What Two Years of ModParl Have Taught Me About AI in Parliaments
From isolated experiments to institutional strategy, ten lessons on legislative modernization
BY BEATRIZ REY
This may sound a little crazy as I write it, but ModParl has now been around for more than two years.
That realization made me stop and ask a simple question: what have we actually learned about legislative modernization — and, in particular, about implementing AI in parliaments?
Looking back, the answer lies in the journey itself.
When we published our first post in April 2024, our understanding of modernization was still relatively narrow. Much of our attention was on the US House Select Committee on the Modernization of Congress, which I followed closely during my time on Capitol Hill as a legislative assistant. We also highlighted interesting initiatives in places like Kenya and Chile that were experimenting with new ways of receiving citizen input. But one thing was notably absent from those conversations: AI.
Our first piece dedicated to AI did not appear until February 2025. It focused on Brazil’s Senate, which at the time was using AI primarily to improve legislative transparency and has since expanded its use into several other areas. From there, I started researching other examples around the world. For more than a year, though, these remained mostly isolated cases. We saw interesting experiments emerge in different legislatures, but not yet a broader conversation about what AI might mean for parliamentary institutions.
Then something shifted.
In March of this year, I attended the Inter-Parliamentary Union (IPU) Assembly in Istanbul. For the first time, I heard parliamentary officials talking not about individual AI tools or pilot projects, but about developing an institutional vision for AI. The conversation had moved beyond experimentation and toward institutional strategy. Legislatures were beginning to ask a different set of questions. What role should AI play in parliamentary work? How should it be governed? What capacities need to be built? And how can institutions adopt AI strategically rather than piecemeal?
But what do parliaments need to begin that process? Here is what I have learned so far.
Foundations
Digitization comes first.
The uneven pace of parliamentary modernization means many legislatures still lack the foundations necessary for meaningful AI adoption. Before parliaments can experiment with AI, they need to digitize and organize their information. After all, AI is only as useful as the information it can access and process.
An informational approach is key.
AI implementation requires parliaments to see themselves not simply as political arenas, but also as information-processing institutions. Legislatures produce, receive, organize, and disseminate enormous amounts of information every day. Yet many struggle to manage that information effectively — making it difficult to retrieve knowledge, connect data across processes, and turn information into actionable insights. Viewing parliaments through this lens makes it easier to identify where AI can add value. In many cases, successful AI implementation depends less on sophisticated algorithms than on improving how parliamentary information and knowledge are structured, managed, and used across the institution.
AI literacy matters.
AI is moving incredibly fast. New tools emerge constantly, and assumptions that seemed reasonable a few months ago can quickly become outdated. For parliaments, this means investing in AI literacy across the institution. Members, staff, and ICT teams need a shared understanding of what AI is, how it can support parliamentary work, and where its risks and limitations lie.
Institutional Capacities
Leadership matters.
I have heard this again and again in different contexts, but one of the clearest explanations came in our latest ModParl piece: a conversation with Annelie Lotriet, Deputy Speaker of South Africa’s National Assembly. As she put it, institutional commitment is essential to transforming parliaments. And that commitment is much easier to build when leaders do more than simply approve change from above. The real difference comes when they act as policy entrepreneurs — championing new ideas, building coalitions around them, and creating the conditions for institutional change.
Parliament’s ICT department needs to work closely with both staff and Members.
I saw how often these groups operate in separate worlds at another IPU event: Artificial Intelligence Conference: The Role of Parliament in Shaping the Future of Responsible AI, this time in Malaysia. But I also saw something encouraging there: people from these different areas sitting together and beginning to recognize the importance of bridging those gaps. That matters because AI cannot be treated as an isolated project to “fix communication with constituents” or automate a single workflow. If parliaments are going to use AI well, they need to approach it institutionally: as an opportunity to rethink how the legislature manages information, supports deliberation, and serves the public.
Governance matters.
Implementing AI is not simply a technical exercise. It requires institutions to establish rules, responsibilities, and safeguards. Questions of accountability, privacy, transparency, and oversight need to be addressed from the beginning. I’m now researching how parliaments are adopting regulations for internal usage of AI – I hope to write about this soon.
Processes of change
Experimentation and learning matter.
Most legislatures do not leap directly into institution-wide AI strategies. They begin with small experiments, learn from early initiatives, and gradually build confidence and capacity. In this sense, pilot projects are not the end goal; they are part of an institutional learning process. I came to appreciate this in conversations with Martin Kamprath of the German Bundestag’s Scientific Services, whose insights unfortunately never made it into a published ModParl piece.
AI is encouraging institutional thinking.
Early parliamentary AI initiatives often focused on individual tools and isolated use cases. As I mentioned, increasingly, legislatures are asking broader questions about institutional strategy, governance, and capacity building. AI is prompting parliaments to think more systematically about how information flows across the entire institution.
Modernization is ultimately about people.
Technology can enable change, but transformation depends on leadership, incentives, collaboration, and institutional learning. In the end, modernization succeeds only when people throughout the institution decide to work differently and have the capacity to do so.
Sharing lessons matters.
Regardless of their level of modernization or AI adoption, parliaments have much to learn from one another. Over the past two years, I have seen beautiful examples of legislatures openly sharing experiences, successes, and even setbacks with their peers. Those moments of exchange have convinced me that parliamentary modernization is, in many ways, a collective endeavor. They also continue to inspire me to write ModParl — and, hopefully, to keep writing it for years to come.
Modern Parliament (“ModParl”) is a newsletter from POPVOX Foundation that provides insights into the evolution of legislative institutions worldwide. Learn more and subscribe at modparl.substack.com.
