AI Won't Solve Your Problems...But It Can Make Your Problem Solving a Lot Faster

I recently sat down with Rich Nave, quality systems expert and founder of The Luminous Group, for the third installment of our ongoing series on the Shop Floor, Top Floor Talk Show. This time, we got into one of the topics I’ve been wanting to dig into for a while: using AI for problem-solving and root cause analysis in manufacturing. Rich didn’t pull any punches. His opening was essentially a warning, and it was the right place to start.

Let’s Start with What AI Won’t Do

Rich was clear: “If you have this illusion that you’re gonna just take a customer complaint, throw it at some AI, whether it’s Claude or ChatGPT or whatever, and it’s gonna spit out an 8D for you, you’re kidding yourself.”

He’s right. AI will give you an 8D. It just might not give you a good one. And a rushed, unvalidated 8D sent to a customer is worse than a slow, accurate one. As I said in that conversation, you can make yourself look pretty foolish, pretty fast.

The point isn’t that AI is useless for problem-solving. It’s that AI doesn’t replace the engineering judgment, the leadership commitment to deploying resources, or the cross-functional work that actually closes problems. What it does is remove friction from that work. And that’s genuinely valuable if you know where to apply it.

Four Areas Where AI Is Actually Helping

1. Getting to a Better Problem Definition, Faster

One of the most common places teams stall in an 8D is D2, defining the problem. Rich sees this constantly. Teams get stuck on something vague like “bad parts” and burn time before the real work even begins.

If your AI tool has access to your scrap data, production reports, or quality records, it can cross-reference the customer complaint against what’s already in those systems and generate a working problem definition in minutes. Even better, have the AI give you five versions of the problem statement and pick the two that are closest. That iteration is fast, and it surfaces angles your team might have missed.

“It’s really easy with AI to not just ask for a problem definition, but ask it to give you five versions of the problem definition. Then you can look at what were some of the different angles it looked at and came up with, and maybe pick numbers two and three and merge those together.”

2. Making Sense of Big Data

Back around 2015, Rich was working with a company that was already collecting over 1,000 data points on every axle they manufactured. Today? Probably 10,000. That’s not an edge case anymore, that’s manufacturing reality.

When you hit an end-of-line failure or a customer complaint, which data point is the one that matters? AI is very good at finding correlation across massive datasets, surfacing which variables are moving together when defects appear. That narrows the field significantly so your engineers can focus on the right suspects.

The important caveat, and Rich made it well: correlation isn’t causation. The engineering judgment that determines which correlation actually caused the problem, that still belongs to your team.

For manufacturers with multi-site operations, this is especially powerful. Solutions found in Europe don’t automatically make it to North America. AI changes that equation.

4. Cascading 8D Outcomes Into Your Quality System

This was the one that really caught my attention. Once you’ve closed an 8D, the findings need to flow into your PFMEA, your control plan, your training, your LPAs, and possibly your standard work instructions. That documentation work can take the better part of a day.

Rich’s team fed the 8D and PFMEA into the AI and asked: where does the PFMEA need to be updated based on this 8D? The AI highlighted the rows it changed in yellow so the team could quickly review and verify. Then they did the same for the control plan, and generated five LPA questions to reinforce the corrective action on the floor: two needed to be cut, three were usable immediately.

“What might take a day of work to implement those four or five documents, we did it in 20 minutes, maybe 30. You are taking what might be six or eight hours of work and condensing that down to under an hour.”

One Thing to Keep Front of Mind: Verify Everything

Rich shared an example that’s worth putting in a frame somewhere: he once reviewed an FMEA, written by humans, no AI involved, that listed “visual inspection for microscopic cracks” as a detection method. Microscopic. Visual inspection.

That’s the kind of logical gap that AI can also produce, and often in more subtle ways. His recommendation: whenever you have AI update a document, ask it to highlight every change. That way your team can do a rapid human review of exactly what moved, rather than reading the whole document from scratch.

Build the verification step into your standard work. AI generates; humans approve.

Where to Actually Start

If I were advising a quality leader looking to bring AI into problem-solving, I’d say this:

• Make sure your AI has access to the right data, scrap reports, production records, historical 8Ds. If the data isn’t accessible, the AI is working blind.
• Don’t expect results in week one. The first time you use AI for a problem definition or an FMEA update, it’ll get you 40% of the way there. The thirtieth time, it’ll be much closer to 80%.
• Start with one application, not a platform. Pick your highest-friction task (usually data retrieval or documentation) and prove value there before expanding.
• Keep engineering judgment in the loop at every decision point. AI surfaces patterns and drafts documents. Your team makes the calls.

Listen to the full episode with Rich Nave on the Shop Floor, Top Floor Talk Show and if you want to see how EASE helps your team move from raw data to real decisions faster, take a look at what we’ve built.