Formulate effective questions

To get optimal results from the AI Assistant, apply these tips when asking questions about your data.

Best practice: Be specific

Note that English is currently the best language to use for prompting the AI Assistant.

DON’T DO

  • Do not omit key details: Missing identifiers or missing context can confuse the agent.

  • Avoid overly broad queries and requests for large amounts of data.

  • Ask about particular part IDs, notification numbers, batch numbers, or time frames, etc.

  • Mention which IDs you are using.

  • If applicable: Always provide the product family.

  • If you would like an aggregation, make sure to request it explicitely.

The following examples illustrate the above tips.

Example: Use labeled IDs and specify the product family
Show me the production time for the product with processed part ID ABC1234567890 and material number 123456789A in product family A48Z.
Example: Request aggregations explicitely
Show me the total production quantities for plant ABC.
Example: Too broad
Tell me everything about quality claims.
Example: Too large amount of data
Show me all quality complaints for the XYZ product.

Limitations and troubleshooting

Learn what to do when responses aren’t helpful.

Known issue: No results for your question

Currently, the data backend service responds to some erroneous requests with "0 results found" instead of throwing an error. This can cause the AI Assistant to incorrectly state that there are no results for your question.

Constraint: No complex analyses

Complex statistical analyses are not supported.

For example, the following prompt will not work.

Example: Too complex analysis
Perform a logistic regression to find the best parameterization for my production line.

Troubleshooting: Improve insufficient responses

If you are still getting insufficient responses, try the following:

  • Start a new chat to reset memory and ask again (the AI gets slower/loses focus when context gets too large)

  • Try rephrasing with more details

  • Break complex questions into smaller pieces