By: Jeevan Thiyagaraja
Sri Lanka's public service, a legacy of British colonial administration, faces persistent challenges in maintaining independence and integrity amidst evolving political landscapes. . While designed to operate on principles of meritocracy and impartiality, it has been profoundly shaped by the political dynamics of post-independence history.
By 2015, growing public dissatisfaction fuelled calls for more independent and corruption-free public services. A decade later, numerous indictments and court cases have implicated public officials, highlighting the ongoing struggle for accountability and integrity. The public service remains a challenging domain, characterised by island-wide transferability, provincial rotations, diverse cultural contexts, and varying demands.
This context raises fundamental questions about the essence of public service, the imperative for its independence, and the ideal balance between autonomy and accountability. In the World Bank's most recent Worldwide Governance Indicators, Sri Lanka's Government Effectiveness score for 2023 was 41.04%, placing it in roughly the lower two-fifths of nations worldwide. This stark figure illustrates the tangible gap between constitutional design and lived reality when institutional stability is compromised.
Why Independence Must Be a Constitutional Component, Not an Administrative Courtesy
Wanasinghe's account already answers, in effect, why reform bodies need independence: every implementing unit he traces — the Salaries and Cadres Committee, the Ministry's stripped-out policy division, the donor-funded RMU — failed not because its mandate was wrong but because it had no protected standing of its own. Each lived inside a ministry, a budget cycle, or a donor project that someone else controlled, and each was reshaped or starved the moment its host's priorities changed. The lesson generalises beyond personnel reform: in Sri Lanka, anything that depends for its survival on an ordinary parliamentary majority or a single minister's goodwill will eventually be unwound by exactly that majority or that minister, regardless of how sound its design was on paper.
This is why the comparison to the Constitutional Council, the Public Service Commission, the National Police Commission, and the Human Rights Commission matters. These bodies show that Sri Lanka already knows how to build independent institutions — the constitutional language and appointment machinery exist. What the country has not yet done is apply that same architecture to the body responsible for administrative reform itself, or entrench the independence already granted to the others so it cannot be reversed by the next government with the right numbers in Parliament, as happened repeatedly across the Seventeenth to Twentieth Amendments.
Treating administrative reform capacity as a constitutional component — rather than a unit inside the Ministry of Public Administration — would mean three things in practice. First, the appointment of its head and core team should follow a process similar to the Constitutional Council's nomination procedure, requiring cross-party concurrence rather than a single appointing authority, so the institution survives a change of government. Second, its core funding should be a charge on the Consolidated Fund or an equivalent ring-fenced vote, not an annual ministry allocation that can be quietly reduced, and not a donor project that expires with its funding cycle — the flaw that sank the UNDP-staffed RMU. Third, its existence and core functions should require a constitutional or, at minimum, a specially entrenched statutory basis, so that abolishing or hollowing it out demands the same supermajority that creating it required, not a simple amendment Act.
Independence of this kind is not an end in itself. It exists to protect a specific capability: the ability to say that a ministry's structure is irrational, that a cadre is bloated, or that a transfer was retaliatory, without that finding depending on whether the minister currently in charge wants to hear it. Wanasinghe's seven Secretaries in six years is the cautionary case — an institution whose leadership changes that often cannot accumulate the standing to make findings stick. An independent, long-tenured body can.
Accountable to Whom?
Independence without accountability is its own failure mode, and Wanasinghe was alert to this — his ARIA proposal paired independence from line-ministry control with a deliberately layered reporting structure: quarterly to a Presidential Cabinet sub-committee, half-yearly to Parliament, and continuous to the public through active dissemination. That three-tier design is worth taking seriously as a template, because each tier answers a different question.
Reporting to a Cabinet sub-committee answers the question of executive coordination — does the reform body's work align with government priorities, and can the executive flag operational problems quickly, without that channel becoming a vehicle for the executive to redirect or quietly bury findings it dislikes. Reporting to Parliament answers the question of democratic legitimacy — can elected representatives, including the opposition, examine what the body has found and what it has done about it, in public session, on the record. Reporting to the public answers the question that the first two cannot: can a citizen, a civil society organisation, or a journalist independently verify that the first two channels are functioning, or whether a quarterly Cabinet report and a half-yearly parliamentary report have, in practice, become a formality nobody reads.
This third tier deserves more weight than Wanasinghe's report could give it in 1994, before routine digital publication was possible. A reform or implementation body accountable chiefly to two closed institutional audiences is accountable in name only if those audiences have no external pressure to act on what they receive. Publishing findings, cadre data, project timelines, and outcome metrics on a standing public portal — not as occasional press releases but as a continuously updated record — converts accountability from a private exchange between the executive and the body into something the public can check. The HRCSL's experience is instructive here: its formal reporting lines exist, but its referrals to the Attorney General have not reliably produced visible follow-up, partly because there has been no statutory deadline forcing a response and no public dashboard making the absence of a response conspicuous. The fix proposed for that case — a statutory duty on the Attorney General to prosecute, decline in writing, or refer onward, within a fixed deadline — is really a generalisable accountability mechanism: pair every independent finding with a forcing function and a public record of whether the forcing function was honoured.
So "accountable to whom" has no single correct answer — it is all three at once, and weakening any one of them reproduces the conditions under which Wanasinghe's earlier reform bodies quietly died, unaccountable to anyone because no one outside government was watching, and unaccountable inside government because oversight had no teeth.
Technology as the Implementing Layer: A Worked Example
Wanasinghe's report is fundamentally about coordination failure — the same problem recurring whether the task was cadre rationalisation across ministries or, in the later devolution discussion, reconciling national policy with provincial authority. Multi-agency coordination problems of exactly this kind are still Sri Lanka's default condition, and they are visible in concrete, present-day tasks like greening urban open spaces or improving the welfare of poor households. Both cases illustrate why technology — used carefully — is less a separate initiative than a tool for solving the institutional-custody problem Wanasinghe diagnosed.
Take a greening example. Responsibility if given of all open space in an urban area would typically split across the Urban Development Authority, the Forest Department, the Irrigation Department, central and provincial planning authorities, water resource board and the relevant local government body — five or more institutions, each with a legitimate claim on part of the decision, none with full authority over the outcome. This is structurally identical to the fragmented-custody problem Wanasinghe documented for administrative reform itself: when responsibility is split this many ways, the default outcome is not balanced coordination but inaction, because no single agency bears the cost of the space staying unused and degraded, and any agency that tried to act unilaterally would be acting outside its mandate.
A shared digital register — a single geospatial layer where every open space in a local authority's jurisdiction is mapped once, with its land-use status, governing authority or authorities, current condition, and any ongoing or planned intervention visible to all five institutions and to the public — does not by itself solve the coordination problem, but it removes the most common excuse for it: that no one had visibility into what the others were doing or planning. Layering simple automated tracking on top of that register — flagging spaces with no recorded activity for a defined period, or surfacing spaces where two agencies have logged conflicting plans — turns a static map into something closer to an early-warning system, prompting a human decision rather than making one. This is a modest, realistic role for AI in this kind of work: pattern-spotting and prioritisation across data that already exists but is currently held in five incompatible filing systems, not autonomous decision-making about which trees to plant where.
The same logic extends more directly to household-level welfare work, where the stakes are higher and the margin for error is smaller. A family's quality of life depends on inputs from Samurdhi/Aswesumna welfare transfers, Social Services Department casework, district health services, school enrolment and attendance systems, Divisional Secretaries and often local government infrastructure — water, sanitation, roads — each tracked, if tracked at all, in a separate system that does not talk to the others. A family flagged as vulnerable in one system is frequently invisible in the others, not because any single agency is failing at its own task, but because no one holds the composite picture. Here the system's job is narrower than in the greening case: identifying, from existing administrative records, which households show multiple compounding risk indicators — for instance, a household combining welfare-eligibility flags with a health-system record of malnutrition risk and a school record of irregular attendance — and routing that composite case to a named caseworker, rather than leaving each agency to notice its own slice of the problem independently and usually not in time.
Two design principles should govern this kind of system, precisely because the underlying task touches vulnerable people and because Sri Lanka's own recent history with both Samurdhi and Aswesumna targeting has shown how badly automated eligibility decisions can misfire when there is no effective appeal channel. First, AI-supported systems of this kind should inform and prioritise human casework, not make eligibility or enforcement decisions unsupervised — the technology's job is to make the existing bureaucracy notice problems faster and more consistently, not to replace the judgment of a social services officer or a Grama Niladhari who knows the household. Second, every automated flag needs a visible, appealable trail: a family or a citizen affected by a system-generated decision must be able to see why it was made and contest it through a human channel, echoing the same logic as the reversed burden-of-proof and mandatory show-cause protections proposed elsewhere for officers facing retaliatory action — due process is not a luxury that gets dropped once a decision is automated, it is exactly the point at which due process matters most.
Equally important, and consistent with the wider reform agenda Wanasinghe pointed to but the 1986–87 Committee never directly addressed, this kind of coordination cannot be purely technocratic. A geospatial registry of open spaces or a household risk-flagging system designed entirely by central planners, with no community input, will reproduce the same blind spots as the centrally-administered economy Wanasinghe argued Sri Lanka's bureaucracy had been built for and had not yet escaped. A mixed model — government systems providing the data infrastructure and the AI-supported triage, community structures (Grama Niladhari divisions, school development societies, community-based organisations, local civil society) providing ground-level verification, contesting flags that don't match lived reality, and proposing interventions the data alone would miss — is both more accurate and more legitimate than either a purely top-down or purely informal approach. It also gives the technology a check against the single greatest risk of administrative data systems: that they become confidently wrong at scale because no one close to the ground was asked to verify them.
Continuous Improvement as an Institutional Property, Not an Event
The recurring failure across every case Wanasinghe documents was that reform was conceived as a finite project with an end date, rather than a standing institutional function. The ARC reported once, comprehensively, in 1987, and the assumption embedded in that design was that implementation would follow and then the work would be largely done. The Organizational Review was a one-off World Bank-funded exercise that produced a report and then, by Wanasinghe's account, simply stopped — "the matter rests there." The RMU's mandate was redirected twice and then terminated rather than renewed, because it had been built as a project with a lifespan, not a permanent function.
ARIA's design, and by extension any reform or coordination body built on the same principles, has to resist this pattern explicitly. A minimum five-year contract for the Co-ordinator, core government funding rather than donor project financing, and standing — not ad hoc — Teams on Personnel Policy, Administrative Systems, Organisation Audit, and Training were all Wanasinghe's attempts to build permanence into the institution's structure rather than leaving it to depend on continued political enthusiasm. The same applies to the technology layer described above: a geospatial open-space registry or a household-risk-flagging system is not a one-time IT procurement that gets switched on and then left alone. It needs the same standing custody — a team responsible for keeping the underlying data current, auditing the AI-generated flags for accuracy and bias on a fixed schedule, incorporating community feedback into how the system is calibrated, and reporting publicly on what the system has and has not caught.
Continuous improvement, in other words, is not a separate fourth item alongside independence, accountability, and technology — it is the property that determines whether the other three survive contact with Sri Lanka's actual political cycle. An independent body with no mechanism for revisiting and updating its own findings becomes a museum of its founding report. An accountability structure that reports the same metrics the same way every year without ever asking whether those are still the right metrics becomes a ritual rather than a check. Concretely, this could mean an annual public audit — distinct from the quarterly and half-yearly reports already proposed — that asks a narrower question: did the system catch what it was built to catch, did any flagged case get successfully appealed and overturned, and what changed in response. Skipping that question is how a calibration choice made in year one quietly becomes wrong by year five, exactly as Sri Lanka's cadre numbers drifted upward by double digits in the years immediately following a reform exercise meant to shrink them. The institutional answer Wanasinghe proposed thirty years ago — a permanently resourced, long-tenured, publicly accountable body whose mandate is reform itself rather than any single reform — remains the right shape for the answer today, whether the task at hand is rationalising ministries, protecting whistleblowing officers, greening a city, or making sure a poor family is seen by more than one part of the state at a time.
