AI is changing how organisations use data. But while businesses are investing heavily in AI models, copilots, and automation, there is a foundational question that too often gets overlooked: can we trust the data those systems depend on?
I’ve spent years working in data governance, and one thing has become increasingly clear to me: the governance practices that were sufficient for dashboards and traditional analytics are not necessarily enough for AI.
AI operates at a different scale, at a different speed, and with potentially much greater consequences. That means organisations need to rethink what data governance looks like in an AI-enabled world.
That is the focus of my LinkedIn Learning course, Data Governance in the Age of AI.
Data Governance Is About More Than Rules
At its simplest, data governance ensures that data is managed as a strategic asset.
That means agreeing on definitions, understanding where data comes from, establishing standards and processes, defining access policies, and making sure there are clear owners and responsibilities.
Consider something as seemingly straightforward as “monthly active users.”
If one team defines an active user as someone who logs in, while another defines them as someone who completes a transaction, both teams can produce technically correct reports while arriving at completely different numbers.
Without data governance, there is no agreed definition or process for resolving the disagreement.
With data governance, there is.
This becomes particularly important when AI systems begin using the same data to make recommendations or decisions at scale.
AI Amplifies Existing Governance Problems
Traditional analytics usually involves several opportunities for human intervention. Someone prepares the data, builds a report, reviews the results, and makes a decision.
AI changes that process.
AI systems can process vast quantities of information and make or influence thousands—or even millions—of decisions far more quickly than a human team could.
That means a data governance problem that once resulted in an inaccurate report could become a much larger operational or financial problem when embedded in an automated AI system.
The three factors I focus on are scale, speed, and consequence.
AI operates at scale, meaning errors can affect enormous numbers of records or decisions.
It operates at speed, meaning problems can propagate before anyone has an opportunity to intervene.
And the consequences can be significantly greater when AI is involved in areas such as hiring, lending, healthcare, or access to public services.
The question is no longer simply whether a dataset is good enough for a report.
We need to ask whether it is good enough to support an AI system that may act on it automatically.
When Access Isn’t Inclusion
Why digital strategies must move beyond coverage metrics to measure skills, trust, identity, and AI readiness as core participation outcomes.
Good Data for Reporting Isn't Necessarily Good Data for AI
One of the most important distinctions organisations need to understand is that data quality for analytics and data suitability for AI aren't always the same thing.
A dataset can be clean, accurate, and well governed from a traditional reporting perspective while still creating problems for an AI model.
Consider a healthcare dataset that accurately reflects decades of patient records. If those records disproportionately represent people who had access to healthcare, the dataset may be technically accurate but not representative of the wider population.
The problem isn't necessarily data.
It's incomplete context.
AI systems can expose these weaknesses because they learn patterns from the data they're given. If those patterns don't represent the population or situation the system is intended to serve, the resulting predictions can be unreliable.
This is where data governance needs to evolve.
Three New Challenges for Data Governance
AI introduces several areas that traditional governance programmes were not originally designed to address.
When Data Governance Drives AI
A deep look at data architecture, governance and composable design as foundations for generative and agentic AI in complex enterprises.
Unstructured Data
Traditional governance programmes have often focused on structured and semi-structured information stored in databases and data pipelines.
But AI can consume much more than neatly structured tables.
Documents, emails, images, conversations, audio, and other forms of unstructured information can all become relevant sources of data for AI systems.
If organisations don't know what information they have, where it came from, who owns it, or how it can be used, they have a governance problem.
AI Data
AI introduces new categories of data that need to be governed.
This can include training data, feature data, input data, and AI-generated outputs.
Organisations need to understand questions such as: Who does this data represent? What biases might it contain? Was it collected appropriately? Were the necessary permissions or consents obtained? How is it being used?
These questions need to become part of the governance conversation.
The AI Lifecycle
Data governance also needs to extend across the AI lifecycle rather than stopping when data enters a traditional reporting environment.
Data can influence an AI system during training, development, deployment, inference, and output.
Governance needs to account for those different stages and establish appropriate controls throughout the process.
AI Self-Improvement Risk Line
Mythos-class cyber tools, US–China rivalry and narrowing human oversight collide as frontier AI races beyond traditional safety controls.
Making Traditional Governance AI-Ready
The good news is that organisations don't necessarily need to throw away their existing data governance programmes.
Instead, they need to extend them.
Existing governance assets such as business glossaries, data catalogs, and metadata stores can provide an important foundation for AI governance.
But organisations may need to expand the types of assets they govern, introduce additional quality dimensions, develop new policies, and clarify responsibilities across the AI lifecycle.
Perhaps most importantly, organisations need to understand where data governance ends and AI governance begins.
They are connected, but they aren't the same thing.
Data governance focuses on the data: its quality, definition, ownership, context, access, and appropriate use.
AI governance addresses the broader behaviour and risks of AI systems.
Keeping that distinction clear helps organisations build governance programmes that are comprehensive without becoming unnecessarily complicated.
Governance Needs to Become Part of AI Strategy
AI readiness isn't simply about selecting a model or deploying a new tool.
It's about creating the foundations that allow AI to operate reliably and responsibly.
That means knowing what data you have, understanding what it means, establishing who is responsible for it, and putting the right controls around how it is collected and used.
It also means recognising that AI doesn't eliminate the need for data governance.
It makes it more important.
In my Data Governance in the Age of AI LinkedIn Learning course, I explore these challenges through real-world AI incidents, practical governance concepts, and the data governance responsibilities that exist throughout the AI lifecycle.
The course also includes AI-powered Role Play, allowing learners to practise what they've learned through interactive simulations of real-world conversations.
When AI Theft Meets Geopolitics
Anthropic’s claims against Alibaba show how model scraping is reshaping export controls, sanctions policy and AI vendor due diligence.
The goal isn't simply to understand data governance as a set of policies and processes. It's to understand how those foundations need to evolve when data starts powering AI systems that can operate at unprecedented scale and speed.
Because trustworthy AI doesn't begin with the model.
It begins with trustworthy data-and the governance that makes it possible.
Comments ( 0 )