Chad Barendse here, co-founder of DGX, and welcome to the April newsletter.

It has been a busy month. ADG 2026 planning is well underway, the speaker lineup is shaping up to be something special, and we are seeing strong early interest from organisations looking to get serious about governing AI. More on that soon.

But the thing I have been thinking about most this month is a shift I am seeing across almost every organisation I work with.

Data governance is becoming infrastructure.

Not in a buzzword sense. In a structural sense. The traditional model, central policy groups, manual registers, compliance reviews run by committees, was built for an era when data moved slowly. That era is over.

Teams are shipping data pipelines faster than governance can review them. In most large organisations, you have 30-40 or more scrum teams shipping changes every two weeks. It is not possible for a governance team to be across all of that. I have seen it firsthand. By the time I find out about something its well into implementation, making it hard to make any changes.

AI has made this worse, or better, depending on how you look at it. You cannot separate governing the data from governing the AI that runs on it. They are the same problem. The organisations still running data governance and AI governance as separate programmes with separate committees are creating gaps they cannot see.

The piece I keep coming back to is the context layer. Metadata, lineage, business definitions, classification rules. For years we have treated these as documentation exercises. Artifacts for audits. What I am seeing work is treating them as a product. A small team owns definitions and lineage. They version them, test them, integrate them into the platform. When new data sources come in, context is built at ingestion, not bolted on six months later.

That changes everything downstream. Quality issues get caught earlier. AI outputs can be traced.

If you are thinking about where to start, three things:

1. Merge your data governance and AI governance into one function. Same data, same risks, same owner.

2. Build your context layer before you automate anything. Automation without context is just faster mistakes.

3. Move your team’s focus from writing policies to embedding controls into the systems where data actually lives.

I went deeper on all of this in a substack post this month. If this resonates, it is worth a read.

Read the full post: Data and AI Governance Is Becoming Infrastructure

Thanks for reading