From responses from respected forum members and online research, I've learned a few things:
- Why shouldn't AI take full control of a website?
1. Lack of intuition and critical thinking.
2. Security Risks.
3. Unpredictable updates.
- Where can AI be a great help?
Instead of providing complete autopilot, it's better to use a "human-in-the-loop" approach, where the AI makes decisions and you approve them.
1. Log monitoring and analysis. An agent can analyze server logs (CPU, RAM) 24/7, promptly detect anomalies (404 error spikes, suspicious activity), and notify you.
2. Backups and sandboxes. AI can automatically clone a website, update plugins on a test copy, check them for errors, and only then recommend an update for the main website.
3. Testing (QA). Using specialized software, AI can simulate real user actions (login to the website, add an item to the cart, checkout) and report errors in real time.
Bottom Line: Automate monitoring, error detection, and preparing update drafts with AI. However, always leave vital operations, such as approving CMS core updates, changing the database structure, or configuring security, under human control.