Run a one-person company on AI agents
Over six months the poster ran a company with agents reading and writing the business as text, including turning one LinkedIn post into multiple pieces of content and managing engagement, which backfired with slop copy.
Done with a custom-built or unnamed agent
- What they did
- The poster put the business into a single git repo as plain-text folders and playbooks, then used one generalist agent to route tasks across sales, CRM, content, outreach, and ops. Agents drafted work into one approval queue while browser and CLI tools handled LinkedIn, Instagram, Google Workspace, and internal apps; the human approved sends, reviewed edits, and handled irreversible actions.
- How it went
- Over six months it closed a first paying client, found and closed two cold leads, replaced about ten subscriptions with owned tools, and reached five-figure revenue. It also misfired: social automation got an account suspended, drafts and queues stalled unnoticed, and AI-written engagement on repurposed content backfired as slop copy.
- Worth knowing
- The poster says multi-agent runs cost about 15x the tokens, so they only use subagents for genuinely complex parallel work.
Try it yourself with Claude Cowork
I run [type of business] and write about it on LinkedIn. Take this post: [paste post]. Turn it into [number] pieces such as [e.g. a newsletter section, a short thread, a blog outline], written in my voice using these samples: [paste 2-3 of your past posts]. Avoid generic filler phrases and keep each piece specific to my experience. Give me drafts to review and do not publish or reply to anyone without showing me first.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: Reddit · Undated
Related use cases
Run a solo company with eight agent departments
A founder set up CEO, CFO, COO, lawyer, accountant, marketing, CTO, and improver agents that share memory and consult each other.
Reconcile bank statements with invoices
Cowork cleaned vendor names, categorized expenses, matched invoices to transactions, flagged missing ones, built an Excel file and renamed invoice files.
Turn email invoices into a live financial dashboard
An n8n workflow scans emailed invoices, extracts supplier, amount, date and description, and feeds a real-time Google Sheets dashboard.
Organize Canvas courses with an in-platform agent
Instructure says early adopters use its IgniteAI Agent to rename modules consistently, bulk-change due dates, and rebuild course pages; this comes from a vendor video with no named users.
Custom agent
Turn booking confirmation emails into calendar events
An n8n workflow extracts details from booking emails and PDF attachments and creates calendar events in a dedicated calendar, built for a CEO's flooded inbox.
Interview candidates based on their resume
A built AI interview agent analyzes a candidate's resume and conducts a personalized interview for HR and recruiters.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.