ROI and time-to-delivery will always kill governance efforts. And 2026 will confirm this more than ever.

The Return to Centralization

AI agents are pushing toward increasingly monolithic modes. The promise of mesh, federation, distributed self-service runs into a reality: agents work better with unified, coherent, centralized context.

Fragmenting data across autonomous domains creates friction for AI. And friction, in an agentic world, is lost time-to-insight.

Whoever Controls Context Controls Output

This is the critical point few anticipate yet: whoever configures the agent controls what it produces.

The context injected into an agent determines its responses. Not the raw data, not the model, but the context. This will become a major political issue in enterprises. Data teams will need to reposition: it’s no longer just about pipelines and models, it’s about information control.

The End of Data Factories

Data teams organized in “factory” mode will have to question themselves. When friction and time-to-insight trend toward zero, the industrial data production model loses its meaning.

Business doesn’t need a factory delivering datasets. It needs answers. Immediate. Reliable. Contextualized.

The data teams that survive will be those that switch to direct value delivery: less production, more context curation. Fewer pipelines, more feedback loops.

Data Quality Becomes Critical

With the inversion of the testing pyramid, data quality takes on a new dimension. We no longer test to validate columns—we test to guarantee agent determinism.

An agent that responds differently to the same question depending on timing is an unusable agent. DQ is no longer a compliance or best practice topic. It’s a functional prerequisite.

The Announced Death of Catalogs

Data catalogs don’t deliver direct business value. Their adoption comes down to three points:

  1. Regulatory compliance — obligation more than choice
  2. Asset size — large organizations need inventories
  3. FOMO and vendor-driven marketing — implemented because everyone else was doing it

Rarely because they delivered measurable value.

Pragmatism will prevail. These tools will be replaced by the natural feedback loops of AI agent interfaces.

Documentation Becomes AI Fuel

This might be the most important shift.

Before, we documented altruistically. ROI was fuzzy, incentive weak, degradation inevitable.

Now, documentation directly feeds AI context. No doc = no context = no reliable answer.

Feedback is immediate: if the agent responds poorly, it’s visible. The cost of non-documentation becomes tangible and measurable.

Consequence: the catalog as an externalized static artifact will disappear. What matters is that context gets injected into the agent—therefore as close to the data as possible, in the code.

The discovery interface is no longer a static web portal. It’s the conversation with the agent itself, with its feedback loops serving as data stewardship.

The New Paradigm

Data governance in 2026 won’t look like what we’ve been sold for ten years. Fewer processes, fewer dedicated tools, less dead documentation.

More living context, more immediate feedback, more direct value.

ROI will have the last word. As always.