Five Whys is a tool. 8D is a whole process around it: who works the problem, how you protect the customer while you work it, and how you make sure it stays fixed.
What it is
8D stands for Eight Disciplines. Ford Motor Company popularized it in the late 1980s, and it's now standard in automotive and much of manufacturing. Customers often require an 8D report when they find a defect.
The steps (many versions add a D0 for planning):
| Step | What happens |
|---|---|
| D1 | Form the team |
| D2 | Describe the problem precisely |
| D3 | Contain it: protect the customer now |
| D4 | Find the root cause, and why it wasn't caught |
| D5 | Choose permanent corrective actions |
| D6 | Implement them and verify they work |
| D7 | Prevent recurrence: update systems, documents, similar products |
| D8 | Recognize the team and close |
Two things make 8D stronger than a quick fix. D3 contains the problem before you understand it. And D4 asks two root causes: why it happened, and why your checks didn't catch it.
Why people skipped it
An 8D report takes many hours of writing on top of the actual problem solving. Teams rush the paperwork, or do a "paper 8D" after the fact to satisfy a customer.
How a bot does it
The bot keeps the 8D moving and drafts each section. The team does the thinking that needs people.
- D2. Have the bot draft the problem description from complaints and data, using is / is not. See Kepner-Tregoe.
- D3. Have it list everything that might be affected: lots, shipments, customers, other products. Then propose containment.
- D4. Have it draft a fishbone and 5 Whys from evidence, for both "why it happened" and "why it escaped."
- D5 to D7. Have it draft actions, and search for other products and processes with the same weakness.
- Throughout: have it keep the report current and flag steps that are overdue.
The bot brief
Draft an 8D report for this problem. Inputs: [complaint, data, records].
D2: problem description with an IS / IS NOT table. Cite records.
D3: list everything possibly affected (lots, dates, customers, related products)
and propose containment actions. Mark this URGENT for review.
D4: two root-cause analyses — occurrence (why it happened) and escape (why it
wasn't detected). Use fishbone categories and 5 Whys. Every "why" must cite
evidence or be marked HYPOTHESIS TO VERIFY.
D5-D6: propose corrective actions, each with how we'll verify it worked.
D7: search [other products/process files] for the same weakness and list them.
Leave D1 and D8 for the team. Mark every section DRAFT.Starting point only. Test it on your own data and model, and adjust as needed before you rely on it.
An example: a bot's own mistake
A bot emailed three customers a quote with an old price.
- D3 (contain): Pull the routine. Find every quote sent in the last 30 days using that price file. Call the three customers.
- D4 occurrence: The skill pointed to a folder, and a copy of last year's price list was in it.
- D4 escape: The person approving quotes checked totals, not unit prices against the current list.
- D5: The skill points to one named master file. Old lists are archived.
- D6: Re-run the last 20 quotes. All match the current list.
- D7: Two other skills point to folders, not files. Fix them too.
D7 is where 8D pays off. One problem fixed in three places.
Where bots differ
Containment (D3) is time-critical and needs authority: stopping shipments, calling customers. A bot can list what to contain. A person must decide and act.
And a bot can write a convincing root cause for a customer that nobody actually verified. Mark unverified causes as hypotheses until someone tests them.
What this means for you
- Use 8D for problems that reached a customer or could.
- Let a bot draft each section from evidence.
- Always do both root causes: occurrence and escape.
- Don't skip D7. Look for the same weakness elsewhere.