Farmers respect the need for harnessing nature’s power strategically. Irrigation, sunlight, toil—all combine to nurture precious plants. For success at any scale, processes must be set, followed, and modified as needed. Growing plants takes time and care.

Building high-quality information systems takes just as much time and attention. First, you must understand the landscape: what kind of data do you have, and where is it housed? Can you effectively scan the entire environment, or only parts?

Understanding the topography of your de facto information architecture is paramount. Luckily, there are many tools, technologies and methods for achieving this. And it’s especially crucial in today’s AI-infused world! The stakes are higher than ever.

That’s because AI tends to have a voracious appetite for data, especially the gritty, granular kind. Don’t think high-level summaries, like those dashboards that have been built over the years. That’s not what AI wants. It wants the granular details, en masse.

Policies Prevail

Assuming you can take a valid inventory of your information assets, you’re then ready for the next big step: policies! This requires a thoughtful approach to the what, why and how of data: What’s its purpose? How will you acquire, persist, manage and use it?

In the United States, there are many state-level laws and regulations that apply. The fact that there are different standards depending upon geolocation makes the job of data governance significantly more challenging. You’ll need layers of abstraction.

The good news is that modern info systems tend to offer the kind of functionality necessary for achieving this. Exactly how policies are put into place and managed will depend heavily on which systems are in play, but there are plenty of knobs and levers.

It’s important to run some processes concurrently, or at least be open-minded about how the program progresses. Reason being: You want to know what you’re trying to do with the data, and also know what’s possible within your systems, but…

… your priorities likely will, and arguably should, change over time. As such, data governance policies should not be etched in stone, but rather codified in rules that can be amended quickly and efficiently.

Until recently, that was not so easy. You either had to enable access at a database level, or within specific applications. That works for small teams, but not for larger organizations. When companies reach a certain size, such manual efforts falter.

But these days, many cloud platforms enable robust policy management for both information and application access. After all, that’s what governance is all about: Controlling access and usage, thus enabling reasonable controls that work.

Evolving Landscape

The rise of powerful AI models, particularly those with generative capabilities like ChatGPT and others, throws stark relief onto the importance of data governance. As companies grapple with the potential and pitfalls of generative AI, understanding how to manage, secure, and utilize the vast amounts of data these models consume becomes absolutely essential.

Here are some key aspects of data governance highlighted by the increasing use of AI:

• Mitigating Bias: AI systems inadvertently “learn” biases from the data they train on. Without proper governance, these biases can propagate harmful stereotypes or lead to discriminatory outcomes. “Data governance practices are essential for uncovering biases. AI cannot overcome bias in the data it has been trained on,” says Bao-Ha Bui, SVP of FPT Software Americas.

• Ensuring Data Quality: High-quality, well-structured data is the foundation for accurate AI results. Governance practices ensure that data is consistent, reliable, and fit for purpose.

• Data Privacy and Security: With growing concerns about data misuse, data governance safeguards sensitive information, ensuring it’s handled ethically and within regulatory compliance. “Effective data governance is an essential part of being a good steward of data entrusted to you,” emphasizes Jeff Witt, Fractional CTO, and a regular collaborator with Bui.

• Interpretability and Explainability: AI models can be complex, raising the need to explain their decision-making processes. Data governance ensures that data lineages and transformations are auditable and transparent. This fosters trust in AI-generated insights.

The Way Forward

To navigate this dynamic landscape, organizations must embrace robust data governance frameworks. This includes clear policies and procedures that cover:

• Data ownership and accountability
• Data access and usage guidelines
• Data quality standards
• Processes for bias detection and mitigation
• Regulatory compliance

“The increasing prevalence of AI does not eliminate the need for data governance – rather, it amplifies it,” asserts Bao-Ha Bui. “Organizations must approach AI development and use with a focus on data governance. This will ensure that AI is deployed in an ethical, responsible, and trustworthy manner.”

Simply put, data governance is no longer optional. As AI advances, data governance serves as the responsible anchor, addressing ethical concerns while facilitating the full potential of this transformative technology.

Getting back to our original analogy, data governance is like the process framework for running a farm. Weather changes all the time, and severe weather can cause real damage if protocols are not in place to protect crops. Don’t get caught in the storm! Careful is as careful does. Governance is all about caring and attention to detail!

Now, let’s make it rain!