Moving AI from pilot to production starts with data

Moving artificial intelligence (AI) from pilot projects into production takes more than a capable model. For marketing teams, it also requires data that AI can interpret consistently across platforms and internal systems.

McKinsey’s 2025 global survey found that 88% of respondents said their organizations regularly used AI tools in at least one business function, up from 78% the previous year. Yet only 7% said AI was fully scaled across their organization, while 30% remained in the piloting phase.

Pilots often work with a narrow task and carefully selected data. In production, an AI system may encounter multiple versions of a metric, with different field names and calculation rules. Without shared definitions, it may choose a plausible but incorrect match. Its answer can sound convincing even when it rests on a mistaken assumption.

Data quality is only part of the challenge. AI also needs the organization’s definitions for key concepts, mappings between platforms, and rules for measuring performance. A knowledge layer can provide that context on top of an existing data warehouse, without requiring infrastructure replacement.

Some context changes with each investigation. A drop in campaign performance, for example, may coincide with a budget change, promotion, or live market test. AI may detect the change correctly but misread its cause if it lacks that information.

The article argues that marketing and data teams should work together on definitions and governance before those rules enter automated workflows. For organizations seeking measurable value from AI, the focus should include whether models have reliable data and the context needed to interpret it.