AI agents: putting AI to work

This year is set to be all about Al agents. Nicole Kobie reveals what this means and whether the backlash is beginning before Al agents are even here

- Nicole Kobie

Hey, assistant! I sent you a link to a form. Can you dig out the sales figures from last month in order to fill it out? I'll be at lunch, so message me any questions.”

Such dull jobs used to be saved for junior staff and interns, but soon AI may be taking them on – well, that’s the promise of Al agents, according to those developing the smart automation tools.

It’s high time for AI to start proving useful, as various surveys suggest a third of businesses continue to be too scared of the risks to consider implementing the technology (Indeed), seven in ten employees aren’t convinced it makes them more productive (Resume Genius), and a Goldman Sachs report suggested the benefits remained too little given the costs.

That’s why Al developers – Google, Microsoft, OpenAl and Anthropic, among others – are beginning to release details about their own Al agents, sparking predictions that the digital helpers will be the big story in 2025. Microsoft and Anthropic have already launched agents, Salesforce believes its customers will have a billion agents in use by 2026, and OpenAI chief product officer Kevin Weil said in October: “These more agentic systems are going to become possible, and it is why I think 2025 is going to be the year that agentic systems finally hit the mainstream.”

But warnings are already sounding on security risks, poor performance and employee trust, while analyst house Gartner says there’s a risk that companies will be overwhelmed by agents that no-one can remember the purpose of – similar to the problems raised by creating too many bots in robotic process automation.

The Al industry seems dead set on making this pivot to agents, so here’s what we know, how it will impact you, and how they’re already being used.

Al agents vs agentic AI