Policy failure in the public sector is often attributed to corruption, weak political institutions, or insufficient resources. These are real problems, but a large proportion of the breakdown in practice traces back to something more specific and more fixable, which is the absence of usable information at the point where it is needed. The everyday operations of government agencies are shaped by information gaps that are structural rather than incidental, and that create the conditions for many of the more visible dysfunctions.
Government agencies consistently operate without the information they need to function well. New leadership arrives at a ministry without access to records of what has been tried and what has worked, so they spend months developing a policy only to discover by chance that a nearly identical initiative had been tried and abandoned a few years earlier, with the lessons from that failure never recorded. Staff develop policies without reliable data on baseline conditions, sometimes proposing entirely new legislation when a simple amendment to an existing law would achieve the same goal, because no one did the basic work of checking what legislation already existed. Departments within the same ministry have no idea what their colleagues down the hall are doing, which means one department conducts site visits to islands without knowing that another department needs information from those same islands, so two trips happen when one could have served both purposes, or one department develops a data collection system without knowing that another department already built something similar.
This lack of visibility across departments also means that staff fielding public enquiries or client requests about services or activities handled by a different part of the organization have no easy way to give an answer, because there is no shared reference for what the organization as a whole does. A frontline worker answering a call about available social services on a given island cannot quickly tell the caller what services exist there, which agencies provide them, and how to reach them, because that information has never been compiled in one place.
Every new policy report, every consultant assessment, every strategic plan begins with the same background research done over again. Consultants search for previous reports and cannot find them, or find some but miss others. They piece together the history and evolution of policies from scattered sources, whatever documents happen to be accessible, and whatever institutional memory still resides in the heads of staff who were around at the time. Reports are published, circulated briefly, and then effectively disappear, existing somewhere on a hard drive or in a filing cabinet but not indexed, not searchable, not accessible to the next person who needs them. Academic research on Maldivian social policy exists but is scattered across journals and repositories that practitioners rarely have time to search. International assessments and guidance documents from UN agencies and development partners are available online but not compiled in any sector-specific way. The collective knowledge that should be accumulating over time instead dissipates, and each new effort starts closer to zero than it should.
Skilled employees leave and their knowledge leaves with them, because critical procedures and institutional knowledge were never documented. Much of what makes an experienced staff member effective is informal knowledge, things like how things actually work as opposed to how they are supposed to work, why certain decisions were made, which workarounds are necessary, who needs to be consulted. When that person leaves, their replacement has to figure all of this out from scratch, making mistakes that the experienced person would have avoided, and the answers to questions that were asked and answered years ago have to be reconstructed from memory each time. Programs that were working well get discontinued because new leadership does not know they exist, and programs that failed get repeated because no one remembers why they were abandoned.
All of this is compounded by a lack of technical capacity. Staff often lack proficiency in spreadsheet software or data management, which means a task that should take twenty minutes takes two hours, and overworked staff will inevitably abandon time-consuming new tasks in favour of their more immediate, familiar responsibilities. This creates a culture of reactive firefighting where long-term policy work is constantly pushed aside by the tyranny of the urgent, and organizations most in need of improvement are the ones with the least bandwidth to pursue it because their staff are consumed by managing the very dysfunctions that need to be fixed.
We counted over forty different, incompatible software systems being used across various government agencies in the Maldives, each developed in isolation to solve a specific problem, creating an environment where data sharing and coordination between agencies was nearly impossible. This is not a problem that can be solved by hiring more people or increasing budgets. The Maldivian public institution we worked with tried that approach, ballooning its workforce without addressing the underlying inefficiencies, and the result was staff who were simultaneously overworked and underutilized. These are all outcomes of organizations that have no infrastructure for holding onto what they know.
Three interventions with a shared logic
The briefs that follow describe three specific interventions that Public Policy Lab has been developing with the Ministry of Social and Family Development. Each addresses a different dimension of the information problem, but the underlying logic across all three is the same: building infrastructure that makes knowledge accumulate over time rather than dissipating with turnovers, political transitions, and gaps in individual technical skill.
The first brief describes a searchable knowledge repository for the social sector. This is a centralized digital archive of all relevant documentation, from government publications and internal ministry reports to international assessments, academic literature, and news coverage. The problem it addresses is that every new piece of work currently starts from scratch because there is no central place where previous work has been compiled. Reports get produced, circulated briefly, and then effectively disappear. Around 200 documents have been compiled and labelled so far, with a searchable portal built to make them accessible, and the next phases involve incorporating the Ministry's own internal reports and research, adding detailed tagging by department and policy area, working with universities to receive any Maldivian student theses on these topics, and eventually adding scanned copies of older paper reports where departments have physical print copies on hand. The repository also includes a searchable archive of news coverage with sector-related keywords, which provides a layer of information that formal reports often omit and which is valuable for anyone trying to build a timeline of developments in a specific area or for onboarding new staff or consultants who need to quickly understand the background of an issue.
The second brief describes the documentation work needed to preserve institutional memory across staff turnovers and political transitions. Where the repository collects external and published knowledge, this effort focuses on the ministry's own operational knowledge: what programs are running across every department, what their status is, what challenges they face, what the reasoning was behind key decisions. Government agencies lose institutional memory in two ways, through ordinary staff turnover where people leave, retire, or transfer and their knowledge of how things work leaves with them, and through political turnover where every change in government brings new leadership that was not present for the previous administration's work and has limited visibility into what was done, what worked, and what failed. The output of this work is a comprehensive reference document that serves as an onboarding resource for new staff, gives everyone visibility into what colleagues in other departments are working on so that coordination and avoiding redundant work becomes possible, and provides a genuine handover document for leadership transitions so that incoming administrations can make informed decisions about what to continue, adjust, or discontinue rather than starting without knowledge of what came before.
The third brief describes a standardized prompting tool for AI-assisted policy work. AI tools can dramatically accelerate drafting, data conversion, and analytical work, but the quality of their output depends heavily on the quality of the instructions they receive. A senior analyst who understands research methodology and document formats can get excellent results from these tools, while a junior staffer using the same tool may produce output barely better than starting from scratch, not because the tool is incapable but because they do not know what to ask for. The prompting tool provides a library of form-based templates covering everything from cabinet papers to UN project documents to GIS workflows, where users fill in the specifics of their task and the tool generates a detailed prompt that embeds best practices, methodological frameworks, and formatting requirements. This means that the quality of the output depends on the template rather than on individual prompting skill, and institutional knowledge about how documents should be structured becomes accessible to all staff rather than residing only in the heads of a few experienced people.
How these connect
We have looked at organizations where data sits siloed across departments and servers in formats that other parts of the organization cannot access or do not know exist. Where a member of staff fielding a public enquiry or a client request about services or activities handled by a different department has no easy way to give an answer, because there is no shared reference for what the organization as a whole does. Where multiple departments carry out overlapping data collection efforts or site visits to the same locations without knowing about each other, because there is no visibility into what other units are doing. Where every new publication or assessment repeats background research and findings that were already produced in an earlier report that no one can find, so the same ground gets covered again and again while actual analytical work gets less of the time and resources it needs. Where accumulated knowledge and evidence and new research findings that should be informing decisions are not easily known or seen or read by the people they need to reach, because there is no system that puts relevant research in front of practitioners. Where knowledge built up over years by experienced staff, including informal knowledge about how things actually work, why certain decisions were made, what was tried and what happened, disappears with every turnover or leadership change because none of it was ever written down in a way that survives the departure of the person who held it.
These are all expressions of the same underlying problem, which is the absence of any information infrastructure within the organization. The repository addresses the part of this where knowledge exists in documented form somewhere but is unfindable, scattered across hard drives and servers and filing cabinets and websites with no central place to search. The documentation work addresses the part where critical operational knowledge was never recorded in the first place, and exists only in the heads of staff who will eventually leave. The prompting tool addresses the gap in technical capacity and familiarity with specific methodologies and document formats that prevents staff from being able to produce and use the knowledge products that would make all of this work, so that output quality depends on the template rather than on which staff member happens to be writing that day.
Each of these addresses a gap that, left unfilled, undermines the value of the others. A repository is only useful if people can produce quality work to put into it and if there is operational documentation feeding into it, and documentation is only useful if it is accessible and searchable and actually maintained, and capacity tools are only useful if there is an institutional knowledge base for the outputs to draw on and contribute to. A significant share of implementation failure is caused not by intractable political problems but by information problems that persist mainly because the infrastructure to solve them has never been built. These briefs describe what building that infrastructure looks like in practice.