Overview
The encounter and the institution
A service is experienced as a chain of requests, explanations, payments, judgements and corrections. Its success also depends on matters the applicant cannot see: agency mandates, budgets, staff capability, data sharing, contracts, risk oversight and the limits placed on new technology. This edition brings both sides into view.
The ten selections include two reports by the Organisation for Economic Co-operation and Development (OECD), African policy briefs, peer-reviewed studies, institutional arguments and a legislative development. They appeared in September 2026 and establish different things. None measures the experience of Kenya's counties directly. Application here requires judgement about our own institutions and conditions.
Can a person reach the service, and can the state see and govern the institutions on which that journey depends?
At the point of service
Several agencies, one person's life
Organisation for Economic Co-operation and Development, Serving Citizens 2026: Delivering Better Public Administrative Services.
The report examines how OECD and partner countries organise services around common life events. In 27 of 31 responding OECD countries an institution leads service improvement, yet 15 report shared responsibility across agencies and 22 have a coordinating committee. A government-wide strategy exists in 26 of 30 countries with data, while services are joined through data sharing across agencies for only one fifth of the life events examined. Seventeen of 31 use experience and performance data to decide improvements.
Evidence boundary. These are reported institutional arrangements and selected service journeys, not a causal test of which design works best. They establish no corresponding rate for Kenya. A named lead, a shared website and a completed transaction are different achievements.
For eCitizen, Huduma Kenya or a county service, name the owner of the whole journey and map the agencies, records and payment systems behind it. Record repeat visits, abandoned cases and whether feedback can change staffing, budgets or procedures.
Shared standards and infrastructure must still permit adaptation to language, connectivity, travel and local administrative conditions. Subnational institutions need a route to change the service, not merely receive transactions.
Coordination is a continuing governing function. Judge it by the person's completed task and the ability to correct an error held by another institution.
The tax is only part of what a small trader pays
International Growth Centre and Mozambican collaborators, Reducing the cost of ISPC compliance: The case for mobile wallet payments in Maputo Province.
The study looks at the practical demands of paying Mozambique's Imposto Simplificado para Pequenos Contribuintes (ISPC), a simplified tax for small contributors. Among surveyed taxpayers who experienced payment difficulty, service unavailability was reported by 27 per cent, lack of money by 22 per cent and long queues by 19 per cent. The researchers identify travel, waiting and time away from business alongside the monetary tax. They examine mobile wallets as an additional channel, with fraud concerns, official confirmation and user support shaping likely uptake.
Evidence boundary. These percentages describe barriers among respondents who experienced difficulty, not all Mozambican taxpayers. The brief supports testing simpler payment arrangements; it does not demonstrate that introducing wallets has already raised compliance or revenue.
For a county permit or small-business payment, measure what compliance costs in time, transport and failed transactions. Make the receipt and correction route as clear as the payment demand.
An additional channel can ease a burden only when the existing route remains usable and assistance is available to people with limited records or connectivity.
Assess compliance policy from the payer's side of the transaction, including the administrative costs that never appear in a tax schedule.
Public care begins before the clinic door
International Growth Centre and Mozambique's National Directorate of Traditional and Alternative Medicine, Traditional medicine practitioners as a bridge to conventional healthcare in Mozambique.
In the surveyed rural area of Inhambane, 61 per cent of households had consulted a traditional medicine practitioner. Provider choice varied with beliefs about the cause of illness. The researchers found that practitioners recognised around one-third of the typical symptoms of human immunodeficiency virus (HIV), malaria and tuberculosis, and have begun a randomised trial of training for recognition and referral.
Evidence boundary. This is baseline evidence about where people seek help. The referral intervention began testing in September 2026; its effectiveness and safety have not been established by this brief. Clinical assessment and treatment remain matters for qualified health services.
In mobile and remote communities, first identify whom households actually approach. Any referral arrangement would require locally grounded design, clear clinical limits, supervision and a record that people reached appropriate care.
Formal services can improve access by understanding trusted routes into care while testing, rather than assuming, whether a new bridge changes outcomes.
The first contact in a person's service journey may sit outside the institution that is ultimately responsible for the service.
The conditions under which judgement is exercised
R. M. Rifat Mahmud, Why do field bureaucrats stay satisfied in Bangladesh?, Frontiers in Political Science.
A survey of Bangladesh Administrative Service officers examines job satisfaction in field postings. Collegial relationships, training, promotion prospects and fair performance evaluation are positively associated with satisfaction in the study's models. The author also reports an association between authoritarian pressure and lower satisfaction, and finds that participation in innovation is associated with lower satisfaction in this setting.
Evidence boundary. The cross-sectional design identifies associations, not causal effects on job satisfaction or service outcomes. The abstract and sampling section report different participant counts; this edition therefore does not quote a sample size. Neither the personnel system nor its estimated relationships can be presumed to match Kenya's.
Ask what training, supervision, workload and escalation arrangements a county officer has before making a consequential frontline decision.
A mandate to innovate may create extra exposure for officials unless managers also provide professional support, fairness and room to exercise judgement.
Service reform should observe the experience of those asked to carry it out alongside user satisfaction.
One roof does not complete a service
South Africa's Department of Public Service and Administration, account of the 2026 Thusong Service Centre Week.
The department identified practical gaps in the functionality, governance, infrastructure, connectivity, monitoring and maintenance of its one-stop service centres. It framed those issues through the experience of citizens seeking basic services and the work of frontline public servants. The account provides a useful set of operational questions for a model intended to bring several services together.
Evidence boundary. This is the department's account of its programme and challenges. It is not an independent evaluation of how many people completed a service, how quickly they did so, or whether the centres reduced exclusion.
At a county or shared service desk, count completed transactions and unresolved referrals alongside visits. A person should know which institution owns the next step.
Physical proximity helps when staff have the authority, records and connections to resolve matters, with a route for cases that cross institutional boundaries.
Co-location should be evaluated as a working service, with attention to uptime, maintenance and the end of the user's journey.
The machinery that must answer
A queue, a failed payment or an unanswered request is the visible face of choices about workforce, ownership, information, contracts and oversight. The next readings examine the capacity behind the encounter and the public limits on new tools.
The capacity the state must keep
Donald F. Kettl, The Urgent Race: Learning from the Challenges of the Public Service—Toward a Pathway for Reform, IBM Center for The Business of Government.
Kettl identifies five roadblocks in the United States federal civil service: contested meanings of accountability, the loss of mission as an organising purpose, merit obscured by process, uncertain boundaries between employees and contractors, and weak performance management. He distinguishes changes agencies can make under existing authority from reforms requiring legislation. The argument places workforce rules at the centre of the state's ability to carry out policy.
Evidence boundary. This is a United States-focused reform argument based on institutional history and administrative experience. It does not test a workforce model for Kenya or prove that a particular personnel change improves services.
National and county governments need people able to specify, procure, supervise and evaluate complex services and technology. Specialist career paths and fair performance systems matter alongside the ability to manage contractors. Public judgement cannot be purchased with the software.
Where expertise moves among government, contractors and partners, identify which capabilities the state must retain, which may be shared across institutions and which can be commissioned without losing institutional memory.
A service statistic may reveal a failure whose cause lies earlier in staffing, incentives or unclear responsibility. Mission needs to be visible in the rules under which officials work.
When risks collect beyond one board
Organisation for Economic Co-operation and Development, Managing Risk Across State-Owned Enterprises.
The OECD asks state owners to examine risks that accumulate across their entire portfolio of enterprises. Individual audits can miss common lenders, suppliers, guarantees, climate exposure and service dependencies. In the data cited by the report, 81 per cent of respondents had rules or expectations for enterprise risk systems, but only 41 per cent had a state-level framework to monitor risk across the portfolio. The report proposes shared language, comparable reporting and a link between risk information and ownership decisions.
Evidence boundary. The publication draws on surveys, interviews, standards and selected cases. The percentages describe reported arrangements in participating jurisdictions; they are not estimates of the risk in Kenya's enterprises or proof that one portfolio model prevents failure.
Treasury and county ownership institutions could examine guarantees, arrears, procurement integrity, climate exposure and shared service dependencies together. The account should distinguish weak performance from a public obligation deliberately assigned to an enterprise.
Portfolio information can clarify when subsidies, restructuring or recapitalisation meet an explicit public purpose, and where several entities rely on the same fragile financing or supplier chain.
The state has to act on joined information. Consolidated reporting matters when it alters capital allocation, supervision or the terms on which an enterprise carries a public task.
Who can require the institution to answer?
The Presidency of South Africa, announcement of presidential assent to the Public Service Commission Act, 2026.
According to the Presidency, the Act strengthens the Commission's independence and extends its mandate to local government and public entities. It provides for directions requiring an executive authority or another person affected by a Commission decision to report on implementation. The development shifts attention from receiving a complaint to whether an oversight decision is followed through.
Evidence boundary. Presidential assent establishes the legislative development. The announcement does not show that implementation has begun, that powers have been exercised, or that services have improved.
Examine how county complaints, administrative justice, public service boards and assemblies hand a matter onward. For each route, identify who can demand a response and verify correction.
An oversight body's reach across levels of government is useful only if its findings lead to dated responses and a visible route for the person affected.
Count implemented remedies, not only complaints received and decisions issued.
Ask what the system is for before buying it
Carnegie UK, Governing AI for Collective Wellbeing.
Carnegie UK proposes that high-impact artificial intelligence be judged against social, economic, environmental and democratic wellbeing. It would place a public-value test in business cases, procurement guidance, assurance and public registers, asking who benefits, who bears the risk, how decisions can be reviewed and which institution remains answerable. The report begins with instruments governments already possess rather than waiting for one comprehensive new statute.
Evidence boundary. This is a policy framework for the United Kingdom. It does not measure whether its proposed tests improve outcomes, and the instruments would have to be adapted to Kenya's constitutional and devolved arrangements.
Before procuring an AI system for licensing, health administration or service triage, state the public purpose, data limits, fiscal and environmental costs, human review and route for redress. A county needs authority to pause a deployment whose effects differ from those assumed in a national business case.
Budget guidance, procurement templates, sector oversight and registers can establish immediate duties. Those duties have substance only where officials can obtain evidence, challenge a supplier and respond to harm.
The public test extends beyond technical performance: what kind of service and distribution of power will this system produce, and which institution can change course?
When trust travels further than permission
Raveena Marasinghe and colleagues, Governing AI innovation in local government: how trust and risk perceptions shape public support, European Journal of Innovation Management.
The authors surveyed 1,183 people across Australia, Hong Kong and Saudi Arabia. Experience with local-government AI was associated with perceived understanding and trust; trust, in turn, was linked to perceived benefits and support for visible urban applications. Favourable views of routine uses were also associated with greater openness to surveillance uses. Perceived risk remained important to demand for responsible governance.
Evidence boundary. The data are cross-sectional and self-reported, with different sampling approaches across the three settings. These are associations, not evidence that exposure causes trust or consent. The samples cannot stand in for opinion in Kenya's counties.
A useful traffic or licensing application earns no automatic licence for surveillance or wider data reuse. Any change of purpose calls for fresh public justification, notice, privacy safeguards, human oversight and a route to challenge.
Test whose experience is missing from consultation, particularly where language, connectivity and prior treatment by authorities shape confidence. Imported opinion data cannot replace local inquiry.
Public trust is evidence to be understood, not permission that can be carried from one use to another. Institutions must keep the boundary between a service and a more intrusive power visible.
Synthesis
The state must be able to follow its own promise
These ten readings bring different kinds of evidence to one institutional question. A person bears the time, money and uncertainty of obtaining a service. Behind that encounter sit decisions about who owns the process, what an official may decide, which information can be joined, which risks the state can see and how an error is corrected. A reform can be sound in design and still fail at any one of these connections.
For a Kenyan county, begin with one consequential service. Follow it from first enquiry to completion or appeal, including people who give up. Set beside that journey an institutional map: the legal owner, budget holder, responsible staff, data custodian, contractor, escalation point and authority able to correct a decision. Record each visit, cost, transfer and reason given. Then ask whether the resulting evidence reaches someone who can change a rule, a staffing arrangement or a budget.
The same discipline applies at another scale to risks shared across public enterprises. It applies before an AI system is bought, when officials must be able to test the public purpose, challenge a supplier and stop an unsafe extension of use. Trust in a useful service may open a conversation with the public; it cannot settle a different question about surveillance or redress.
Public authority earns credibility when it can follow the person's journey, account for its own capacity and risks, and correct what its systems have done.
Reflections
- Who owns the whole service?Follow a person across offices and channels. Name the institution responsible for completion, for reasons when it fails, and for carrying a case across a boundary.
- What can the state still judge for itself?Test the skills retained to commission work, question a supplier, see risks across public entities and change a failing process.
- Where does permission end and remedy begin?Separate acceptance of a useful service from consent to a new use of data or power. Give people reasons, a route to challenge and evidence that correction followed.
References
The selected sources
- OECD, Serving Citizens 2026: Delivering Better Public Administrative Services, 24 September 2026.
- Manhique and colleagues, International Growth Centre, Reducing the cost of ISPC compliance, 15 September 2026.
- Budillon, Gray, Lowes and Cumbi, International Growth Centre, Traditional medicine practitioners as a bridge to conventional healthcare in Mozambique, 22 September 2026.
- R. M. Rifat Mahmud, Why do field bureaucrats stay satisfied in Bangladesh?, Frontiers in Political Science, 18 September 2026.
- South African Department of Public Service and Administration, account of Thusong Service Centre Week, 17 September 2026.
- Donald F. Kettl, IBM Center for The Business of Government, The Urgent Race: Learning from the Challenges of the Public Service—Toward a Pathway for Reform, September 2026.
- OECD, Managing Risk Across State-Owned Enterprises, 15 September 2026.
- South African Presidency, announcement of assent to the Public Service Commission Act, 16 September 2026.
- Carnegie UK, Governing AI for Collective Wellbeing, 22 September 2026.
- Marasinghe and colleagues, Governing AI innovation in local government: how trust and risk perceptions shape public support, European Journal of Innovation Management, 11 September 2026.
The ten sources are interpreted for Kenya, the Global South and the wider world without treating their findings as directly transferable. Dates, source types and evidence limits accompany each reading. Selections from 11–17 September remain within the preceding fortnight and add practical weight to the issue.