An industrial AI discussion should begin with a useful operational question. Which decision needs better evidence? Who will act on the answer? What information can the team trust? We recommend treating the data foundation as part of the application itself: a defined scope, understandable context and clear ownership that can support both today’s operational decisions and future opportunities for improvement.
Start with the decision.
Choose a production question that a named team needs to answer. Where are losses occurring? Which conditions deserve investigation? How should a planner respond to a constraint? Write down the decision, the person responsible and the action that better information would help them take next.
Agree how the team will judge whether the proposed application is useful. Consider decision quality, investigation effort or the clarity of an operational handover. Establish a starting point using information you already have, then choose an initial scope that people can examine and discuss together.
Make the context visible.
Map the information needed for that question across equipment, production systems and enterprise records. Ask whether each signal has an understandable name, unit, source and owner. Include relevant operating context, such as the product being made, the production order or the equipment’s current operating state.
Review gaps with the people who know the operation. Which values are missing, delayed or manually entered? Which terms mean different things across departments? Record these limitations alongside the data. A useful deliverable is a shared map of what is available and what needs attention.
Build a first step people can review.
Start with an application that makes its inputs and purpose easy to inspect. This might be an operational dashboard, a structured loss review or an assisted investigation. Give users a way to challenge the information, flag exceptions and explain when the suggested interpretation needs revision.
Use the findings to shape the next stage. Decide which data improvements deserve priority, where integration is needed and who will maintain the resulting service. Expand when the operational team can explain the value and the limitations, with clear ownership for the decisions that follow.
Build the foundation around the decision you want to improve.