- Work
- Insights
Sharing opinions, ideas and knowledge.
All our news
Opinions
•
5 min. read
A good model is only a start
AI 'agents' seem to be the next thing in AI, but what are they? Simply put: agents are AI implementations that have some freedom to act autonomously.
That's right, AI can be more than just that handy chatbot that does your homework. When set-up right, AIs can act on their own, scary perhaps, but this blog is not about emotions and ethics. Let's focus first on how it works conceptually.
The good, the bad & the AI
AI models is what you hear about a lot these days, but the AI model is just a component of an AI agent; its brains basically. You can compare it with your laptop or smartphone: the processor is surely important, but on itself it's just a piece of baked sand.
You'll need other components to create the magic, such as a good User Interface, memory and data to work with. In case of an AI agent you'll also need to train it, to do the task that you want it to do.
So, yes, there's definitely some work to do when selecting the right AI model, but there's also a lot work to do in fitting the AI agent in your workflows, and creating a user interface that enables smooth human-agent cooperation.
The biggest step to take right now in AI, is to move from generic models and applications, to tailored ones that smoothly integrate with work processes, app stacks and people's lives really.
There's a lot of work to be done here and as people are used to smooth interactions with ChatGPT, the work needed to tailor an AI agent and truly integrate it, may come as a surprise. Surely, we'll live an learn, get better at it or learn AI to do it, but for now human designers will still make the difference.
