Legal Tech Service

Scoring service based on internal and external data to automate human decision making.

About the client

The client is a legal tech company for flight claims. Users can submit a claim and, if approved, they receive a payment within 24 hours. The company earns money by forcing settlements or winning in court.

Background & Challenges

Inspecting cases is a complex process and requires a deep understanding of each case’s respective legal context. Paying the user immediately poses high financial risks. Many employees were needed to manually check a variety of conditions and data points, which is error-prone.

Pythomation’s Solution

We built a machine learning based classification service to automate human decisions:

  • We designed a decision tree to automatically accept or reject basic cases
  • Decisions are based on a customer’s web form information and 3rd party data, like weather or flight statistics
  • More complex cases are assessed by a random forest model using xgboost, which assigns an acceptance score per case
  • Employees are now assisted by the new algorithms, which has reduced the time per case drastically
  • Additionally, we provided them with a small live reporting application to monitor automatic acceptance/rejection rates


  • Increase in number of accepted cases by 24% (based on the same amount of leads) due to improved risk assessment
  • Reduction in the FTE costs of assessing cases by 50%

Get in touch

Write us



PYTHOMATION is a software company based in Berlin with a focus on the efficient development of high-quality Python applications using verified tools like Django/Flask and Pandas. We fell in love with Python because of its simplicity and elegance. We are Python & math experts with a long history in the digital world. Specialised in process automation for eCommerce platforms, we build custom-tailored Sales & CRM software, Marketing Automation tools, and scale complex decisions and processes easily with customised algorithms that utilise thousands of data points within seconds.