If your team cannot reliably see where production problems occur or understand what is causing them, establish better visibility before committing to an AI deployment. For plant managers, operations directors, and manufacturing technology leaders, the first question is whether clearer process information would help employees address the issue. AI may support a subsequent analytical or optimization task, but it should not replace the work of understanding the process.

Live Connected Manufacturing discussions emphasized a practical sequence: establish process discipline, capture useful data, and make performance visible before applying advanced analytics. Vijay Gurav, Senior Industrial Engineer at Boston Whaler, advised, "Pick one small paper-based or whiteboard process, [then] capture it digitally." He described building an issue-tracking application that recorded production problems, troubleshooting steps, and solutions. That information helped him identify which departments needed attention, while an agentic troubleshooting assistant remained a future goal.

In Connected Manufacturing Insights' 2026 report, The State of Connected Manufacturing, 27% of the 100 manufacturing leaders surveyed identified a lack of real-time visibility as their greatest challenge in converting data into actionable decisions. The finding reinforces the importance of assessing information gaps alongside opportunities for more advanced technology.

Start with one recurring problem, such as inconsistent output or repeated production interruptions, and ask frontline employees what they need to understand it. Capture the relevant process information, establish a baseline, and assign responsibility for responding to what the data reveals. Test whether better visibility and a clearer response process improve the outcome. Then evaluate AI against any remaining, clearly defined analytical need.

Join the upcoming Connected Manufacturing Forum to compare practitioner approaches to improving operational visibility and deciding where AI can add value.