OpenAI chairman Bret Taylor lays out the bull case for AI agents

We still didn’t get a straight up definition of exactly what an AI agent is during Bret Taylor’s Mobile World Congress fireside chat in Barcelona on Tuesday. The Sierra founder and OpenAI board chair preferred to side-step CNN moderator Anna Stewart’s question asking how “agentic AI” is “any different to a GenAI chatbot” by suggesting everyone hates the former but is delighted by the “empathetic” responses AI agents can serve up.

Given his new startup is building a customer service AI agent, you’d expect Taylor to be evangelical about the tech’s potential. And he did not disappoint: “I am more excited about large language models and this current wave of technology more than any technology I can remember, perhaps since I discovered the internet when I was a teenager,” he told conference delegates.

The step change with generative AI-fuelled customer service AI agents versus earlier iterations of AI chatbots is just a much higher level of capability — such as AIs that can be “multilingual and instantaneous.”

“I think we’re in this era now where these AI solutions are actually better than the alternative,” he said, adding: “We work with companies like SiriusXM in the United States, or ADT home security, where if your alarm stops working an AI will help you fix it, and you don’t need to wait for a field service team to come to your house.

“And what’s remarkable about these agents is people actually really like them.”

Supercharging customer experience

These more capable AI service bots are helping companies shrink the costs of customer service, which Taylor suggested will help raise the bar overall. “I think it’s just going to improve the consumer experience for so many brands,” he said.

Bots that are too capable can lead to fresh challenges, as well, though, he conceded, noting examples where customer support AI agent have “hallucinated” refund policies that don’t exist in response to a customer bereavement.

Brands developing appropriate “guard rails” for their AI agents is thus an important piece of safely implementing the tool, he said. But he was bullish that this challenge will shrink as customer service agents become increasingly tailored to each brand’s use-case and policies.

“In general, my philosophy is, don’t wait for the technology to be perfect. In fact, it may never be perfect — but narrow the domain that you’re working on so you can take these intractable problems and make them solvable,” he said.

“Rather than trying to solve all the world’s AI problems, you narrow it to a domain and say, ‘Hey, we’re going to put in some practical guard rails around this AI so we can solve problems right now.’ And I think that’s an opportunity for every company at this conference,” he said. Alongside his own customer service focused AI agent company, he name checked AI code assistant Cursor and OpenAI-backed legal tech Harvey as examples of AI specialization that’s successfully applying AI agents in a defined domain.

Taylor’s take on how seminal AI agents could become for brands in the future was also unsurprisingly maximalist. “I think most companies, AI agents will actually be as significant as their website or their mobile app in terms of the percentage of interactions they have with their customers,” he said. “It wouldn’t surprise me for most brands here if, in fact, if you fast forward five or 10 years, their AI agent is their main digital experience, which I think is kind of hard to imagine right now. But I really do think that’s where the world is going.”

How people interact with AI agents is likely to shift, he also suggested, envisaging that user interfaces for interacting with these bots will fade more into the background as technologists look for ways to make it even more effortless to tap into the tech’s utility.

“I do think that — I’m hopeful — that everyone staring at their screens all the time will start to melt away as a social habit. And with the advent of conversational AI, when software can truly understand how we speak, that computers will sort of melt away, and devices will kind of melt away, and I think that will be very exciting,” he said. As a parent, he said, he hopes his own kids “don’t need to stare at a screen their entire life to engage with technology.”

Responsibility for reskilling

What about the disruption that customer service AI agents could have on jobs?

Taylor said it’s a valid concern but again expressed optimism that the shift will ultimately be good for humanity — anticipating that while some job roles will go away, new ones will open up in their place. But he added that “technology makers have a responsibility to have that conversation and not just simply deliver the technology.”

The big risk with an AI-fuelled jobs shift is that the necessary reskilling won’t be able to keep pace with the rate of change, he said. “When disruption happens faster than society can reskill it is a disruptive force. So fundamentally, I think it requires public, private partnership.”

The moderator also asked the OpenAI board chair about the AI giant’s plan to switch from being a nonprofit to a for profit venture, which has attracted some critical attention.

Taylor said OpenAI’s stated mission to develop artificial general intelligence that benefits humanity hasn’t and won’t change — even as he also said it hasn’t yet settled on what its future structure will be — but he chose to highlight the costs of developing AI technology, which he said are “quite high.”

“Whatever we do, we want to amplify that missionm and that’s the bar that we’re holding ourselves,” he said. “The mission won’t change. And in fact, the structure … will, I hope, enhance that mission, and that’s the way we’re thinking about it.”