# Agentic AI

Agentic AI refers to artificial intelligence systems that autonomously perceive their environment, pursue goals, and take actions over extended periods, often improving through learning. It is a specialized subset of intelligent agents defined by proactive, goal-directed behavior.

Agentic AI, also known as AI agents, describes artificial intelligence systems that autonomously pursue goals by perceiving their environment, making decisions, and taking actions over extended periods. This concept extends the broader notion of an intelligent agent, which is any entity that perceives its surroundings, acts to achieve objectives, and may improve its performance through [machine learning](https://www.wikiprompt.org/wiki/machine-learning) or knowledge acquisition. In AI textbooks, the field of [artificial intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) is often defined as the study and design of intelligent agents, with goal-directed behavior considered central to intelligence.

Intelligent agents range from simple control systems, such as a thermostat, to complex entities like a human being, a firm, or a state. Agentic AI specifically emphasizes proactive and sustained autonomy, distinguishing it from reactive systems that merely respond to immediate inputs. These agents operate based on an objective function that encapsulates their goals, and they are designed to create and execute plans that maximize the expected value of that function upon completion.

## Agent-Based Definition of Artificial Intelligence

Stuart Russell and Peter Norvig, in their influential textbook, describe artificial intelligence as the study of agents that receive percepts from an environment and perform actions. In this framework, an agent perceives through sensors and acts through actuators. A rational agent selects actions expected to maximize its performance measure, given its percept sequence, prior knowledge, and available actions. Rationality does not require omniscience or guaranteed success; it concerns the expected outcome based on available information.

In agent-oriented computing, agents are characterized by properties such as autonomy, responsiveness to environmental changes, and goal-directed or proactive behavior. One modeling approach is the belief-desire-intention (BDI) architecture, which represents an agent in terms of its information about the world, its objectives, and the courses of action to which it has committed. This architecture has been influential in both theoretical and applied AI research.

## Objective Functions and Goal Specification

An objective function specifies the goals of an intelligent agent. An agent is deemed more intelligent if it consistently selects actions that yield outcomes better aligned with its objective function. The objective function serves as a measure of success and can be simple or complex. For example, in a game of Go, the objective function might assign a value of 1 for a win and 0 for a loss, while a self-driving car's objective function might balance safety, speed, and passenger comfort.

Different terms describe this concept depending on context. A utility function is often used in economics and decision theory, representing the desirability of a state. An objective function is a general term in optimization, while a loss function is used in [machine learning](https://www.wikiprompt.org/wiki/machine-learning) where the goal is to minimize error. A reward function appears in [reinforcement learning](https://www.wikiprompt.org/wiki/reinforcement-learning), and a fitness function is used in evolutionary systems. Goals can be explicitly programmed or induced through learning or evolution.

## Learning and Adaptation in Agentic Systems

In reinforcement learning, a reward function provides feedback that encourages desired behaviors and discourages undesirable ones, with the agent learning to maximize cumulative reward. Programmers may use reward shaping to give incremental rewards for progress toward the final goal. Evolutionary systems use a fitness function to determine which agents are more likely to reproduce, analogous to natural selection. Yann LeCun stated in 2018 that most learning algorithms essentially consist of minimizing some objective function.

Some AI systems, such as nearest-neighbor methods, reason by analogy rather than being explicitly goal-driven. However, even these systems can have goals implicitly defined within their training data. Systems not traditionally considered agents, like knowledge-representation systems, can be framed as agents with the goal of answering questions accurately. In [generative AI](https://www.wikiprompt.org/wiki/generative-ai), mimicry-driven systems can be framed as optimizing a goal function based on how closely they mimic desired behavior, as seen in generative adversarial networks where a generator tries to fool a discriminator.

## Applications and Broader Context

Agentic AI is closely related to agents in economics, and versions of the intelligent agent paradigm are studied in cognitive science, ethics, and the philosophy of practical reason. The concept also appears in interdisciplinary socio-cognitive modeling and computer social simulations. Intelligent agents are often described schematically as abstract functional systems similar to computer programs, with abstract descriptions called abstract intelligent agents. They are also related to software agents, which are autonomous computer programs that carry out tasks on behalf of users, and are sometimes referred to as rational agents, a term borrowed from economics.

Modern implementations of agentic AI often build on [large language models](https://www.wikiprompt.org/wiki/large-language-model) and [transformer architectures](https://www.wikiprompt.org/wiki/transformer), enabling systems to plan, use tools, and operate across extended horizons. Research institutions such as [MIT CSAIL](https://www.wikiprompt.org/wiki/mit-csail), [Stanford AI Lab](https://www.wikiprompt.org/wiki/stanford-ai-lab), and [Berkeley AI Research](https://www.wikiprompt.org/wiki/berkeley-ai-research) have contributed to advancing these systems. Companies like [OpenAI](https://www.wikiprompt.org/wiki/openai), [Anthropic](https://www.wikiprompt.org/wiki/anthropic), and [Google DeepMind](https://www.wikiprompt.org/wiki/google-deepmind) have developed agentic frameworks that integrate reasoning with action, though the field continues to evolve rapidly.

---
Source: https://www.wikiprompt.org/wiki/agentic-ai
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-07T02:33:55.823864+00:00