This reflection was sparked by two Gartner reports that, at first glance, seem unrelated. However, when read together, they paint a very telling picture of the current state of AI adoption.
In practice, these two theses often collide in a common scenario: companies logically deduce that "AI-ready data" is the key prerequisite for AI success. This leads to a dangerous shortcut: the belief that if we build a robust data platform and improve data quality, AI will simply "plug in" and start working. In other words, D&A is viewed as the primary enabler of AI adoption.
But that is only half the truth.
Modern data platforms do an incredible job with the essentials: integration, transformation, quality assurance, traceability, and consistent reporting. This is the necessary foundation. Without it, AI hits a wall in almost any organization.
However, a data platform is inherently a layer above operational reality. It works with prepared views, projections, and "data products" optimized for analysis and decision support. This works perfectly for generating insights.
The problem arises when the goal is not an insight, but a result—meaning autonomous or semi-autonomous behavior within business processes.
The moment an AI is expected to act—execute a process step, create a record, change a status, make a binding decision, or communicate on behalf of a specific role—it must interact with Systems of Record. These are the systems that own the authoritative state, business rules, permissions, and audit trails.
This is where "AI-ready data" hits its limit. Having a clean data source is not enough. You must resolve the enterprise-level questions that determine the safety and trustworthiness of the entire solution:
These are not primarily "data engineering" questions. While data engineering can implement controls (contracts, observability, policy enforcement), the definition of authority, boundaries of Systems of Record, and decision rights is a matter of Architecture, Governance, and the Operating Model.
If it’s true that AI without AI-ready data often stalls at the MVP stage, then investment in D&A is mandatory. If it’s also true that governance fails without clear business anchoring, then using "AI" as a vague, universal motivation for D&A might result in a collection of "correct" activities that lack a connection to concrete results.
For autonomous AI, we must add a third dimension: Authority Readiness.
Authority Readiness is the organization’s ability to manage who (and for whom) can act, what they can access, how contexts are isolated, how everything is audited, and where the liability lies. Without this, pilots will continue to look impressive, but production deployment will be bogged down by risk management and "catching up" on architectural requirements that should have been clear from the start.