Why does my AI prompt condition treat capability questions as feature requests?
Because the model focuses on the positive yes-case and under-weights any exclusion written inside it. Move exclusions into a separate Don't match for clause and the condition becomes reliable.
The problem
You write a workflow If/Else AI prompt condition whose yes-case includes an inline exclusion, for example that capability questions like "is there a way to do X" should not match. Those exact questions still get classified as yes.
This is a prompt writing failure, not a bug. A negation inside the yes-case competes with the positive picture the model builds from it, so a message that resembles the core intent still tips toward yes.
Separate the yes and no cases
Use the pattern Match if [yes criteria]. Don't match for [no criteria]. Define both sides explicitly, using positive statements where possible. List specific examples of what should not match, especially ones close in meaning to the yes-case. Avoid double negatives and nested logic.
Before, unreliable:
Answer yes if the message is a feature request or product feedback. It DOES NOT include questions about the product which could also be feature requests, such as "am I able to do X" or "is there a way to do Y".
After, reliable:
Match if the customer is explicitly requesting something the product does not do: a new feature, an improvement or a product wish. The intent must be clearly to suggest a change in the product.
Don't match for:
- Capability or availability questions such as "is there a way to...", "am I able to...", "can I..." or "does X support..."
- Bug reports or anything where the product is expected to already work
- How-to or setup questions
- Billing or account issues