What comes next after AI agents? We had to ask
Agentic AI is so 2024. What's next?
At first, ChatGPT and generative AI were the buzzwords to describe the life-changing tech of large language models. Next came copilots. Then agentic AI, which hit the mainstream last year.
This all happened in less than three years. Logically, then, agentic AI should be hitting its expiration date, with a new flavor of generative AI coming out faster than H&M, Zara and Shein cycle through polyester prints. We should be seeing the next phase now, or at least very soon, as the technology -- and the buzzwords around it -- evolves.
Comparing enterprise AI to fast fashion might sound six kinds of crazy, considering that it's still so new. Just this month, Adobe, Qualtrics, Oracle, OpenAI and Deloitte launched agentic AI platforms, joining others such as Microsoft, AWS and Salesforce, all of which had released their own in the previous months. But here we are. Before some companies could even evaluate copilots, we moved on to autonomous agents, which themselves could be left behind for the next thing in fairly short order.
This is the speed at which generative AI is moving, according to experts at Adobe Summit last week in Las Vegas and at the online MarTech conference.
"I think all of these things are fads, and they do have shelf lives," Forrester Research analyst Brandon Purcell said. "AI as a concept is a moving target. First it was predictive stuff, and then generative, and now it's agentic. I don't know what the next thing is going to be. I think agentic probably has more than a year of staying power just because there's automation involved, there are real efficiency gains, and there might even be some top-line growth for the companies who do it well."
Regardless of what it's called, agentic AI is here to stay, according to Klaasjan Tukker, senior director of product marketing for Adobe Experience Platform. Eventually, it will just run, and its users won't think about it or talk about it -- much like we use a map app on our phone and don't think about the vast server technology it takes to plan a route from Point A to Point B. Or like we get behind the wheel of our car and don't think about the engine or drivetrain.
Cars in general are a frontier for AI's next evolution.
"Self-driving autonomous activity is full of listening, reasoning, verification ... building a plan," Tukker said. "They're already off what we now call agentic capability."
It's still unclear what 'agent' means
Yet at this point, vendors haven't agreed what agentic AI even is. Common definitions are, well, uncommon. Adobe defined an agent as something that reasons, is interactive and is autonomous. But that's just Adobe.
"I'm not entirely sure I really know what agentic means at this point; it gets abused in so many ways," said Scott Brinker, founder of the MarTech conference and current vice president of platform ecosystem at HubSpot. "[It's the] idea of, like, we can take some AI automation magic and incorporate it in things like workflows. So, we don't have workflows now, we have agentic workflows."
Whatever it is, and whatever it does, agentic AI needs to be interoperable to survive. Agents need to talk to each other and access each other's data repositories in order to do their autonomous work, which Adobe is trying to enable.
Chances are, said Phil Regnault, a partner and Global Adobe Alliance leader at PwC, that Salesforce users will use Salesforce agents in the Salesforce environment, Adobe users will use Adobe agents in the Adobe environment, and so forth. But in many cases, data from outside those vendor-specific silos will be needed for agents to take action. Standards for agentic data exchange would help keep the technology around.
"I've never deployed Adobe anywhere where I didn't see Salesforce and myriad other applications," Regnault said. "Interoperability -- whether it's at the data layer, the intelligence layer, the application layer -- is important for stacks to work."
So, with all that in mind, it's still early days for agentic AI, which seems to be another name for autonomous tech -- we think, although Adobe appears to be one of the few vendors with enough chutzpah to actually tell us what its agents do.
But if current trends hold, we'll be talking about something else by this time next year. Fear not, though: Tech buyers are evaluating or investing in AI agents now. They'll probably hang around forever, but we'll go back to calling them something unfancy, like "software bots."
As for what we will call the next generative AI advancement, "Autobot" has a nice ring to it. So does "Singularitron." Unfortunately, both of those have already been taken.
Got a better idea? Have your agent get in touch with my agent.
Don Fluckinger is a senior news writer for Informa TechTarget. He covers customer experience, digital experience management and end-user computing. Got a tip? Email him.
