Late last week OpenAI introduced Operator, an AI agent that can go to the web and perform tasks for you. This essentially gives ChatGPT its own browser so it can look up webpages, scroll, type into them, and otherwise interact with the web on your behalf. It’s currently a research preview and limited to people with a ChatGPT Pro account, currently costing $200 per month.

Now that the gauntlet has been thrown down, I expect that we will see “The March of the Agents” from OpenAI, Google, and Anthropic, rolling out over the course of this year. I fully expect by the end of 2025 to be able to have an AI agent successfully order a pizza for me without any interaction from me aside from the instruction. And I expect this will be possible without a $200 a month account(!).

Up until now, our interactions with LLMs have been through an app or a web window where we direct a text conversation. Agents will change that, and it’s coming soon.