You need intents so whatever code is running Lex knows to do something. Especially for a conversation string.

If you want text to speech or text to text? You'd make a polly utility, or do something like "open dictate" and then just capture everything said after.

That being said the duration is limited too.