Resend API all in one massive canvas. When it broke, debugging was a nightmare. The fix: I now build small, single-purpose workflows. A "Generate Blog Outline" workflow. A "Commit to GitHub" workflow. A "Send Product Email" workflow. Then I chain them together via webhooks or a central orchestrator workflow. This makes failures obvious and fixes trivial.

Wrapping Up

This business works because it's built on systems, not my time. Each revenue engine is automated and feeds the others, creating a compounding loop. The tech stack is boring, reliable, and cheap. The biggest lesson was to start selling the process behind the outcome, not just the outcome itself.

Want to build the same system? Start with our Digital Products, explore AI Employee Kits, or sell workflows on FlowStore.

You don't need a revolutionary idea—just a system that runs without you.