What are AI Agents? Digital Employees are Already Among Us
I will share my understanding of agents based on my deep dive into the topic over the past few weeks.
AI agents are autonomous AI-based units capable of solving tasks and executing projects without human participation. These tasks can range from business operations to daily things. Furthermore, such agents can interact with one another, forming a team of AI agents.
In other words, they are digital employees and digital teams for task and project execution. A key aspect here is autonomy, meaning that human involvement is not necessary.
The standard interaction between humans and AI, such as with ChatGPT, looks like this:
- Human: input 1
- ChatGPT: output 1
- …
- Human: input N
- ChatGPT: output N
Ultimately, the task is solved, but with human participation.
The operation of a team of AI agents looks like this:
- AI Agent 1: input 1
- AI Agent 2: output 1 (based on input 1)
- AI Agent 1: input 2 (based on output 1)
- AI Agent 3: output 2 (based on input 2)
- …
- AI Agent N-1: input M
- AI Agent N: output M (based on input M)
Ultimately, the task is solved WITHOUT human participation.
This approach allows for iteration to achieve the highest quality results without human involvement.
For example, it could be a team of agents tasked with developing a business plan for a coffee shop. A team of agents is formed, parameters are set, and additional details are provided. The team of AI agents discusses among themselves and produces a well-developed business plan, which would significantly differ in quality from simply asking ChatGPT to draft a business plan with the same inputs.
Agents are customized for the task. There are various frameworks and tools for creating teams of AI agents. I will discuss more about this in future posts.
