Conversely, every person calling LLMs "agents" never had any credibility to begin with. Despite my many attempts to coerce an example out of people here, I am still waiting for one (1) singular demonstration of "agentic software" that is capable of replacing production-grade software at scale. Software created by agents that solves a real-world problem and is used by tens of thousands or millions of people. Candidates include an OS, a web browser, an IDE, image editing software akin to Photoshop, a fully-featured Discord/Slack/etc. replacement, a non-trivial video game, music production software, enterprise-grade database software etc., really anything that isn't just another AI tool to produce another AI tool to produce another AI tool culminating in complete psychosis and detachment from anything people do in the real world. If you would like to be the first to provide evidence that these things are more capable than chatbots in concrete terms, by all means, go ahead.
I build agents for banks for a living and can tell you for sure that you are wrong.
LLMs as chatbots are the least interesting application of LLMs.