A public service moves its application form online. Completion becomes faster for many people, yet someone with a screen reader cannot reach the final button, and another person cannot tell how a decision was made. The institution may report that the service is digital. Whether it is human-centric digitalisation is a different question. The term describes a way of shaping digital change around people’s rights, capabilities, needs and meaningful opportunities to influence outcomes. It concerns governance and practice across the whole service, not only the appearance of an interface.
Human-centric is a demanding label. It asks whose task becomes easier, whose burden moves elsewhere, who can contest an error, and whether staff have the authority and time to respond. It also asks what happens to people who cannot or do not wish to use the digital route. A service can improve average speed while worsening access for a small group; an automated recommendation can aid a worker while hiding the evidence needed to challenge it. The concept becomes useful only when these tensions are made observable and answerable.
From a digital channel to a human outcome
Digitalisation is the use of digital technologies to change how an activity or institution works. Scanning a paper form is one step; redesigning how a service communicates, records, decides and learns is a much larger change. Human-centric digitalisation evaluates that change by its consequences for people. A portal’s uptime and number of accounts matter, but they are not sufficient. A resident must be able to understand what is asked, complete a real task, obtain help, and correct a wrong outcome. A worker must be able to see relevant information and exercise judgment rather than carry responsibility for a process they cannot influence.
The European Declaration on Digital Rights and Principles places people and their rights at the centre of digital transformation. Its themes include solidarity and inclusion, freedom of choice, participation, safety and sustainability. These are policy commitments and evaluative directions, not a single empirical scale of human-centredness. They can guide concrete questions: Does a person have an accessible route? Can they control personal data? Is there a meaningful choice when a system is unsuitable? Does participation include people who encounter the service at its hardest point, not only those who join a consultation workshop?
Several neighboring terms are easy to confuse. User-centred design studies the needs and behavior of intended users to improve a product or service. Accessibility concerns whether people with varied abilities can perceive, operate and understand it. Inclusion asks who can participate and benefit under real social conditions. Human-centric digitalisation draws on all three, while adding rights, institutional responsibility and the distribution of costs and authority. A polished user journey can still be exclusionary if identification rules deny access, appeals are obscure, or staff lack permission to make an exception.
The language of centring humans should not mean that every difficult choice is transferred to an individual. A person can be given five settings and still have no workable way to protect privacy. A frontline worker can be told they remain responsible while a recommendation arrives too late to check. A meaningful choice needs information, time, authority and an alternative that works. Without those conditions, human control becomes a symbol in a process whose consequential decisions have already been made elsewhere.
European and Swedish policy context
The EU AI Act states among its purposes the promotion of human-centric and trustworthy artificial intelligence while protecting health, safety and fundamental rights and supporting innovation. Its requirements vary by system and risk category. The Act is a legal framework; it does not certify that a specific implementation is fair, accessible or beneficial. A public body still has to identify the system’s actual purpose, its users and affected people, applicable obligations, and the evidence that oversight works in that workflow. The presence of a human reviewer on an organizational chart does not by itself establish meaningful review.
Sweden’s 2026 AI strategy sets ambitions for national capability, innovation and public-sector adoption. Its official page describes a starting point of high digital maturity, access to public-sector data and competitive infrastructure. Those are government assessments and goals. They do not mean every municipality, clinic or individual has equal access or that each proposed use is appropriate. A national strategy can mobilize resources, but local evidence is still needed about service outcomes, staff capacity and groups whose experience the average conceals.
The European Commission’s 2025 State of the Digital Decade report monitors progress toward 2030 objectives and identifies persistent gaps in skills, infrastructure and advanced-technology adoption. It is useful as a policy dashboard. It cannot tell us whether a particular appointment system lets a person with limited language proficiency receive care on time. Macro indicators and lived service quality answer different questions. Both matter: a society needs infrastructure and skills, and it needs grounded checks that these resources make institutions usable.
Human-centric digitalisation thus belongs to a wider regional debate about trust and public value. Nordic countries may begin with comparatively mature digital infrastructure, yet maturity can create its own blind spots. When a digital identity or shared platform is widely assumed, people who cannot use it may become less visible. A high level of public trust is valuable but should not replace clear routes to explanation, correction and appeal. Trustworthy institutions earn trust partly by showing what happens when technology fails.
Policy also changes over time. Rules may be amended, guidance revised and administrative practice interpreted by courts or supervisory bodies. An article about this concept should be reviewed when major legal or policy changes alter the relevant obligations. It should not turn a 2026 strategy page into a permanent statement about the future. The stable principle is the need to relate a digital system to real people’s rights and capacities; the exact legal route must be checked in the jurisdiction and date of use.
What a human-centred service would measure
Consider an online benefit application. A conventional dashboard might count completed forms, average processing time and staff hours saved. A human-centred evaluation would keep those measures and add others: how many people start but cannot finish, how errors are detected, whether a refusal can be understood and challenged, whether an assisted route works, and how long resolution takes after a problem. It would disaggregate results where lawful and appropriate to see whether a change helped some groups while disadvantaging others. Numbers alone would still need interpretation through interviews and observation.
Measurement should follow the causal path of a service. If the goal is timely access, the clock should not stop when the user presses submit while the case waits elsewhere. If the goal is informed consent, a checkbox is weak evidence unless people understand the choice and can act differently. If the goal is meaningful human oversight, an override count by itself is ambiguous. A low count might indicate accurate recommendations, pressure to agree, or an interface that makes dissent too costly. Evaluators need case-level evidence about what reviewers saw, what alternatives they had and what happened after disagreement.
These examples are Institute synthesis drawn from the policy principles and evaluation literature; they are not a validated universal index. The purpose is to translate broad commitments into falsifiable local questions. An institution can set a threshold for a particular task, examine failures and revise the design. It should also ask whose work is hidden. People may spend more time gathering documents, managing passwords or repeating their story while the agency’s processing metric improves. A human-centred account includes the user’s time and the staff member’s practical burden.
Access has several dimensions. A service may be technically accessible to assistive technology yet impossible to use with slow connectivity. It may offer plain language but require a document that an applicant cannot obtain. A translated interface may be undermined by untranslated notices or support channels. Testing must follow the full journey, including identity verification, the decision and any appeal. It should involve people with varied needs without making them responsible for proving that an institution’s design is flawed. Accessibility is a continuing obligation of design and maintenance, not a one-time test at launch.
Participation also needs careful design. A consultation dominated by confident, digitally fluent users may improve a service for that group while leaving its hardest cases unexamined. The institution can recruit participants through trusted community routes, compensate their time and show how their input changed decisions. Even then, participation is not a substitute for representative outcome data or rights protection. A service should not be able to justify a harmful feature because a small panel once approved a prototype.
Technology, work and accountable authority
Digitalisation changes professional work as well as the public interface. Automated triage can direct attention to urgent cases, but it can also narrow what a worker notices. A dashboard can summarize evidence, but it may omit context known to the person affected. Human-centred design should give staff a way to inspect sources, record uncertainty, ask for more information and change a proposed outcome. These capabilities must be supported by training and time. A formal power to override is thin if the worker is rewarded only for speed or cannot view the underlying evidence.
Institutional accountability cannot be delegated to a vendor or a model. Procurement should specify which outcomes matter, what information will be available for audit, and how the system will be corrected or retired. An organization needs to know where data came from, when a system performs poorly, and who has authority to pause it. Security and privacy matter because the cost of a breach falls on people whose information was collected in order to receive a service. A human-centric approach treats those risks as part of service quality.
There can be genuine trade-offs. A highly tailored service may need more personal data, while privacy protection calls for minimization. Strong identity checks may prevent fraud but block people without conventional documents. Automation can reduce waiting time while changing the relationship between staff and residents. No slogan resolves these tensions. The institution should state the trade-off, identify whose interests are at stake, compare feasible alternatives and record why one design was chosen. It should later test whether the predicted benefits and harms occurred.
Energy, devices and procurement also belong in the account. The European Declaration includes sustainability as a digital principle. A system that offers a small convenience while requiring avoidable replacement of working devices or large ongoing resource use may shift costs into another domain. This does not imply that every digital service is environmentally worse than its predecessor. It means the comparison should include material consequences and the people who bear them.
What the evidence does not show
The evidence does not show that an online service is human-centric because it is popular, fast or used by many people. Aggregate uptake can coexist with exclusion and failed appeals. A positive survey of users says little about people who could not become users. Nor does a rights declaration prove that every implementation meets its principles. Policy texts establish commitments and obligations; service evaluations must show how those commitments work in practice.
Sweden’s AI strategy does not prove that a particular AI system improves public value, and the Digital Decade indicators do not measure every local barrier. The AI Act does not replace a contextual assessment of a system, its legal basis or its actual effects. A person clicking the final button is not necessarily exercising oversight. Each claim needs a defined population, task, comparison and outcome. Costs to staff, inaccessible routes and delayed correction should not disappear because an average processing metric improved.
The concept also does not demand that all services become digital. In some circumstances a telephone, in-person or assisted path is the human-centred choice. A multi-channel design can preserve agency if the alternatives are reliable and not punitive. The aim is to use technology where it helps people act, understand and obtain a fair response. The correct digital share of a service is an empirical and democratic question, not a target that can be set by the word modernisation.
From principle to revision
The Institute treats human-centric digitalisation as an institutional capacity: the ability to translate rights and technical possibilities into services that people can use, question and improve. This is an editorial synthesis. The cited policy sources establish aims and monitor broad trends; they do not validate one Institute measure of that capacity. A practical test would examine a real task from entry to outcome, follow people who struggle, and ask whether errors can be corrected without exceptional personal effort. It would then revise the service and repeat the test.
The opening application form offers a simple benchmark. It should work with the screen reader, offer another route when it does not, explain what information drives the decision, and provide a timely human path to remedy. The worker receiving the case should have usable evidence and authority, not merely a responsibility statement. The institution should record failures in a form that designers and accountable leaders can learn from, and tell affected people when a correction has been made. These conditions are harder to deliver than a launch announcement. They are also the conditions under which a digital change can credibly claim to serve people. Human-centric digitalisation is sustained work on those conditions, measured at the point where a person’s life meets the system.
Research references
- European Parliament, Council & European Commission. (2022). European Declaration on Digital Rights and Principles for the Digital Decade. View source
- European Parliament & Council. (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence. View source
- Government of Sweden. (2026). Sveriges AI-strategi. View source
- European Commission. (2025). State of the Digital Decade 2025 report. View source
- European Commission. (2025). Digital Decade 2025 Special Eurobarometer. View source