In short
Let’s explore what agency is and why this key property distinguishes an AI agent from a classical deterministic script. The difference lies not in the technology, but in how the task is carried out: a rigid algorithm versus independent decision-making.
When people talk about AI agents, they often confuse them with ordinary automated scripts. Formally, both are programs that perform a task without constant human intervention. But there is a fundamental difference between them, and it’s called agent-based behavior.
Let’s consider the classic approach: a programmer writes an algorithm that visits three specific websites, extracts headings from specified HTML blocks, concatenates them, and sends the result to the user.
Such a script behaves predictably but lacks flexibility:
This is a tool: it’s useful, but requires constant monitoring and manual adjustments whenever conditions change.
Now consider another scenario: a user asks the agent to “monitor topic X, select the most important information, and send a summary in the evening.” In this case, the program does the following on its own:
If the source website changes, the agent adapts. If nothing significant has appeared on the topic that day, it will honestly report this, rather than sending out an empty template just for the sake of formality.
This is no longer a set of rigid instructions, but an agent capable of making decisions within the framework of the set goal.
Agency isn’t about using neural networks per se, but about the degree of autonomy a system possesses. A program can use a language model and still remain a deterministic script if all its steps are strictly defined. Conversely, an agent only becomes an agent when it chooses its own path to achieving the goal, rather than simply executing a predefined scenario.
This distinction determines when to use simple scripts and when an agent is truly needed—one capable of adapting to changing conditions without constant human intervention.
Source: Habr