← Back to Field Notes

April 14, 2026 · The Bloomfield Team

Why Manufacturing Workers Are Right to Question AI

A PwC and National Association of Manufacturers survey published in April 2026 found that manufacturing workers are skeptical of AI adoption in their operations. PwC and NAM recommend that leaders offer meaningful training and demonstrate tangible benefits to move forward. Both recommendations are correct.

The earlier step is the one most manufacturers skip.

Workers who have lived through an ERP rollout that created six months of chaos, a scheduling tool abandoned after three weeks, or a quality dashboard that produced reports nobody read, have learned something specific: technology projects in their operation tend to arrive as mandates, not conversations. The skepticism is accurate. It is measuring something real.

What the Skepticism Is Actually Measuring

Worker resistance to AI is not resistance to the technology. It is resistance to the way the technology gets deployed.

A production manager learns about an AI project affecting her team through a hallway conversation. An estimator finds out a new quoting tool is live when someone sends him a login link. A maintenance technician is told to use an AI dashboard without being told what the system does with the information he enters. In each of these cases, the technology might be good. The implementation is a failure before it starts.

The PwC/NAM findings point to a structural pattern. Manufacturing leaders are deploying AI tools without involving the people who perform the work in the design process. The result is tools that solve the problem leadership thinks exists rather than the problem that actually slows things down.

Twenty years of management literature describes this gap. It does not close because the people with the budget to build the tools and the people with the knowledge of the actual work are rarely in the same room when decisions get made.

The Fix Is in the Design, Not the Convincing

The shops where AI adoption succeeds share a structural feature. The people who do the work were involved in defining the problem before anyone started building the tool. Their knowledge of what slows things down, what the workarounds are, where the process actually breaks, is treated as the primary input. The AI handles the friction they described.

When a worker sees her daily workflow reflected in a system that removes something she has been working around for three years, the skepticism dissolves. The resistance was never about the technology. The resistance was about being ignored.

Henry Kaiser built Liberty ships in 42 days during World War II by asking the welders what slowed them down. The technology available in 1943 was not exceptional. The process of surfacing and removing operational friction was. The principle has not changed.

The Question Worth Asking Before Training Anyone

PwC and NAM are right that meaningful training matters. Workers who understand what an AI system is doing with their inputs, why it surfaces certain information, and what happens when it gets something wrong, will use it differently than workers who received a login and a 30-minute onboarding session.

The question that comes before training: What would you build differently if you could? Ask it of the estimators, the schedulers, the setup specialists, the production managers. Document the answers. Build against the documented friction.

The workers who are skeptical are not the obstacle to AI adoption. They are the best source of information available on where the technology needs to go. The manufacturers who treat that skepticism as a signal rather than a barrier will build tools that get used. The ones who treat it as a communication problem will spend the next three years trying to convince people to adopt systems that do not match the work.

The survey is not a warning about workers. It is a warning about implementation design. The two things are different, and the distinction matters.