What is an AI Agent?

An agent is an LLM with access to tools that runs in a loop, able to think on its own, take actions, and work toward an end goal without predetermined steps. It makes decisions based on available information and uses whatever tools you give it to accomplish objectives, with its effectiveness determined by the quality of the model used.

An AI agent is something where the model itself dictates the program flow, making decisions about what actions to take next rather than following a predetermined path. It's a system where the model chooses between different possible actions or tools in a flexible structure that might be different for each query or run, as opposed to deterministic workflows with fixed steps.

An AI agent is a piece of software that uses LLMs connected to tools and data, making them more deterministic by implementing logic on the tool side rather than on the agent itself. These single-purpose agents help users perform specific tasks like research, market analysis, or data extraction, with IBM focusing on providing ways to deploy and host these agents at scale.

An AI agent is something that can take an open-ended task, make a plan, run actions or tools, check its own work, and produce an output while potentially interfacing with humans. It involves decision-making, planning, and executing with a set of capabilities to achieve a specific goal or outcome.

Any application that uses an LLM. Josh believes the definition doesn't need to be complicated, and that there are already too many definitions out there, making it a "lost cause" to try to narrow it down further.

An agent is a layer framework around the underlying language model that can represent the current state and action space of the environment you're trying to get the LLM to interact with. It autonomously executes a goal, helping you achieve tasks by breaking them into planning and execution steps.

An AI agent is a large language model with access to tools, meaning with the ability to modify something outside of its context window. It receives input, decides based on that input whether to use tools and which ones to use, then uses those tools and either waits for more input or completes the task.

An AI agent is a piece of software that can accomplish high-level goals by figuring out long-tail use cases and working around roadblocks, similar to a self-driving car for software. From a UX perspective, it represents a new species of software that moves beyond traditional interfaces, allowing users to interact through chat or other modalities rather than clicking buttons and filling forms.

An agent is a set of multiple prompts and multiple non-deterministic interactions designed to generate directed user behavior. It involves adding non-deterministic complexity through a collection of prompts and models working together, going beyond just a single prompt to create more complex, purposeful systems.

A piece of software you can delegate tasks to, empowering and augmenting human agency rather than acting as a separate entity. These agents use LLMs and APIs to act in the world based on requests and can be easily replicated. They are fundamentally tools extending human capabilities under human direction.