21. Generative AI


Generative AI refers to Artificial Intelligence (AI) and Machine Learning algorithms that use existing content to generate new content. Generative AI can generate text, sound, images, etc. Based on models stored in a database, it can produce its own similar model. For example, today’s artificial intelligence systems can be trained to recognize a distribution network component in images, whereas generative AI systems can be trained to generate an image of a distribution network component. Generative AI is likely to have the biggest impact on knowledge work, particularly activities involving decision-making and collaboration, which previously had the lowest potential for automation.


According to Bloomberg Intelligence, the generative AI market will reach $1300 billion by 2032. It was close to $40 billion in 2022 and should reach $67 billion in 2023. The rapid growth of the generative AI market is best illustrated by the success of ChatGPT. When it was launched in 2022, ChatGPT had one million users in 5 days. The EU AI Act adopted by the Parliament in March 2024 contains specific requirements for generative AI.

Challenges and opportunities for DSOs


  • Enhance employee productivity across DSO processes. Generative AI has the potential to automate routine, repetitive content-related tasks and augment work activities that require problem-solving and abstract reasoning skills.
  • Support personal experiences. Generative AI can tailor content, products, and services to individual preferences for DSO customers, partners and employees.
  • Automate human-like content creation, support informed decision-making and augment human creativity across DSO processes.
  • Augment software development.


  • The use of generative AI will require specialised resources and adapted validation processes (bias/accuracy monitoring, privacy and security management) before establishing confidence in processes critical to the distributor.
  • Data quality. Generative AI relies on high-quality training and input data (e.g., grid data).
  • Intellectual property. Information entered in generative AI services can become part of its training set, determining ownership of content created by generative AI is complex.

EDSO Considerations

  • Generative AI solution performances are directly based on data availability and quality and emphasise data privacy concerns.
  • Generative AI algorithm performance control requires special attention and dedicated monitoring tools.
  • Generative AI technologies require specific skills and companies will have to develop specific training programs.
  • Development and implementation of Generative AI solutions must be carried out with the aim of reducing their carbon footprint.
  • Particular attention must be paid to the ethical aspect of Generative AI (potential concerns about justice, fairness, accountability etc… ).
  • It will be necessary to ensure that the results of Generative AI are always identified as being generated by a machine.
  • It will be necessary to ensure that a “human is in the loop” when Generative AI is used for decision-making and communication

Potential use cases

  • Improve knowledge management: retrieve stored internal knowledge for better-informed decisions across DSO processes, clarify complex regulations etc.
  • Boost grid management and operations by automating & augmenting tasks: incident planning & response, network development studies, grid engineering, asset performance & planning, demand load forecasting, network control, etc.
  • Enhance customer experience by generating personalized content, recommendations, and automating responses to customer requests.
  • Faster product- and software engineering: create novel designs, prototypes, code and testing them.

Ongoing projects

  • Stedin is piloting virtual assistants for knowledge discovery and augmenting content creation (e.g. question and answer to internal data including grid data, creating minutes of meetings, and drafting documents) to boost the productivity and creativity primarily of office workers.
  • Enedis DeepCourboGen is a synthetic load curve generator that preserves the statistical distribution of real curves while completely masking actual data.
  • ČEZ Distribuce developed a proof of concept including the dedicated implementation of an OpenAI-based Generative AI model as a chatbot for connection requirements information prepared for customers and tested by internal employees. Future features include a chatbot integrated into the public website, analytics of customer feedback, an internal chatbot for corporate documentation, Human Resources (HR) onboarding etc.
  • UFD is working on two projects to leverage generative AI. Both projects aim to provide UFD call center and control room operators with a co-pilot to support their activity and personalize responses to customer calls and incidents, respectively. Additionally, it is planned to use AI as a simple manager for advanced queries to the UFD information repository.

Last update: 17 May 2024