SA Schreiber AI

AI for root cause analysis: faster 5 Whys, fishbone and 8D, without fooling yourself

AI is good at the parts of problem-solving that are tedious: finding similar past failures, organising evidence and drafting the report. It's bad at the part that matters most: going to see and proving the cause. Use it for the first, never for the second.

By Ben Schreiber · Schreiber AI ·

The short answer

Feed AI your history (NCRs, 8Ds, maintenance logs, customer complaints) and it can quickly find similar past problems, suggest fishbone categories and candidate causes to check, and turn your notes into a clean 8D or A3. The team still goes to the gemba, tests the hypotheses and verifies the root cause with data. An AI-generated "root cause" nobody verified is just a well-written guess.

Where AI speeds things up

  • "Have we seen this before?" Searching years of reports in seconds, including ones described in different words.
  • Brainstorm prompts. Suggesting causes across Man, Machine, Method, Material, Measurement and Environment, so the team doesn't anchor on the first idea.
  • Structuring the 5 Whys. Flagging when a "why" jumps to a conclusion or blames a person instead of a process.
  • Drafting the 8D or A3 from notes, photos and data, in your template.
  • Tracking actions and reminding owners until containment, corrective and preventive actions close.

Where it misleads, and how to guard against it

  • Confident but wrong. Language models produce plausible explanations. Treat every suggestion as a hypothesis to test.
  • Garbage history. If past reports say "operator error" every time, AI will suggest that too. Clean up how you close reports.
  • Skipping the gemba. The fastest way to a wrong root cause is analysing from a desk. AI makes desk analysis tempting; don't let it.

Worked example

A cracked housing shows up at a customer. Searching 4 years of NCRs by hand would take a day. AI finds three similar cracks in 20 minutes, all after a mould maintenance event. The team goes and checks, confirms a cooling-line issue after maintenance, and adds a post-maintenance first-article check. The AI didn't find the root cause; it pointed the team at the right place to look, a day sooner.

A simple way to start

  1. Export your last few years of NCRs, 8Ds or complaint records into one folder.
  2. Use AI search over them the next time a problem hits. Did it surface something useful?
  3. Add AI drafting to your 8D template, with a required "verified by data" field before closure.

Common questions

Can AI find the root cause of a quality problem?

It can suggest likely causes and point to similar past failures, but a root cause must be verified with data and observation on the floor. AI speeds up the search; people prove the cause.

Which problem-solving methods does AI work with?

Any structured method: 5 Whys, fishbone (Ishikawa), 8D, A3 and DMAIC. It helps most with searching history, organising evidence and drafting reports.

Is it safe to put quality records into an AI tool?

Use a business-grade tool that doesn't train on your data, and check customer agreements before sharing customer-specific information.