Timeframe and Context

The project was carried out in January 2026 at the Institute for Applied AI and Robotics.

The goal was to automate and optimize e-mail processing by developing an AI-based e-mail agent.

As the sole developer, I bore full responsibility for the project, from conceptual design to the development of the agent.

Implementation and Tech Stack

The implementation involved developing an AI-based e-mail agent in Python. LangChain was used to orchestrate various functions such as reading, processing, and generating e-mails.

For language processing, a locally hosted Large Language Model was integrated via Ollama. This enabled privacy-friendly and efficient processing of e-mail content.

Challenges and Results

Ensuring consistently high response quality and correct system behavior was a key challenge. In addition, combining autonomous decision-making with human-in-the-loop approaches required careful coordination to balance efficiency and control.

The e-mail agent was successfully developed and can read, filter, summarize, and label e-mails. It can also independently create, edit, send, and delete drafts and replies.

As a result, the agent offers more efficient e-mail processing and a privacy-friendly implementation through the use of local models.