This post was originally published on this site.
Are public servants in Poland ready to embrace AI, or does the technology still spark uncertainty? For Pamela Krzypkowska, AI presents an opportunity to enhance public administration.
In this interview, Pamela discusses Poland’s digital transformation and the development of PLLuM, the country’s first large language model tailored for public administration. Designed to streamline processes and support overstretched government teams, PLLuM is a step toward making AI an integral, practical tool for public service.
But integrating AI isn’t just about technology — it’s also about trust. With public attitudes towards AI shifting, clear communication and responsible implementation are essential. Pamela shares how Poland’s focus on social research wants to track these perceptions and explains why transparency is key in this environment. This conversation explores how Poland is working to ensure AI is a tool that works for the public, not against it.
Q: What are some of the biggest challenges in leading Poland’s digitalisation strategy, particularly when integrating AI across public services?
One is that policy and regulation never follow technological innovation on the same timeline. We always need some time to adapt, which creates challenges for public administration. On the one hand, public administration needs to be flexible to accommodate technological advancements. But, on the other hand, it is rightfully not so flexible because we are dealing with public money, which comes with a great level of responsibility.
The major challenge in driving AI adoption is public fear, especially regarding job security. With ILO (International Labour Organization), we are just conducting research on the impact of AI on jobs, in the context of the Polish economy. On top of it we are planning a deeper research on public attitudes towards AI. We started to call this the “AI barometer” — where we can track how perceptions of AI evolve. Media narratives often focus on extreme concerns, from AI taking over jobs to more dystopian scenarios, which fuels anxiety. Our goal is to communicate that AI is simply a tool–one that should work for us–and to highlight the possibilities for collaboration between humans and AI.
The second challenge is implementation. We have just completed our first sprint of developing the PLLuM model, a large language model (LLM) designed specifically for public administration. You can now find it on HuggingFace, as an open-source family of models. Given that GPT technology has been around for about two years, we are only now reaching the stage of integrating it responsibly into government services. Doing things properly takes time, but we are committed to delivering AI-driven services that are reliable and align with legal and ethical standards.
Q: How has your research allowed you to understand citizens’ attitudes about technological innovations?
In our department, we operate within three main areas: research, policy, and implementation. We see them as a complementary triangle–research informs policy, which in turn enables implementation. Understanding citizens and civil servants is key to creating effective services. We don’t want to implement solutions just because they are trendy; we want them to genuinely serve people’s needs.
In the research we are conducting, we observe a growing fear of job loss related to AI. The research with ILO and further communication connected with it, are trying to target these challenges. At the same time, public trust in AI seems to be diminishing rather than growing.
One major reason for this is the increasing popularity of generative AI. Many people use the terms “AI” and “generative AI” interchangeably, which creates confusion as for “hallucinations” — incorrect or misleading outputs — have strongly contributed to scepticism. While reliance on AI is increasing, so is awareness of its potential flaws, especially when not properly supervised. That’s why we see our development of PLLuM and simple yet informative communication as crucial projects in the AI sphere.
Q: How important is communication to AI innovation?
Public fear of AI is largely driven by the narratives people encounter — many of which are negative, influenced by sci-fi movies and sensationalist media. When innovation is poorly understood, it becomes intimidating.
Breaking concepts down into simple, relatable terms helps build trust. People need to see the direct benefits. For example, our digital wallet initiative in Poland resonated with citizens because it was framed in terms of convenience — having an ID in your phone is an easy-to-grasp concept. Once people see the value, they are more likely to adopt new technologies.
Q: What emerging technologies will have the most significant impact on public services in the next five years?
AI is definitely at the top of the list. Other technologies, such as quantum computing and connected devices (Iot), have a role to play as innovation progresses. However, AI is the most impactful for public administration because of the high level of bureaucracy involved. Since we spend public money and need to document every process, AI can help automate many tasks while still allowing for human oversight, which is crucial.
Many ministries are already implementing AI, particularly in financial services. One interesting use case involves generative AI for identifying potentially abusive clauses in contracts. The AI provides recommendations, but the final decision still rests with legal professionals or citizens. This aligns with our focus on trustworthy AI — it supports decision-making rather than replacing human judgment.
If there were a way to “clone” civil servants, that would be a game-changer for public administration! And AI can hopefully help us deliver on that, helping people deliver the final product with much of the process being automated and validated. There is a common perception that bureaucrats have ample time and resources, but digital departments are often understaffed, as this is still a relatively new field. We need to maximise the time these digital gems are driving strategies and narratives, instead of finalizing paperwork.
Q: Your career spans both the private sector and government. What are the biggest lessons you’ve brought from the private sector into your current role?
At first, adjusting to the rules and structures of the public sector was a challenge. But one key lesson I brought from the private sector is the importance of communication and transparency. In my previous company, whenever we completed a project, we made sure to communicate it effectively.
Public administration can benefit greatly from this approach. Transparency can be achieved not just through conferences but also through simple updates on social media or other platforms that engage citizens. We need to build trust, as only with trust in the government’s involvement and expertise, can our message be really effective.
Q: Can you share a specific project or initiative that has significantly impacted Poland’s digital landscape?
In the AI landscape, The PLLuM model, our public service LLM — we recently completed its first version, and we have just published it on Hugging Face. PLLuM is designed as a trustworthy LLM that can be used across public administration and even by the private sector. One of our main goals is to support smaller municipalities, which often struggle to hire AI experts. By providing them with a ready-made chatbot tool tailored to their needs, we hope to ease their administrative burden.
All tools we build for central administration — we want them to be publicly available, open source and as ready as possible to use. The goal of the chatbots integrated into digital government services is for them to be tailored for civil servant use cases — easy to use for both smaller and larger players. The Swedes were actually one of the first to build a country-specific LLM. But not only Europe (Spain, France) has worked on these projects, countries such as India are working on analogous initiatives due to their many languages and dialects. As we see the trend, more and more countries are developing their own sovereign AI models.
This article is reproduced from Apolitical.
READ MORE: