Spend enough time talking to tech executives and you might come away thinking generative artificial intelligence - the technology behind ChatGPT and other services that can create text, video and images - is about to upend every aspect of our lives.

AI co-pilots, assistants and agents promise to boost productivity with helpful suggestions and shortcuts. Today, AI tools can do simple, specific tasks such as to craft emails, make presentations or write code, but soon, tech executives say, they will usher us into a sort of efficiency nirvana, where digital assistants anticipate and deliver our every need before we have to ask.

One problem: Tech’s hype machine moves faster than anyone can actually build the tech. The technologists are not waiting for it to catch up to its promises - they’re already on to the next thing. And what comes after AI agents? Agentic AI.

It is a fancified way to say something acts like an agent. Unlike chatbots, which require a human to type in a prompt before it can spit out a response, agentic AI can act on its own. A customer could create a complex goal, like predicting which factory machines will need maintenance or booking a trip, and the AI would automatically complete the required tasks.

Or at least, that’s the idea. Most agentic AI is still in the “possibility” stage. And that means it’s a great time for tech companies to promote the heck out of it.

Agentic AI “heralds a transformative era,” a startup called Humans.AI declared on its blog. “We now live in an agentic economy,” the CEO of Humanic AI, a different startup, wrote on LinkedIn. Snowflake, a provider of cloud computing services, recently released an agentic product that it says will let customers “talk to their data.”

Startups are even naming themselves after the idea. Agentic.ai, founded in 2021, makes technology that can play video games for testing or to play alongside humans. Agentic Labs, founded last year, makes software for building software. Agentic Systems, founded this year, is operating in secret.

Andrew Ng, a prominent AI researcher, helped popularize the term this year as a way to quell debates over which technology should be considered an agent and which should not. “There’s a gray zone,” he wrote in a June newsletter. He concluded that “agentic” was an umbrella term encompassing tech that wasn’t strictly an agent but that had agent-like qualities.

In other disciplines, agentic has meant very different things. Psychologists have used it to describe humans’ capacity to exercise control over their lives. And privacy advocates have used it to describe technology that gives people control over how their data can be used.

Zoe Weinberg, a venture capital investor whose firm invests in the privacy- and security-focused kind of agentic tech, said it was ironic to see a term that started out describing human agency being used to talk about its opposite - technology that operates with little or no human oversight.

It was, she said, “a very classic Silicon Valley trope.”

 - Ny Times