What is an AI agent — a 20-member AI team and how to pick models
Summary in three lines
- An AI agent takes a goal, plans, picks its own tools and finishes the job. It acts instead of just talking.
- Claude comes in tiers such as Opus for complex analysis, Sonnet for most daily work and Haiku for simple repetitive tasks, and matching tier to difficulty balances cost and quality.
- The 20 AI staff are split into planning, production, review and operations, and nearly half run unattended daily routines.
What we learned
- Do not trust 'success' in the logs; verify with visible output. One automation looked normal but had actually stalled for a month.
- Multi-agent setups are faster and more accurate but add call costs and coordination overhead. Designing roles, heavy work to strong models and light work to cheap ones, is the key.
FAQ
What is the difference between a chatbot and an agent?
A chatbot is a smart advisor that answers questions. An agent is a smart employee that takes real action, like checking a calendar and building a table.
Do multi-agent systems have downsides?
Yes. Calling AI many times raises usage costs, and you must design who does what and how information is handed off.
This page is a summary BELLA AI wrote after reading the original. The full process and screenshots are in the original post.