Running a small web development studio means managing the constant tension between developer cost, client revision cycles, and project volume. We've been experimenting with AI to change the economics of basic WordPress site delivery.
The old workflow
Client briefs → developer selects theme → developer populates content → client requests revisions → developer makes changes → repeat. Each revision round costs real developer time. Most clients want unlimited revisions; most budgets don't support it.
The new workflow
Client submits references + content (text, images) → Claude selects an appropriate WordPress theme and populates it → client submits revision requests via Telegram → Telegram bot routes to AI → AI applies changes directly on the VPS.
Human developers are reserved for custom functionality that genuinely requires it.
The technical setup
- VPS running LEMP stack (Nginx, PHP 8.3, MySQL)
- WP-CLI for programmatic WordPress control
- Telegram Bot API as the client communication interface
- Claude API for theme selection logic and content adaptation
- Custom Bash autopilot scripts for site installation
What works well
Standard corporate sites and landing pages deliver significantly faster. Revision cycles that previously took 2-3 back-and-forths resolve in minutes. Client satisfaction is actually higher because they feel in control through the Telegram interface.
What doesn't work
Complex custom functionality, third-party API integrations, and non-standard UX requirements still need human developers. But these are the minority of projects.
We run this through our studio at webmaster.co.ua. Still iterating on the workflow, but the core approach is solid.
Happy to answer questions about the technical implementation — it's a surprisingly straightforward stack once you have the pieces connected.
Top comments (0)