AI and Factory Automation: Plant Managers on the Return-on-Investment Gap

On 15 October 2026, Manufacturing Dive published an article by Sakshi Udavant collecting the views of operations leaders at companies including Pella and Hexion on the boom in artificial intelligence and automation. The opening message is blunt: most manufacturers have tried AI and automation in some form, but not all have yet achieved a good return on investment.

A note on the limits of this summary: the version of the page we reviewed was cut off after the opening section. We could not read the comments from Pella, Hexion and the other companies' leaders, and no survey data surfaced. What follows is based only on the readable portion.

AI as access to expertise, not just automation

In the accessible part of the article, Jared Pfeiffer, principal client partner at Augury and a former plant manager, makes his case. His argument is that industrial AI goes beyond automation because it widens access to expertise.

Plants, he observes, have historically relied on the experience of people with many years on the job. As that generation retires and labor shortages persist, AI can help fill the knowledge gap. According to Pfeiffer, it can combine sensor data, operating conditions, maintenance history, standard operating procedures (SOPs), manuals and company knowledge, helping a less experienced operator understand what is happening and where to intervene.

The benefits he lists for plant managers are:

• shorter downtime;
• faster fault diagnosis;
• greater consistency from one shift to the next;
• less dependence on a few experts.

Why it matters for those running a plant

The return gap is a useful reality check for plant and operations directors evaluating AI projects. Experimenting has become easy; turning experiments into measurable production results is another matter.

It is also worth remembering that Pfeiffer works for a solution vendor. His statements are the view of someone selling technology, not independent evidence, and should be read as such. Still, the way he frames the problem is interesting: AI as a tool to distribute expertise, not to replace people.

AI that is useful to shop-floor teams depends on the data it receives.

The prerequisite: shop-floor data in order

The benefits described assume that machine signals, maintenance records and procedures are available, connected and reliable. If sensor data is not linked to the production order, if fault history is incomplete or procedures are scattered documents, even a sophisticated tool will have little to work with. This is where machine connectivity and manufacturing execution systems (MES) become the foundation, rather than a side concern, of any AI project.

Practical guidance

Before launching or relaunching an AI project, it may be useful to check:

• which concrete problem you want to solve (downtime, diagnosis time, shift-to-shift variability) and which indicator will measure it;
• whether the necessary data, from machines to maintenance, is collected continuously and with clear context;
• whether procedures and expert knowledge are documented and up to date;
• who on the floor will actually use the output, and at what point in the shift;
• which vendor claims are backed by data from your own plant.

In short, the signal from the 15 October 2026 story is a cautious one: interest in AI is high, but returns are not guaranteed. For now, those getting the most value appear to be those who start with orderly data and a precise operational goal.