Not sure if this has been discussed or not but using the Data Schema to get structured data has been yielding semi innaccurate results, when it comes to names and emails. I have it so that my agent requests for the user to spell out their name, which made things a bit more accurate but It'll still get it wrong. does anyone have any tips or tricks to get the most accurate results. Agent information: Model: gpt-4o-mini transcriber: deepgram Nova-2-General I'm not entirely sure how the structured data schema stuff works under the hood so, I don't know if upgrading to 4o instead of 4o-mini would fix the mistakes or not, or if it could be fixed by prompting for the structured data? is there a generic structured data prompt that I could put for all my agents that would increase accuracy w/ consideration that the data is dynamic between all agents? all help is appreciated