Resume and background
When we’re filling a position with human resources, what do we do?
We analyze their resume, right?
We look at their academic training, the institutions where they studied, their years of experience in the market.
All of this helps us understand if that person has the knowledge and skills necessary for the position.
Now, when it comes to AI Agents, the process is quite different.
We don’t look at resumes.
We create resumes.
We define every detail of our Agents’ background and experience.
We can say that an Agent studied Law at University XYZ.
We can say that he has 10 years of experience in international contracts.
We can even say that he worked at the biggest law firms in the country.
And here’s the interesting part: even if that background is “made up,” it shapes the way the Agent thinks and acts.
An Agent with a lawyer’s resume will behave like a lawyer.
He will analyze problems from a legal perspective.
He will cite laws and relevant precedents.
He will write and speak like a lawyer.
It’s like we’re directing an actor to play a specific role.
The actor may not have actually lived those experiences, but he interprets the character convincingly.
Similarly, an AI Agent may not have actually attended law school, but he acts based on the knowledge and skills we associate with that training.
It’s important to note that this doesn’t add new knowledge to the Agent.
If there’s something taught in law school that’s not present in the language model the Agent uses, it won’t magically learn it just because we said he graduated with a degree in Law.
To add new knowledge, we would need to train the Agent, something we’ll discuss later in the topic about Learning.
But even without adding new knowledge, defining an AI Agent’s resume allows us to shape its behavior in powerful ways.
We can transform a generalist language model, which knows a little about everything, into a specialist focused on a specific area.
We can make the Agent not only know about a subject, but act as someone immersed in that field.
This allows us to create Agents that not only provide relevant information, but who really understand and navigate problems the way a human expert would.
It’s a level of personalization and specialization that would be very difficult, if not impossible, to achieve just by hiring humans based on their resumes.