AI Agents: Friend or Foe?
Are Agentic AIs Worth the Trouble?
AI agents are very big news.
Excitingly, they move us away from AI systems that merely carry out individual tasks, into a world of systems that work toward larger, multi-focus goals, by assigning individual tasks to other AI tools.
They do this without the need of human prompts for each new instruction.
You can now have your own powerful, far-reaching AI agent working 24/7 as your personal assistant, business manager, fitness manager, travel agent, education planner and more.
An agent could research and book your next vacation from start to finish, or organise a digital tutor for your children. (That is, of course, if you trust the AI tutor not to become your child’s “best friend”, which is already proving a problem for some parents.)
Designing an AI agent is something pretty much anyone can do, for free, using online platforms. You don’t need much knowledge of or experience with programming, either.
Obviously, there are many benefits. But they come at a cost.
For a start, there’s the obvious challenge of powering all these agents. We’re starting to realise just what data centres - and any supporting power plants - will cost, for the economy and the environment.
There are challenges with privacy and online security, too. In the next couple of years, as many as 80% of businesses will adopt AI agents to handle at least 15% of daily decisions.
That’s going to require the handing over of huge amounts of extra data to AI platforms. This increases the risk of surveillance and corporate espionage, not to mention personal ID theft.
The same will be true on a personal level. If you did want your AI agent to research and book your next vacation, you’d need to share all your membership data for things like airlines, car hire companies and hotels, plus your dietary preferences and more.
The agent will likely also need access to your email, calendar and credit card details. Even with the very best data regulations, you’ll have very little control over what happens to your data once it’s in the system.
Ethical implications of widespread agent use include concerns about digital addiction and reality distortion, where people confuse the virtual world for reality.
The widespread use of AI agents will also add to the digital disadvantage felt by people who have little or no experience with digital tools, or the money to pay for them. In the workspace, people who have very little natural talent for computers, or who aren’t given adequate training, are going to struggle.
There will be challenges with trust. Today, only 27% of employees fully trust autonomous agents. More than half of organisations that use AI agents report growing levels of anxiety among their employees, as well as fears about job loss.
The bottom line is this: we need to be wary about injecting AI agents into the nitty-gritty of our daily lives. The tech might be a good servant, or even a collaborator, but it could prove a terrible master.
We might easily, if unwittingly, slide into the latter situation by trusting agents to represent us without our constant oversight.
Oversight requires deliberate scrutiny, and a certain level of data literacy. Following the trail of an agent’s activity on your behalf will sometimes be a tricky process.
AI agents represent exciting new possibilities but, as with most things involving advanced technology, they’re also a potential menace.
| 
