10-Step Framework for Running a One-Person Company with AI Agents
Michaelbetterecycle · reddit · 2026-07-17
The author shares a 10-step framework distilled from running a one-person company with AI agents for 6 months. The focus isn't on "AI writing emails," but on having agents participate in real operations like marketing, sales, CRM, content, and outreach.
1. Put the Company in a Repository
He advocates migrating business states into standard files and folders for AI to read and write directly, which is more effective than juggling ten SaaS tabs. The trade-off is fatter context, so load on demand rather than stuffing the entire company into the context window.
2. Manage Tasks with a Router Brain and Department Playbooks
The root directory acts as a router, opening the corresponding department's CLAUDE.md / playbook to temporarily turn an agent into the department head. The author believes:
- A general agent + a clear playbook is usually more stable than a bunch of fragile specialized bots.
- Subagents are only worth it for genuinely complex, parallel, multi-step tasks.
- Multi-agent token costs are about 15x that of a single agent, so use them sparingly.
3. Take Control of Core Workflow Tools
He rebuilt internal SaaS dependencies into small tools:
- LinkedIn client runs via a real browser session.
- Instagram has its own CLI.
- Google Workspace is operable from the terminal.
- Core business relies on his own databases, brand assets, and processes.
He also hit platform risk controls: an agent once performed high-speed batch operations on social media, leading to an account ban. He subsequently hardcoded limits, such as:
- 20 connects per day
- 40 DMs
- 80 profile views
He also checks DOM / element counts before and after sending to prevent the platform from faking a "success" without actually sending.
4. Conclusions
The author's overall takeaways:
- Control core processes in-house, rent peripheral capabilities.
- Write limits into code, not prompts.
- Don't trust platforms' self-reported success; verify what actually happened.
More from coding & agent
- Coding agents are heading toward an AI-writes, AI-reviews, human-approves workflow — aftahi_ai · 2026-07-22
- oMLX 0.5.2 adds Mac menu-bar stats, low-bit decode kernels, and faster downloads — awnihannun · 2026-07-22
- GitHub review bot hits its PR limit and forces a 39-minute cooldown — DanielLockyer · 2026-07-22
- Max reasoning effort appears to be mobile-only in Codex Remote, not desktop — GabGarrett · 2026-07-22
- A Reddit demo argues online stores should expose carts and pricing through MCP — gelembjuk · 2026-07-22
- Open-source AI SDK provider routes Vercel apps through a local Codex subscription — lgrammel · 2026-07-22