Why didn't a presumed use case get triggered when an advanced AI agent identified a new use case mid conversation?

Question

Why didn't a presumed use case get triggered when an advanced AI agent identified a new use case mid-conversation?

Answer

In a conversation, the advanced AI agent may detect a new use case based on the user’s input. That might suggest an immediate switch to the newly identified use case.

A presumed use case is a potential or predicted path that the AI agent identifies from the user’s intent or behavior, but it isn’t confirmed or active yet. It shows what the AI expects the user might need next and lets the system prepare an appropriate response or transition if required.

However, a presumed use case takes over mid-conversation only in these scenarios:

• The conversation reaches an intent predicted block within the current dialogue
• The current dialogue flow can’t continue with the user’s input

If the current dialogue has a valid way to handle the user’s response, the AI agent remains in that dialogue. If no path exists to proceed in the current dialogue, the AI agent activates the newly predicted use case.

Here is an example of how that works

Let’s say the user first reports login issues. This triggers the Login Issues use case, and the AI agent offers options and dialogue branches tailored to login-related concerns.

Midway, the user types "I need help with a return".

At this point, the AI agent predicts a new use case: Return process and issues.

However, it doesn’t switch because the current Login Issues dialogue includes a conditional branch that handles free-text input.

Even though the user’s message relates to returns, the AI agent stays within the Login Issues reply because it can still process the user’s input within that flow.

If you remove the free text branch in the Login Issues dialogue, the AI agent no longer has a path to continue when the user says “I need help with a return.” In that case, the current dialogue breaks. The AI agent recognizes this and activates the presumed Return process and issues use case instead and redirects the conversation.

In summary:

• Presumed use cases detected mid-conversation don’t always interrupt the current flow
• If the current dialogue can handle the user’s input with a valid branch, it stays active
• If it can’t proceed, the bot switches to the new predicted use case