DATA GOVERNANCE
Data governance is not a set of rules.
It is an operating system.
Invisible as long as it works.
Painful as soon as it is missing.
In many organizations, data governance is only discovered when something goes wrong:
- when data cannot be explained,
- when its “ownership” is disputed/unclear,
- when auditors ask uncomfortable questions,
- when new laws such as the EU AI Act and the Data Act, which will come into force on September 12, 2025, enter the stage,
- when AI systems suddenly make autonomous decisions that no one wants to take responsibility for.
I don’t see data governance as a separate project, but rather as a cross-cutting concern that must underpin every data-driven project. And that means right from the start.
Compliance by design.
Whether it’s cloud, AI, analytics, or platform building:
without data governance, value creation quickly becomes a high-stakes game of chance.
WHY DATA GOVERNANCE IS DIFFERENT TODAY
Data governance used to be an “Excel problem.”
Today, it is a legal, architectural, and management problem, or rather, a challenge.
With the GDPR, Data Governance Act, Data Act, and AI Act, one thing is clear:
Governance is no longer a voluntary arrangement; it is subject to verification.
And it doesn’t end with data protection.
It encompasses data quality, roles, processes, tooling, documentation, traceability, and above all:
the ability to make decisions under uncertainty.
MY APPROACH
I don’t come from the traditional “data office” corner.
I come from architecture, cloud, AI – and law.
That’s why I don’t design data governance as a theoretical target,
but as a workable structure that:
is regulatory viable
remains technically compatible
is accepted by the organization
and makes economic sense
I am currently working at GIZ, among other places, on a global data platform
where data governance is not a side note,
but a prerequisite for collaboration across national, organizational, and legal boundaries.
MATURITY LEVEL INSTEAD OF QUESTIONS OF FAITH
I deliberately work with maturity models, not with absolute claims.
Not every organization needs to be at level 5 right away.
But every organization needs to know where it stands –
and what risks it is consciously taking.
For me, maturity progress means:
- moving from reactive data chaos
- to explainable, controllable data usage
- with clear responsibilities and reliable evidence
ROLES THAT DON’T DIE ON PAPER
Data governance rarely fails because of technology.
It fails because of unclear roles.
Data owner, data steward, AI officer, data trustee.
These roles do not exist to make organizational charts look prettier.
They exist because EU law enforces them.
I help define roles in such a way that they work in everyday life:
- with clear tasks
- realistic competencies
- and genuine responsibility
Not everyone needs a new committee.
But everyone needs clarity about who is allowed to make decisions and who is not. Or as a data scientist once said in one of my workshops on data governance (including for GFU, here: https://www.gfu.net/s3711):
“We have the problem that no one knows who is responsible for the data we are supposed to process.”
DATA QUALITY AS A LEGAL OBLIGATION
With the AI Act, it is now clear:
Data quality is no longer a “nice to have.”
It is a prerequisite for approval.
Representativeness, completeness, timeliness, bias control—these are not buzzwords, but criteria by which systems will be measured in the future.
That’s why I don’t treat data quality in isolation, but as an integral part of governance:
- with measurable rules
- quality gates
- monitoring
- and documented decisions
Not perfect.
But explainable. And defensible.
TOOLS ARE AIDS – NOT SAVIOURS
Data catalogs, business data dictionaries, compliance tools:
They can help a lot.
But they are no substitute for decisions.
I always advise on tool strategies from an architectural perspective:
- Interoperable instead of proprietary
- Auditable instead of convenient
- Connectable instead of shiny
Technology is an enabler.
Governance arises elsewhere: in minds, processes, and responsibilities.
CHANGE MANAGEMENT: THE UNCOMFORTABLE PART
Data governance changes power.
Who has access.
Who decides.
Who has to explain.
That’s why it often fails quietly, not infrequently passively-aggressively, through resistance, ignorance, or withdrawal.
I consciously work with change mechanisms that don’t “sell” governance as control (fear works well in dictatorships but is deterrent and thus reduces the workforce of intrinsically motivated people),
but as a space of opportunity:
- for better decisions
- for trustworthy AI
- for genuine collaboration
Governance that no one understands will be circumvented.
IN A NUTSHELL
I build data governance for organizations
that want to use data without having to justify it later.
Not to the maximum extent.
But appropriately.
Not dogmatically.
But sustainably.
If you don’t see data governance as a compulsory exercise, but as a prerequisite for responsible value creation, then we are probably on the same page.