Donna

Inspiration

Our team noticed that legal professionals spend a huge amount of time on repetitive administrative tasks such as gathering documents, running research tools, updating files, and confirming communication steps. These tasks take away time that could be spent helping clients. We wanted to create something that keeps the legal workflow moving smoothly without constant human supervision. That idea inspired Donna, our automated case orchestrator.

What it does

Donna is a case workflow orchestrator that automates the administrative side of managing personal injury cases. Instead of requiring a user to run each tool manually, Donna coordinates the entire workflow from start to finish and ensures every step is completed properly.

How we built it

We built Donna as a modular workflow orchestrator using Python. Instead of creating one giant script that tries to handle everything, we designed Donna to control separate helper tools that each focus on a specific legal task. This makes the system easier to maintain and extend.

Challenges we ran into

Initially, we tried to build Donna using a complex multi-agent system (A2K/A2A), where different AI agents would pass data and instructions to each other. It quickly became clear that this created more problems than it solved. Agents struggled to share consistent information, formats broke often, and debugging was almost impossible. We realized that we were overengineering the solution, so we simplified to a single orchestrator model. That change made the workflow far more reliable and allowed us to focus on solving the real problem.

Accomplishments that we're proud of

We are proud that we were able to simplify a complex idea into a working automation tool that improves the legal workflow. We built a system where multiple scripts can run in sequence without human supervision, and we ensured that data is stored cleanly in a final master file. We are especially proud of how we handled challenges like file synchronization, JSON formatting, and restructuring the project after abandoning the multi-agent system. Despite setbacks, we created a reliable orchestrator that resets itself and is ready for the next case. Seeing Donna run the whole process from start to finish on her own was a huge win for our team.

What we learned

We learned a lot about designing automation systems that depend on accurate data flow. Working with multiple scripts that modify the same file taught us the importance of consistent formatting and clear ownership of each step in the workflow. We also gained practical experience with version control and resolving merge conflicts, which helped us better understand collaborative development. Trying and then moving away from a multi-agent architecture showed us that simpler solutions are often more reliable and easier to maintain. Most importantly, we learned how to break a complex legal process into smaller parts that can be automated effectively while still keeping humans in control of the important decisions.

What's next for Donna

Next, we want to give Donna a real interface and connect her to secure databases instead of local files. We also plan to expand the automation steps she can handle and improve how she communicates status updates to the legal team. Our goal is for Donna to become a fully integrated tool that streamlines even more of the case workflow.