Can AI Write Better Reports Than Experts?
Can AI Write Better Reports Than Experts? A Breakthrough Publication from PARP and SWPS University Redefines the Future of Public Sector Analysis
The idea that a report created by artificial intelligence could outperform one prepared by experienced human experts has long seemed far-fetched. But a new groundbreaking publication from the Polish Agency for Enterprise Development (PARP) and SWPS University titled “Generative AI in Research: Practical Applications in Public Policy Evaluation” shows that this is not just a provocative idea — it’s now a measurable reality.
For the first time in Poland, researchers conducted a practical, experimental comparison between AI-generated reports and traditional expert-written analyses. The outcome? Even the most skeptical observers were surprised.
Subscribe to Startup Digest to stay ahead with the latest news, investments, and must-attend events.
A Game-Changing Experiment
In a controlled experiment conducted by PARP’s Evaluation Unit and the National Evaluation Unit, participants evaluated two versions of an analytical report — one written by experts and one created by generative AI. Importantly, evaluators were not informed about the authorship of each document. They assessed the reports based on clear content and editorial criteria.
The result:
AI-written reports received higher ratings across all categories, with scores between 4.3 and 4.4 out of 5.
Reports authored by human experts scored lower, between 3.8 and 4.1.
Reviewers highlighted the clarity, coherence, and completeness of the AI-generated work, a revelation that challenges traditional assumptions about the irreplaceability of expert craftsmanship in analytical writing.
Beyond the Experiment: A Practical Guide for Public Institutions
The publication goes far beyond just presenting the experimental results. It offers a comprehensive, practice-oriented guide for institutions looking to integrate generative AI into their research and evaluation processes.
Key takeaways include:
Use cases for AI across the research lifecycle — from defining knowledge needs, planning studies, and collecting data, to analyzing findings and communicating results.
A new model of collaboration, not replacement: AI is seen not as a substitute for experts but as a powerful partner that enhances human capabilities, speeds up workflows, and improves the accessibility and quality of research outputs.
A new “architecture of cooperation with generative AI” built around four choices: assigning AI roles, choosing platforms and technologies, designing interaction strategies, and curating information resources.
A New Era for Policy Evaluation and Public Sector Innovation
The publication argues that the strategic use of AI can:
Accelerate timelines for research projects,
Increase the accuracy and quality of evaluations,
Improve communication with diverse stakeholder groups through more tailored and accessible reporting,
And ultimately enhance evidence-based policymaking.
Concrete examples illustrate how AI tools have already been applied to both qualitative and quantitative analyses in public institutions, from automating document reviews to assisting in data coding and helping synthesize findings.
Polish AI on the Rise: Local Models for Local Needs
The report also spotlights how Poland is building its own AI ecosystem, including large language models like Bielik and PLLuM, designed specifically to handle Polish administrative language, legal contexts, and public sector challenges.
Unlike global AI models dominated by English, these local solutions are being developed to meet the specific needs of Polish public institutions — from automated document generation to enhanced legal analysis.
Challenges: Responsibility, Ethics, and Trust
However, the authors caution that deploying AI in public administration brings significant responsibilities:
Ensuring transparency of AI-generated content,
Maintaining human oversight at key decision points,
Establishing clear ethical standards and security protocols,
Developing new digital competencies among public employees to work effectively with AI tools.
They propose concrete frameworks such as internal auditing systems (“second eyes” mechanisms), explicit governance models, and trust-building strategies to accompany AI integration.
The overarching message is clear: Responsible AI adoption is not just a technological shift — it’s a cultural and strategic transformation.
The authors, including Jacek Pokorski (Deputy Director at PARP’s Department of Analysis and Strategy) and Dr. Karol Olejniczak (Head of the Center for Policy Design and Evaluation at SWPS University), invite all public sector professionals to experiment, learn, and innovate responsibly with generative AI.
As they emphasize, the future belongs to those who understand how to work with AI — developing co-intelligence strategies where human expertise and machine power amplify each other.
In a rapidly changing world where agility, evidence, and communication are more important than ever, embracing generative AI could become one of the greatest opportunities for the public sector — if implemented thoughtfully and ethically.
Download the full publication (in Polish) here: https://www.parp.gov.pl/component/publications/publication/publikacja-ai
