• Challenge

    Collecting data on rates and contract terms from over 50 energy providers on a monthly basis was very time-consuming, due in part to varying document formats and update frequencies. Although an LLM pilot project showed promise, it did not achieve the required accuracy and traceability.

  • Our Solution

    Development of an AI-powered workflow for the automated analysis of rate schedule and contract documents. Relevant information is extracted, validated, and corrected as needed in separate steps: EBF Flow.

  • The Result

    The manual process of data collection, preparation, and analysis has been automated and is now available more quickly and with significantly greater reliability. The AI process is transparent and, at the same time, meets all data protection requirements and ISO standards.

Challenges at EnBW-ODR

EnBW Ostwürttemberg DonauRies AG is a regional energy provider in Ostwürttemberg, DonauRies, and Hohenlohe. By 2026, it will serve 185,000 electricity customers and 18,000 gas customers. To ensure its offerings remain competitive, EnBW ODR continuously monitors the rates and contract terms of more than 50 electricity and gas providers in southern Germany. The data sources consist of documents from independent companies, each with its own reporting style, unique document types, and varying update frequencies. Reliably capturing this heterogeneous dataset and transferring it into internal systems regularly required a great deal of manual effort.

A pilot project using a Large Language Model (LLM) was intended to automate the process. However, the results were not accurate enough and led to time-consuming manual follow-up checks.

Our Solution

As a result, EnBW ODR brought EBF on board to rethink the approach. Instead of entrusting the entire task to a single model, we broke the process down into individual, traceable steps—this way, any errors become visible directly at the respective process step and can be corrected early on.

The advantages of EBF Flow in detail:

  • Smaller, more efficient model: A smaller model than a monolithic LLM is sufficient for the project. This makes operation more cost-effective, faster, and more resource-efficient.
  • Step-by-step: Documents are checked for specific values in individual workflow steps and are documented in a traceable manner. Any errors thus become immediately visible, rather than only in the aggregated final result.
  • Early human correction for iterative improvement: Users are part of the learning process: Corrections are incorporated into the instructions for the next iteration, permanently improving the workflow.
  • Adaptive: If, for example, a document’s layout changes, the workflow can be easily adjusted without having to set up the entire workflow from scratch.
  • Interface: A clear interface facilitates ease of use and traceability.
  • ISO-compliant & audit-ready: Every step is documented and traceable, ensuring that governance and transparency requirements are met.
Modern Work im Energiesektor

The Result

With EBF Flow, EnBW ODR has replaced a time-consuming manual process with a transparent and largely automated workflow. Tariff and contract data are now available more quickly, consistently, and in a more traceable manner. The solution meets data protection and compliance requirements and can be reliably maintained even when document structures change, without the need for fundamental redevelopment. And thanks to processing transparency (logging, user approval, documentation), data protection and governance requirements are met, ensuring that the AI system is audit-ready and ISO-compliant.

Our work makes a difference

“Our competitive analysis is now fully automated using AI. With EBF Flow, we have the key data from a wide variety of documents processed, saving us many hours of manual work. The combination of security, data protection, human validation, and an efficient small language model has won us over and proven its worth. We are completely satisfied with our collaboration and highly recommend the EBF team!”

Martin Hahn System Administrator EnBW ODR AG

ContactReady to get started? Contact us.

Let’s discuss how we can support your team and get your modern workplace to its best.

References: Further references

  • deutsche rueck referenz logo

    Deutsche Rück

    Targeted optimization of the BlackBerry UEM configuration reduced the burden on IT and noticeably improved the user experience.
  • stm referenz logo

    St. Thomas More School

    At St. Thomas More School, homeschooling was smoothly implemented in no time thanks to EBF and Jamf School.
  • richter frenzel referenz logo

    Richter+Frenzel

    90,000 contacts from the customer, dealer, and freight forwarder address book always at your fingertips—with automatic number resolution.