What is an AI agent and what will they do? Experts explain

In future, artificial intelligence (AI) assistants could transform the way we make decisions and even how we interact.

"An AI agent can respond to your messages 24/7, when a person is still asleep at 2AM," according to Nancy Xu, Chief Executive Officer of AI recruitment company Moonhub.

But what exactly are AI agents or assistants, how will they be applied and what are the potential implications for humans?

This was the subject of a highly anticipated panel session – What Can AI Assistants Do? – at the World Economic Forum's Annual Meeting of the New Champions in Dalian, China in June.

On the panel alongside Xu were: Xi Kang, Assistant Professor, Owen Graduate School of Management, Vanderbilt University; Darko Matovski, Founder and Chief Executive Officer, causaLens and Liu Jiren, Chairman, Neusoft.

The definition is open to interpretation, said Kang, but for her, an AI agent or assistant is "any algorithm or models powered by AI or related technology that can help people make predictions about the future or make decisions, if it is approachable enough for laymen to interact with it, to get some insights from it".

For Matovski, there is a broader definition that requires us to rethink the relationship between humans and machines: "We're going to end up in a world where 99% of all intelligence is artificial and the 1%, the human intelligence, will act as a backstop to the 99% of artificial intelligence.

"Now, we're not there yet, and we're not ready for that world, but that's the world we are embarking to create. In that world, we will need to choose, depending on a use case where is the line between autonomy and control?

"There's going to be many applications in which you can have fully autonomous agents, and there's going to be many situations in which you want a high degree of oversight of the humans. So I think the most interesting area will be where do we spend that 1% of human intelligence to oversee the agentic workflows."

Xu said she thinks of AI agents as "digital companions" or, for an enterprise use case, a "digital employee". When was studying for a PhD in Computer Science at Stanford, she said it was about building intelligence in Large Language Models (LLMs) – and today it's about giving them agency.

AI agents need to be autonomous, explained Dr Liu. "It is able to make its own judgement and then take actions. It can sense other AI agents and the environment – the sensing process is a way for them to learn."

In future, AI assistants will "unleash a tremendous amount of productivity and potentially a massive improvement in decision-making", said Matovski.

Today decision-making is based on "human gut feel", he said, and even the most advanced organizations have a very small number of real-world use cases of AI when it comes to decision-making.

But that will change: "There's this tremendous opportunity to use AI agents and more advanced forms of AI, as we develop them to transform completely how we make decisions in our society, which will lead to more equitable, more efficient, better societies."

Matovski explained the difference between causal AI and LLMs: "We want an AI that can understand the cause and effect relationships in the real world like we do, like humans ... It will be grounded to a causal understanding of the world. And we believe this is really the breakthrough technology that will unlock most of the use cases in enterprise and in policy decision-making."

Xu said: "I think the really exciting opportunity for AI agents is actually to compress the time scale that it takes to build ideas or companies to impact and make it possible to do that in a much more compressed time frame."

In the next 5 to 10 years, we will reckon with the question of what is actually uniquely human labour and what is something that AI can do, added Xu.

"I think about it as a sort of combination of silicon talent or AI agents, and biological talent, which is what humans can do. And creating that synergy between the two."

While concerns abound that AI could replace jobs in future, we do need to think about the "AI talent revolution", Xu said: "We'll have to have the societal systems in place to help people find new opportunities and also upskill.

"One of the big challenges in the next five years is people will realize that 80% of the time they normally spend on a day-to-day basis is now done by an AI."

But, agreed the panellists, AI agents have the potential to help solve talent shortages and boost growth.

"I believe that talent has traditionally been the bottleneck for growth for many of the major initiatives that we have as a society, what [would] a world of talent abundance look like? I think that's where we're headed, and that's what I'm super excited about."

Matovski said: "There's a huge opportunity for GDP (gross domestic product) growth, better distribution, all sorts of opportunities that we can create. That for me is the most interesting area, although it's not as interesting as talking about our personal lives, I think that's where we're going to make the biggest changes."

Key to the development of AI agents and all future AI technologies, will be regulation to ensure AI is safe and equitable, said Kang.

"This is a joint effort and the developers, the big players, seem to be fully aware of the importance of developing responsible AI, not just for a greater humanity, but also for the sake of business."