Salary and compensation
When I was a child, I didn’t understand the expression “sell like water.”
I thought: “I barely buy water. I almost never go to the bakery to buy a bottle of water, since I have water at home.”
But as I grew up, I began to understand that the water at home was also purchased.
It’s just that it was purchased in a different way.
It was so “purchased” that there was a water pipe at my house, already selling the water.
And my parents paid at the end of the month for what we used.
With AI Agents, it’s the same thing.
There’s an “intelligence pipe” connected via the internet to your agent, bringing its “brain.”
What you’re actually buying is the intelligence.
When you “buy” your agent’s brain, what you’re paying for is that intelligence that comes through the “internet pipe.”
The intelligence is processed there in the data centers, and the result of that intelligence comes through the pipe.
And you keep paying based on usage.
So, your AI Agent’s “salary” will depend directly on how much you use it.
Unlike a person, who is usually paid by the hour, here you literally pay the agent per thought.
Big thoughts, small thoughts, and the quantity of them.
And also how smart the brain that’s thinking is.
There are several ways to set how much you want to pay for that intelligence.
But the basic principle is the same: you pay by usage, by processing.
Now, it’s also possible to run an AI brain locally, without needing the internet.
Some AI brains, like those from Meta (Facebook) or DeepMind, are available for download.
You can download these models and run them on equipment right within your own company, without needing an internet connection.
Yes, artificial intelligence can run on a computer right in front of you!
In this case, your investment in the AI Agent is an investment in hardware, in a computer powerful enough to “think.”
And from there, the cost is basically the electricity consumed at each “thought.”
But I’ll tell you upfront that, unless you invest millions in such a machine, you’ll only be able to run “dumb” intelligences.
However, there’s still room for agents that run locally, mainly for doing small simple tasks in multi-agent teams.
In this kind of arrangement, the more complex tasks go to the more intelligent agents that run in the cloud.
While the more basic tasks can be delegated to these simpler local agents.
It’s all a matter of strategy and resource optimization.