Agentic AI – everything under control?

Risk Management in the Era of Autonomous AI Systems

Agentic AI represents a form of artificial intelligence capable of independent thought, action, and learning. Today, we already see applications in fields such as robotics, traffic management, and automated agriculture. However, these opportunities bring new challenges. In our white paper, you'll discover how businesses can harness these technological advancements, the associated risks, and the important role of well-structured AI governance. How can one fully leverage the benefits of Agentic AI without losing control?

Generative AI systems have made rapid progress in a short space of time. A wide variety of models, each with its own strengths and weaknesses, is now available on the market. Furthermore, the integration of generative AI systems is steadily advancing. As most AI models operate in the cloud and come with appropriate interfaces, connecting them to other IT systems is quite straightforward. These conditions have organically ushered in the next evolutionary stage of AI: Agentic AI.

Risks and Challenges in Deploying AI Agents

• Technical Risks: Uncontrolled "cascade failures"
• Legal Risks: Liability issues in autonomous erroneous decisions
• Organisational Risks: Unclear decision-making and responsibility structures
• Ethical Risks: Loss of human oversight

Approaches for Advancing AI Governance

To adequately address risks and prevent loss of control, the deployment of Agentic AI systems in a corporate context must be accompanied by an advancement in AI governance. The good news is that the same principles apply as with existing control frameworks for artificial intelligence. Companies can build on common security and control practices that are ideally already in place.