AI Agents
Definition
AI agents are systems that perceive their environment and act upon it; in LLM-based agents, a language model controls actions such as tool calls.
AI agents perceive their environment and perform actions to pursue given goals. In language model-based agents, the model controls, for example, tool calls. Their results feed into further steps until a final answer is produced or the process ends.
Simply explained
A digital assistant, for example, researches travel options and prepares steps while you explicitly approve specified decisions yourself.
Why it matters in practice
Agents can access knowledge sources and execute work steps. The application defines which decisions they are allowed to make independently.
Example from business
Possible example: A wholesaler has incoming complaints checked automatically, compared in the ERP and prepared credit notes.
Limits and pitfalls
- Unlimited loops can run for a long time. Input and output checks as well as human approvals only take effect if explicitly set up; tool calls require appropriate limits.
Origin & history
ReAct was introduced as a preprint in 2022 and published at ICLR in 2023. A survey on language model-based agents followed in 2023. These contributions mark developments but not a clear origin of AI agents.
Use in mid-sized companies
- Possible fields of application: CRM data maintenance, invoice verification, supply chain monitoring, and research across multiple internal sources, each with defined rights and approvals.
Frequently asked questions
Can agents make mistakes?
Yes, language models can generate false content. Tool calls can help but do not reliably eliminate the problem.
What distinguishes ReAct from pure reasoning?
ReAct combines reasoning with actions through which the model obtains additional information.
Sources
- ReAct: Synergizing Reasoning and Acting in Language Models · Yao et al. (2023): ReAct: Synergizing Reasoning and Acting in Language Models
- Running agents: OpenAI Agents SDK · OpenAI (o. J.): Running agents: OpenAI Agents SDK
- A Survey on Large Language Model based Autonomous Agents · Wang et al. (2025): A Survey on Large Language Model based Autonomous Agents
Sources checked on 14.10.2026. Editorial application examples are hypothetical; no evidence of real customer results.
Editorially reviewed on: