ARIC Insights: LLMs and Agentic Coding

Background 
Large language models (LLMs) are increasingly being used not only as conversational systems but also as agents that can plan and execute tasks in pursuit of a defined goal. In software development, this development is particularly visible: instead of merely generating code suggestions, LLMs can write, modify, test, and refactor code, and interact with external tools and environments.

These capabilities lower the barrier to creating small, task-specific applications, accelerate prototyping, and enable more individualized software solutions. At the same time, they raise questions about how software development practices, roles, and quality standards may evolve.

Content of the event 
The talk will begin with a brief conceptual framing of how large language models work and how tool integration enables agent-like behavior. Building on this foundation, Pascal will discuss practical examples from software development and reflect on broader implications for the field.

Topics include:

• A short conceptual overview of LLMs, tool use, and LLM-based agents

• The spectrum of AI support in coding: from autocomplete to agentic workflows

• Practical experiences and observed limitations

• Possible implications for software engineering processes and roles
• Open discussion and exchange of experiences

Target audience 
This session is primarily aimed at technically inclined participants and practitioners who are interested in the practical use of AI in software development. No programming knowledge is required, but a general familiarity with technical concepts will be helpful.

About the speaker 
Dr. Pascal Wichmann is AI Manager at ARIC and Head of AI at Caps & Collars

This event 
This session is part of the ARIC Insights series, in which ARIC offers in-depth perspectives on current AI topics. The event will take place online and participation is free of charge.

The webinar will be conducted in English.