Let's discuss what problems may arise from technology-driven data governance.

The biggest problem in data governance lies in technology because technology determines the goals we need to optimize. This process may be public or indirect, but it will indeed happen. So, you may start your data governance journey because you understand its value, because you see it as a compliance task, or because with the emergence of artificial intelligence, your organization realizes the need to improve data quality, and governance seems to have become a necessity. Where should you start? Usually, some common elements mark the start of a plan or project that will shape your data governance system. These elements may be:

Define some goals based on the triggering conditions of the project

Form a team

Evaluate your vision and maturity

Develop a framework for data management policies, standards, and procedures

Set roles and responsibilities

The question is: How can an organization clearly define these contents before considering implementation? And at this time, the situation often changes: things become more concrete and require implementation, whether it's security policies, quality metrics, metadata repositories, etc. For implementation, tool vendors like to provide guidance and even jointly define the competitive environment. This seems reasonable. After all, vendors know their tools best, don't they? They provide implementation support and usually position themselves as data governance experts.

But the problem is that vendors optimize the functions of their tools rather than the actual data governance functions your organization needs. This inconsistency usually causes data governance work to focus more on policy execution rather than strategic support due to the tool's functions.

With the development of artificial intelligence, this problem becomes even more serious. We've heard it many times: the quality of artificial intelligence depends on the data it receives. A data governance model that optimizes around tools rather than purposes will only perpetuate the classic "garbage in, garbage out" trap, especially at the speed and scale of machines.

You can think of it as the GIGO chain: Poor planning + Vendor-optimized implementation + Opaque AI models =?

Viewpoint 1: Data governance is not a tool

Vendors sell tools, and these tools are built to implement data governance, not to define it. Therefore, when vendors take the lead, the focus quickly shifts to:

Metadata tagging

Rule execution

Access control

Monitoring dashboards

These are all valuable capabilities, but only if you clearly understand the original intention of governing data. Data governance is about the direction, supervision, and accountability of how data is used in an enterprise information system, not about tool configuration.

My definition of data governance takes this into account:

Data governance is a human-centered system through which data assets in an enterprise information system can be guided and supervised, and the organization is required to be accountable for achieving its established goals.

This definition does not start with tools but with people and goals. Tools are the result of our well - considered long - term and short - term choices. Short - term choices are based on business needs, regulatory compliance, market conditions, and limitations. Long - term choices are based on our vision for the organization. Finally, equally important: tools are a way to ensure accountability and supervision, not to define them.

This is the basis for defining the three roles that data governance must play in a modern organization: Data Director, Data Steward, and Data Auditor. This is exactly what is needed to define data governance independently of execution in an organization.

Viewpoint 2: Governance is not just about implementing systems

When data governance implementation becomes vendor - led or tool - dependent, it's easy to optimize tool performance by defining the requirements that data must meet. This may be because the tool comes with a domain structure, predefined settings for metadata collection, or for example, the need to confirm the definition of a data product.

If you start with tools, the focus of data governance will be on implementing policies rather than balancing business goals, regulatory requirements, market pressures, and technical limitations. The resulting governance framework will have the following characteristics:

Give priority to compliance over availability

Optimize policy compliance over innovation

Create checklists instead of cultivating a data culture

You may end up with an excellent execution workflow, but there will be no common understanding of what is being executed or why.

Once artificial intelligence becomes part of the process, decision - making becomes faster and more autonomous. At this time, the lack of clarity in data governance will cause systemic risks. A superficial, vendor - optimized, tool - first data governance model will only accelerate the "garbage in, garbage out" cycle. Think about our GIGO chain. Suddenly, even if our data governance implementation has defined clear rules for data security, data quality, data lineage, etc., biases will be amplified, errors will be magnified, and responsibilities will be blurred, all because the basic data governance framework was not stably defined from the start.

Action 1: What should be optimized

So this means you need to put in some work before turning to tools. Effective, human - centered data governance requires:

1. Start with goals

What goals does your organization hope to achieve with data? What is your vision? What risks do you have to manage? What value do you have to realize? Data governance should stem from these questions, not from product features.

2. Establish a supervision mechanism, not just monitoring and control

Clarify who has the decision - making power, how to resolve disputes, and how to track accountability. Data governance is an accountability system that goes beyond operations. It is a governance topic and should be consistent with the way the company is governed.

3. Use tools for implementation and operation, not for definition

Once the governance framework is in place, tools can speed up execution. But they must serve the human - defined data governance system, not the other way around.

4. Data governance as a living system

Data governance is not static. As business models evolve, regulations change, and new technologies emerge (such as artificial intelligence), governance must be adjusted accordingly. This means continuous reflection, measurement, and iteration to explore the core reasons for data governance, not just how to do it.

Action 2: Don't outsource the difficult parts

The fact is that defining data governance is not easy. It involves continuous difficult communication, weighing pros and cons, conflicting incentive mechanisms, and cultural change.

This is why people tend to let vendors fill in the gaps, starting with implementation and then working backwards. But data governance cannot be reverse - engineered through tools. It must be carefully designed by people who understand the organization's mission, risks, and values.

Stop seeing data governance as a plug - and - play feature. It is the core of an enterprise's data - driven approach. This system defines the interconnection between machines and people and requires guidance, negotiation, and accountability. When you hand it over to a vendor, you are not just buying a tool; you are outsourcing your values, direction, and supervision.

This article is from the WeChat official account "Data - Driven Intelligence" (ID: Data_0101), written by Xiaoxiao, and is published by 36Kr with authorization.