What is the difference between Ask and Save and Extract Dynamic Variables, and which should I use?

Use Ask and Save for structured details such as name, email and phone number. It is the most common node when building agents. The AI still asks the questions conversationally, on the phone or by text message, and saves each answer to the contact in real time. It has built-in validation (a valid email, a valid phone number), and you can add custom questions with different data types, such as a text field, yes or no, or a checkbox. For a custom question you choose which contact field to save it to, a built-in field or a new custom field, so no extra step is needed. It is good practice to keep the answers on the contact even if you also send them elsewhere.

Use Extract Dynamic Variables when the conversation is not linear. It does not ask questions. It lets the AI look back at the conversation and pull data out of it, and it can run during a live call or at the end. It is best for things you would not ask out loud, such as how interested the buyer is on a scale of 1 to 10, and for messy conversations such as a restaurant order where callers ask about hours, ingredients, gluten free options and specials while ordering. For a restaurant, name a variable such as order items, tell it to list all the items the caller wants to order, and run it after the caller confirms the order is complete.

We do not recommend Extract for structured data like emails and phone numbers: if the caller gives a bad email or number you will not know until you try to extract it, and you might lose it.

If you are not using the workflow builder, the equivalent is the Ask and Save setting on the agent, where you list the questions. The three most common steps when building agents are Ask and Save, Bookable Service and Smart Calendar Booking.