AI Agents For Beginners

Intuitive explanation that even your dog understands

Tereza Tizkova
4 min readJun 15, 2023
Source

Recently I joined e2b — a startup focused on “deployment, testing, and monitoring of AI agents”. When I heard about the AI agents for the first time, I asked: “Why isn’t anybody else obsessing about this?” Turns out a lot of people are actually obsessing about AI agents, at least within a tech community bubble in San Francisco and Palo Alto.

I have been learning about AI agents for 10 days. I am still a rookie, but I will try my best to explain why AI agents are a thing, such that even your dog understands it and you will never have to google it again. Don’t worry, I will not use any buzzwords that you also may not understand, like “scalable”, “agile”, “language models”, “metaverse”, “blockchain”, “platform” or “digital”.

What are AI agents

You already do know some agents which you have met in real life. You’re familiar with travel agents, real estate agents, or insurance agents. The term “agent” appears even in popular culture — in movies, books, and video games, an “agent” embodies an individual acting upon someone’s command, having special skills to accomplish their mission. A perfect example of this is Agent 007, who possesses exceptional combat abilities, problem solving, technology knowledge, and charm, using all that to counter global threats.

Source

Even in real-life instances, agents usually act for someone else, to complete an assigned “mission”, using special skills and characteristics. Real estate agents help people with buying, selling, or renting properties. In legal terminology, an agent is someone who has been legally empowered to represent a client in negotiations, implying some special capabilities too.

AI agents are no different in principle, except that they carry out their unique ‘missions’ within the online world. Inevitably, the initial pioneers experimenting with AI agents are the ones who navigate and shape the digital realm most proficiently: developers. Being early adopters of all new technologies, they have naturally directed their recent focus towards experimenting with AI agents.

That’s why it may take more time until AI agents become a normal part of everyone’s life. But they already have big potential for that. In the tech community, people have already started building and using them to make their life easier.

In San Francisco, I met a founder of MultiOn, a startup developing AI agents. Just for a glimpse, here are some tasks their agents can already accomplish:

  • Booking a flight autonomously
  • Sending emails autonomously
  • Searching for vacation rentals and checking pricing for upcoming trips
  • Scheduling a car wash
  • Searching for a wedding venue and initiating the wedding planning process.

How do AI agents work

Why isn’t everyone already utilizing AI agents’ help, in a similar manner to how we employ various assistants, travel agents, insurance agents, legal agents, and many more in real life? AI agents are still in their onset, and must overcome numerous challenges before they become easier to use for “normal people”. However, I believe that in the near future, we will all be seamlessly incorporating them into our lives without even realizing it.

The last part of this article is a glimpse into what the internals of an AI agent look like. Not surprisingly, it is just a bunch of code. This screenshot is part of my other favorite agent — the “smol developer”.

smol developer

When the agent is given a task, they make a “call” to an AI model for help. An AI model is like a “supertool” generating the — mathematically determined — optimal outcomes for any request you give them. The agent’s special skills lie in communicating with such a model, deciding how to use it (together with other tools) and “chew” what the model provides. All of this is done to deliver the best possible results to you — whether it’s planning an ideal vacation or creating a stunning web page for your project. How the agents “think” and act is a topic for another article.

So until next time, folks.

--

--