The intelligence of the state is measured by the quality of its responses: which questions it can answer accurately, based on what evidence, and within what timeframe. That capacity for response depends entirely on the data the institution holds. When data is properly managed, the state becomes intelligible to itself; when it is not, intelligence is a baseless narrative.
The promise of public AI depends entirely on the quality, lineage, ownership, and standards of the data that feeds it. Before the model comes the data; before the data comes data governance. That sequence explains why the path to institutional intelligence begins with the raw material that will sustain it.
Intelligence is based on data
A smart government makes decisions based on evidence, designs programs based on real data, and anticipates risks based on observed patterns. Each of these capabilities relies on data that reflects the world, flows between systems, retains its history, and has someone accountable for it. Intelligence is the effect; governed data is the cause.
The four pillars of data governance are well-known, proven, and replicable: quality, lineage, ownership, and standards. Each one addresses a specific problem, and together they transform data into an institutional asset on which intelligence can rely with confidence.
Quality: Data That Deserves to Be Used
Quality means that the data is accurate, complete, up-to-date, and consistent. It means that the citizen’s name matches across systems; that the registration date reflects reality; that the codes contain no typos or gaps; and that the records are updated in step with the actual process. Each of these attributes may seem minor on its own, but ceases to be so as soon as a model learns from flawed data: it inherits the errors and multiplies them.
Quality is built on rules, metrics, and routines. It involves defining what “valid data” means in each domain, measuring how much of the data meets that definition, and designing processes that detect and correct deviations in a timely manner. Quality is a continuous practice sustained over time, not a project with a fixed end date.
Lineage: knowing where each figure comes from
Data lineage refers to the traceability of data through its various transformations: from the system where it originates, through the processes that clean, join, or aggregate it, to the dashboard or model where it is ultimately consumed. Knowing the lineage allows you to answer basic questions— Where does this number come from? When did it change? Which process modified it?—and diagnose errors in minutes rather than weeks.
When a lineage is recorded, the state gains institutional memory: anyone with permission can trace the path of a piece of data and understand what decisions were made regarding it. Auditing ceases to be an archaeological exercise and becomes a consultation. This is the foundation that makes it possible for AI to be explainable and for algorithmic decisions to be subject to appeal.
Ownership: every piece of data has an owner who is accountable
Ownership is the most institutional of the four elements. Every relevant dataset needs a designated owner: a person—backed by their department—who defines its conceptual model, authorizes its use, ensures its quality, and communicates any changes. Without that assigned role, the data remains in no man’s land: anyone can modify it, no one looks after it, and the organization loses the ability to explain itself.
Ownership works when it is accompanied by a data steward — a technical specialist who handles day-to-day operations alongside the functional owner — and by a data governance committee that resolves issues that span agencies or departments. It is purely institutional work, and it is just as valuable as any strategic role: it defines what information will underpin the next decade of public policy.
Standards: the common language that prevents us from reinventing the wheel
Standards are the shared language of data: formats, unique identifiers, taxonomies, catalogs, metadata, reference codes. When they are defined and followed, two systems that have never met can communicate; when they are missing, each new project creates its own dictionary and adds another layer to the puzzle.
In the public sector, standards are underpinned by an institutional agreement: someone defines the standard, someone maintains it, and someone ensures that new systems adopt it. Leveraging international standards (ISO, OGC, FHIR, digital government schemas) speeds up the process and connects the country to the regional ecosystem, saving years of work in each sector.
All four pieces in a single photo
- Quality: The data reflects reality and remains useful over time.
- Lineage: Each figure can be traced back to its origin and through every transformation.
- Ownership: There is a designated owner who is responsible for the content and authorizes its use.
- Standards: common formats, identifiers, and catalogs that enable systems to communicate with one another.
The data debt
Just as there is technical debt in software, there is also data debt: the accumulation of shortcuts, duplications, and makeshift definitions that each project has left in its wake. Accumulating silently over the years, this debt suddenly comes due the day an organization tries to apply AI to its records and discovers that half the effort goes into cleaning up what should have been clean from the start.
Why does this matter?
Because the promise of a smart government rests on well-managed data. Every predictive model, every automated agent, and every management dashboard depends on the quality and traceability of the data it uses. Investing in data governance is investing in the only foundation that makes the intelligence built upon it accountable.
At Sofis, we support institutions in their day-to-day work: defining quality standards, implementing lineage, assigning owners and stewards, adopting standards, and integrating all of this with the systems already in production. This work is both institutional and technical, and it forms the foundation upon which government intelligence moves from being a promise to becoming an established capability.