Artificial Intelligence has been, quite literally, the word of the past two years — if not the decade. However, as its popularity soars, the spotlight has been particularly on Generative AI (GenAI), which has the power to generate original and creative outputs like high-quality text, images, and videos. While its impact is already well recognised in the private sector, its potential within the public sector remains nascently explored. Some governments, like those in Italy, Germany, France and the Netherlands have taken the lead with ambitious AI initiatives. However, widespread implementation of Generative AI in public administration still faces challenges, and its true potential is yet to be fully realised. 

At the GovTech Connect Assembly on 28 November, we will be tackling these critical implementation challenges in our dedicated session on Generative AI in the public sector. Through real-world use cases and expert discussions, we’ll explore how to overcome barriers to adoption and unlock GenAI’s full potential for transforming public services. Don't miss this key opportunity to join the conversation and be part of shaping the future of AI-driven public administration! 

Implementing Generative AI in the public sector brings a set of challenges that span regulatory, operational, and technical domains. A major concern lies in the regulatory and risk management aspects. GenAI systems often function as "black boxes," where the decision-making process is opaque, even to their developers. This lack of transparency can complicate their use in public decision-making or in communication with citizens. Furthermore, the technology’s effectiveness heavily depends on the quality of its training data, which can lead to biassed or inaccurate outcomes, especially if the data is incomplete or flawed. Privacy concerns are also heightened due to the sensitive nature of public sector data, with the risk of data leaks adding another layer of complexity. 

Operationally, the cost of training and maintaining GenAI systems can be considerable compared to other digital tools. These models demand substantial computational power, which increases energy expenses and requires investment in robust infrastructure. Managing and storing large datasets also contributes to high costs, and public institutions must also invest in skilled researchers and developers.  

Civil servants may be unfamiliar or uncomfortable with such rapidly evolving technologies, and the novelty of GenAI highlights the need for significant upskilling. Governments will need to dedicate time and resources to build teams with the necessary expertise to oversee GenAI implementations effectively.  

Lastly, the procurement of GenAI technologies presents its own set of hurdles. The market is rapidly evolving, with many new suppliers, making it difficult for governments to identify reliable partners.  

Rather than viewing the challenges of implementing Generative AI (GenAI) as insurmountable, policymakers should follow best practices to ensure that the procurement and implementation process in the public sector is as efficient as possible. Key steps include defining project scope and risk appetite as well as evaluating and selecting the most suitable approaches for building or acquiring generative AI models. Policymakers are encouraged to assess potential suppliers through prototypes and pilot programs, while also fostering innovation through collaborative challenge formats. Additionally, enhancing the capabilities of commercial teams to evaluate both functional and technical aspects is crucial. Emphasising data security and ownership rights, as well as effective contract management, will further ensure successful integration and alignment with technological advancements. 

The potential rewards of successfully integrating GenAI into public services could be transformative. A BCG report suggests that the global productivity improvements driven by GenAI could be worth $1.75 trillion per year by 2033 across all levels of government. 

Generative AI can reshape public sector operations by automating processes, analysing vast amounts of data, and improving both internal efficiencies and external services for citizens. Internally, GenAI enhances the productivity of civil servants by automating routine tasks, such as document summarisation, which frees up time for more complex work. This not only allows agents to focus on higher-value tasks but also leads to quicker response times and more accurate information for citizens. Additionally, by driving innovation in policy-making through streamlined data analysis and improved planning, GenAI supports better decision-making, resulting in policies that are more aligned with the needs of the public. Several noteworthy use cases illustrate the benefits of GenAI for public agencies and citizens. 

A few European countries are leading on the implementation of GenAi within public services, paving the way for a broader adoption. 

In Germany, the city of Hamburg is piloting a large language model (LLM) named LLMoin within the city’s public administration. This LLM, integrated as a text assistant, is helping around 100 city employees with functions like research, text generation, and document summarisation.  

Similarly, in France, the government has developed "Albert," a sovereign GenAI designed to accelerate administrative processes and provide accurate, clear responses to citizens’ queries.  The tool assists agents by suggesting potential responses to fiscal inquiries and environmental project evaluations, citing relevant regulations, and providing useful links and practical guides.  

In Italy, the National Social Security Institute (INPS) has introduced a similar generative AI-powered virtual assistant, that simplifies the access to pension and social security schemes, helping citizens easily compare different schemes and check their eligibility for them, in an intuitive way. 

Generative AI holds promising potential for enhancing public sector operations, though some governments across Europe are pioneering its implementation, applications are still in the early stages. It is important that governments act proactively, starting with small-scale implementations, learning from initial experiences, and adapting their approaches as needed. 

As the technology and research in this field continue to evolve, addressing these challenges collaboratively and learning from emerging best practices will be crucial. We invite you to join us at the GovTech Connect Assembly on November 28th in Brussels, where leading experts from the government technology sector will present examples of GenAI experimentation in Government and explore how to effectively implement and maximise its potential in the panel discussion, "From Vision to Action: Overcoming Implementation Challenges of GenAI in Public Services”.