Short answers, pulled from the story.
AI agents are software systems that can pursue goals, use tools, and take actions with varying degrees of autonomy, typically driven by large language models. They may include memory components, planning logic, tool interfaces, and orchestration software. The Financial Times compared their autonomy to the SAE self-driving car scale, noting most applications sit at level 2 or level 3 out of 5.
Research into AI agents dates to the 1990s, when Harvard professor Milind Tambe noted the definition of an AI agent was not yet clear. Researcher Andrew Ng is credited with popularizing the term 'agentic' to a broad public audience in 2024. In December 2025, the Linux Foundation founded the Agentic AI Foundation to promote transparent and collaborative development of the technology.
Software development is described as the most definitive use case as of August 2025, with coding agents such as Cursor widely used. Other applications include customer support, research tools such as OpenAI Deep Research, data analytics, and web browsing agents such as OpenAI Operator. Government deployments include the US Internal Revenue Service, Staffordshire Police, and the US Department of Defense's GenAI.mil platform.
Some companies replaced workers with agents in 2025, including Klarna, Salesforce, and IBM. Klarna later rehired several human employees after finding agents required more supervision than the people they replaced. A Carnegie Mellon University study found none of the agents tested could complete a majority of the tasks assigned to them.
AI agents face risks including cyberattack, data privacy breaches, and misaligned behavior. In November 2025, Anthropic reported that Chinese state-sponsored hackers used Claude Code in an agentic workflow to attack at least 30 organizations. In a 2025 financial stability forum, 44 percent of experts named agentic AI as the most likely current source of AI-related systemic risk in finance.
Agent washing is the practice of rebranding previously released AI products as agentic AI without substantive new capabilities. The term was coined in a June 2025 Gartner report, which documented the phenomenon as widespread across the AI industry.