According to new Deloitte research, about 52% of respondents considered AI agents systems and 42% believed multi-agent systems were the technology that would “drive the future”. So what is an AI agent? Will you have to use AI agents in your job?

So here’s the thing about AI (short for Artificial Intelligence): When people talk about AI, they are usually referring to Large Language Models (LLMs). LLMs predict what word comes next, so the more words they are ‘trained’ on, the better are their predictions. This is where the ‘Large’ in LLMs comes from. Being large, LLMs need lots of computing power, which is why you normally need to connect to the Cloud in order to use them. By contrast. Small Language Models (SLMs) are small enough to fit on small devices like an iPhone, but their accuracy suffers.

Because LLMs (and SLMs) predict the next word and the word after that and then the next word and so on, they aren’t actually intelligent and have no ‘understanding’ of what they are saying. This is why they are usually helpful, but can get things badly wrong.

Anyway, back to AI agents: Satya Nadella, the CEO of Microsoft drew a lot of attention to AI agents recently (Dec 2024), by saying the “notion that business applications exist” could “collapse” with the arrival of AI agents and went on to explain that applications are basically a (graphical) user interface that uses business logic to interface to a database. In this sense, at some time in the future, an end user may use plain English to tell an AI what it wants. The AI then acts as an agent for the end-user by creating the user interface and setting up the business logic to interface with any database housing the required data.

In other words, everyone may end up using AI agents, instead of the programs we’ve grown up with (like Word and Excel), not just IT specialists.