AI Is Becoming a Data Problem Before It Becomes a Model Problem

[romquesta]

The conversation around AI often starts with models. In reality, long-term value comes from how well data pipelines, governance, and workflows work together.

The teams moving fastest seem to be simplifying their stack instead of expanding it.

What trends are you seeing?

[pragya17]

Yes, fastest moving teams are focussing more on unified data platforms, strong governance ,workflow-native AI , measurable operational outcomes. Few clear trends are - smaller and specialized models for cost and reliability ,focus on structured , governed environments like Databricks (unity catalog ), Retrieval/content quality, embedded AI workflows . So AI is moving towards data governance and discipline more than AI model race.

[amirabedhiafi]

Few weeks ago I already wrote an article about it Are we still keeping the same data warehouse modelling standards in 2026 with all this AI?

AI changes the way people interact with data but it does not remove the need for modelling.

A bad data model with AI is still a bad data model.

Actually, it is worse, because AI can make wrong answers look confident.

Because in the AI era, trusted data is not optional.

It is the product.