Interconnected digital systems represented as networked data nodes. The image symbolizes the use of AI agents and highlights the opportunities and risks associated with increasingly autonomous artificial intelligence systems.KI-generierte Darstellung (TAB)

The potential and risks of using AI agents

Thematic background

AI agents are a technology based on the combination of agentic AI and large language models. These systems can act either autonomously or semi-autonomously, completing tasks and producing outputs independently without the need for continuous interaction with a user. Such systems can search for and store relevant information on the internet, on servers, or in local storage systems, and can interact with external systems or other AI agents.

AI agents are currently undergoing rapid development and have the potential to fundamentally transform the way work is organised. They also present opportunities and risks for political opinion formation. In addition, their use poses challenges for consumer protection and data protection. Further advances in multimodal AI and reinforcement learning, as well as the development of domain-specific agent systems, are likely to increase the autonomy of these systems considerably in the years ahead.

Objectives and approach

As part of a TA-Kompakt study, an overview of the key application areas for AI agents will first be prepared. This will include a description of different types of AI agents and their technical characteristics, as well as two to three key use cases from various fields of application. The study will focus on three application areas:

  • AI Agents in the Workplace: Possible use cases include applications in office and knowledge work, such as autonomous project management, e-mail processing, and news monitoring.
  • AI Agents for Consumers (Agentic Commerce): The focus here will be on personal shopping assistants that receive instructions via voice commands or chat and independently carry out purchases across multiple platforms. Another example is the use of AI agents for travel planning and booking, whereby an agent searches the websites of multiple providers of accommodation, restaurants, and leisure activities and generates booking recommendations tailored to the interests of users.
  • AI Agents and Democracy: This area focuses on the automated generation of content, such as news articles and social media posts, with minimal human oversight and tailored to specific target groups, as well as its potential misuse to create an artificial appearance of consensus. The study will also examine systems designed to strengthen citizens’ ability to make well-informed decisions and to support public discourse.

Building on the analysis of technical characteristics and use cases, the study will examine and describe the opportunities and risks associated with AI agents. Particular attention will be paid to the mechanisms that may cause AI agents to disregard human instructions and mislead users. The analysis will address both cross-cutting challenges, such as loss of control, data protection, and data security, and challenges specific to individual application areas.

Given the rapid pace of development in this field and the limited number of available scientific studies, the analysis will also draw on reported experiences from news portals, online forums, and social media. To this end, media monitoring will be used to track relevant feeds, news reports, and actors and may be supplemented by analyses based on natural language processing (NLP) methods. Subject-matter experts will be consulted to assess the opportunities and risks within the three application areas.

Finally, the TA-Kompakt study will identify areas for policy action arising from the increasing use and dissemination of AI agents. To address questions relating to data protection and security, consumer protection, and liability, an external expert report has been commissioned.