Fraunhofer Institute for Cognitive Systems IKS
Fraunhofer Institute for Cognitive Systems IKS

Training »Trustworthy Agentic AI«

(Multi-)Agent systems put to the test

From the AI vision to controllable autonomy in practice

Are you considering using AI agents or multi-agent systems (MAS), but unsure how to reliably evaluate and validate their behavior in real-world operations? Traditional testing methods quickly reach their limits when it comes to AI, and safety, compliance and liability issues can arise. Without controllable safeguards and proven safety architectures, valuable AI innovations can get stuck in the proof-of-concept loop.

The “Trustworthy Agentic AI” training course from Fraunhofer IKS helps you identify precisely these risks early on and develop a robust decision-making basis for the next implementation steps from an AI idea.

In this interactive half-day workshop, our experts will work with you to analyze an exemplary use-case scenario to develop an initial feasibility and implementation roadmap. Together, we will lay the methodological foundation for a safe standards-compliant system architecture and then jointly assess the technological maturity of your use case.

Your benefit: confidence in your decisions for your aI use case

After the training, you will receive a concise management summary that outlines the risks, safeguards, and next implementation steps for your AI use case, serving as a well-founded basis for decision-making.

1. Risk transparency: Initial identification of critical regulatory and functional safety risks for your use case.

2. Compliance & safety: Application of established safety methods (HARA, STPA) and requirements of the EU AI Act to AI.

3. Safety architecture: Design of specific control mechanisms (guardrails, runtime boundaries, fallbacks).

4. Decision confidence: Evaluation of an exemplary AI use case as a well-founded basis for investment and management decisions.

Agenda & schedule

1. Technology & classification

You will learn about the current state of LLMs, tool-using agents, and multi-agent systems, and discover why traditional testing methods are often insufficient for autonomous systems.

2. Use case analysis & risk assessment

Together, we will structure an exemplary use case and identify critical interfaces, black boxes, and liability risks (e.g., in the context of the EU AI Act).

3. Architecture guidelines

You will learn how to apply safety concepts such as HARA/TARA or STPA to agents and discuss appropriate guardrails, runtime boundaries, human-in-the-loop approaches, and fallback/degradation strategies.

4. Roadmap & next steps

You will receive an initial assessment of the feasibility and technological maturity of your use case, along with specific recommendations for next steps (funding opportunities, ways to avoid cost and time pitfalls).

5. Conclusion

We address any outstanding questions, summarize the results, and determine the appropriate next steps.

Target audience

This training is designed for specialists and executives who wish to evaluate, validate, or implement AI agents or multi-agent systems. It is particularly well-suited for:

  • Innovation & Digital Managers
  • Heads of Automation / Robotics / Engineering
  • Safety & Quality Managers
  • Senior Management
  • AI/Data Engineers
  • IT Project Managers
  • Product Managers
  • Data Analysts

Requirements

Participants should have a basic understanding of technology, some experience with generative AI, and a good knowledge of their own business processes (e.g., in production, logistics, or system architecture), as well as potential AI use cases. They should also be able to assess the bottlenecks and risks associated with current AI projects. In-depth programming skills are not required.

Why Fraunhofer IKS?

We are not a traditional IT service provider looking to sell you off-the-shelf software licenses. As an independent engineering partner, we bring a deep understanding of functional safety to the table and empower you to manage the risks of autonomous systems from the very beginning.

Request your preferred date now

Would you like to assess whether your AI agent use case can be safely implemented? Contact us directly using the contact form. We look forward to hearing from you.

Thank you for your interest in the Fraunhofer IKS.

We have just sent you a confirmation e-mail. If you do not receive an e-mail in the next few minutes, please check your spam folder or send us an e-mail to business.development@iks.fraunhofer.de.

* Required

An error has occurred. Please try again or contact us by e-mail: business.development@iks.fraunhofer.de