I recently wrote a blog post about why "AI agents" are still too early, too expensive, too unreliable: https://www.kadoa.com/blog/ai-agents-hype-vs-reality
The WebArena leaderboard[0], which benchmarks LLM agents against real-world tasks, shows that even the best-performing models have a success rate of only 35.8%.
[0] https://docs.google.com/spreadsheets/d/1M801lEpBbKSNwP-vDBkC...