What Let AI decide does
Let AI decide reads something from the run and chooses one answer from a list you provide. Then the workflow branches based on that answer. For example, it can tell an urgent complaint from a routine one, or a real question from spam.
Set it up
- What to decide: The question, like "How urgent is this complaint?"
- Choices: The answers it can give, separated by commas, like
urgent, normal, ignore. - What to read: The part of the run it looks at, like the message someone sent.
Then add a When way out for each choice that should lead somewhere, like data.<step id>.choice = "urgent", and an Otherwise way out for the rest.
What it saves
data.<step id>.choice: Its answer, spelled exactly as in your Choices.data.<step id>.reason: One sentence about why.
If the information doesn't clearly point to one answer, it's told to choose the most cautious option. If it gives an answer that isn't one of your choices, the step fails instead of guessing, and data.<step id>.ok is false.
It uses your account's monthly AI allowance, the same one the assistant uses.