top of page
Mad-Ai Logo.png
Search

Why AI Deserves the Same Scrutiny as a New Gage

Writer: Rick Heilshorn
Rick Heilshorn
12 hours ago
2 min read

Quality & AI | Wednesday, October 7, 2026

Manufacturing quality engineers are used to dealing with variation.


When a new gage, inspection method, measurement system, or process is introduced, we don't simply ask whether it works. We ask how much variation it introduces.


That discipline exists for a reason.


If three trained inspectors evaluate the same 30 parts using the same standard and agree only 36.67% of the time, that is not something we'd shrug off. We'd investigate the measurement system and determine whether the variation was coming from the people, the method, the equipment, or some combination of the three.


We understand that unmeasured variation creates risk.


But there is a part of modern quality work where we are starting to behave differently: engineering judgment assisted by AI.


Consider the work engineers already do every day:

  • Grading an 8D

  • Reviewing a PFMEA

  • Evaluating a SCAR

  • Assessing a nonconformance

  • Reviewing corrective-action logic

  • Evaluating whether a disposition is adequately supported


These are not simple data-entry tasks. They involve judgment. And human judgment varies.


AI does not automatically remove that variation. In some cases, it introduces another variable: the person interacting with the AI.


Give an engineer a general-purpose AI chat window and ask it to evaluate an 8D.

  • What information does the engineer provide?

  • What evidence do they attach?

  • What do they leave out?

  • What questions do they ask?

  • What do they challenge?

  • And at what point do they decide the answer is good enough?


The AI may be the same. The process is not.


One person may provide the complete 8D, the relevant customer requirements, supporting evidence, and detailed context.


Another may paste a few paragraphs and ask, "Is this 8D good?"


Both can receive a polished answer.


That is a process problem.


In manufacturing, we'd recognize an uncontrolled input immediately. We would not expect a measurement system to be trustworthy if the inputs changed arbitrarily from one appraiser to another without understanding the effect.


AI-assisted quality work deserves the same scrutiny.

The question is not whether AI is useful. It clearly can be.


The question is whether the process surrounding the AI is controlled enough for the work we are asking it to perform.


That means asking the same questions we already ask elsewhere in quality:

  • What are the required inputs?

  • What evidence is required?

  • What criteria are being applied?

  • How repeatable is the result?

  • How reproducible is it between people?

  • And what happens when the system gets something wrong?


AI should not be exempt from the discipline manufacturing already applies to every other source of variation.


If anything, the fact that AI can produce a confident, finished-looking answer makes that discipline more important.


The first step is recognizing that the AI interaction itself is part of the process.




 
 
 

Comments

bottom of page