AI-Agents in der Praxis

Module 1 – What Determines an AI's Response?

• Model, instruction, and context: understanding how they work together

• Claude, ChatGPT, and Microsoft Copilot compared

• Why context makes the biggest difference

Module 2 – From Chat to Agent

• Assistant, scheduled task, agent: the terms explained simply

• Levels of autonomy: suggest, prepare, execute

• What types of roles are agents suited for in the maritime industry

Module 3 – Practice I: Your Own Assistant

• Build an assistant for your own task, no prompting experience required

• Define instructions, knowledge, and tone yourself

• Generate, test, and refine results as finished documents

Module 4 – Agents in the Workflow

• From a single chat to an automated process

• Connections to folders, email, and calendar: read permissions, write permissions, data flows

• What happens when an agent gets access to company data

• Data protection: where data is processed and what is EU-compliant

Module 5 – Practice II: Your Own Agent at Work

• An agent handles a task independently, e.g. in the email inbox

• Define approvals: what may the agent do on its own, with your consent, or never?

• Deliberately provoke errors and apply them to your own task

Module 6 – Critical Reflection and Outlook

• Limits of DIY development: when system integrations, data sensitivity, and the consequences of errors require professional support

• Evaluate your own use cases by value, data sensitivity, and effort

• Outlook: what companies can expect over the next twelve months