Connect an AI conversation to a form goal

Do not maintain a separate AI question list. Connect a landing AI CTA to a real Form. AI then uses the same custom fields and server validation as the visual form. Both surfaces create a normal Response.

What the connection reuses

When a form goal is active, the validated form submission replaces separate generic customer or reservation writes. One conversation retry converges on one response instead of creating duplicate operations.

Before you start

• Complete the public test in Create and publish a consultation form.
• Use only a Public form as an AI goal. Never pass a customer-bound Token link into chat.
• Prepare a synthetic customer scenario.
• Check the organization's availability rules for a reservation form.

1. Write the Form Prompt

Open the form in Forms. Put only operating context that fields cannot express in Prompt.

Help the customer state a clear consultation goal. Keep questions short and calm.
If the customer mentions both career and study topics, ask which one has priority.
Read the exact reservation date and time back. Submit only after explicit confirmation.

Do not copy field names, options, required rules, conditions, storage instructions, or access rules into Prompt. The form contract already supplies them. Do not put secrets or customer personal data in Prompt.

2. Connect the landing AI CTA

1. Open Sections.
2. Select the landing section that starts the consultation.
3. Set the Primary or Secondary CTA kind to AI.
4. Enter the localized CTA Label.
5. Search for and select the Consultation request Form.
6. Save. Select the button on the real landing page.

Form is a reference to a real form. Do not keep a copied slug in another free-text setting. A button-level form overrides the page-default goal. An AI CTA without a form starts a general organization conversation.

3. Test the conversation

1. Describe the consultation goal in natural language.
2. Check that AI does not ask again for a value already supplied.
3. Omit one required value. Check that AI asks for it.
4. Select Other. Check that AI asks the conditional follow-up.
5. Give an invalid option. Check that AI requests an allowed value.
6. Select a reservation slot. Check that AI reads back the exact time and asks for confirmation.
7. Check the pre-submit summary against the customer's answers.
8. Change the topic or decline. Check that AI does not pressure the customer to submit.
9. After completion, check that one response exists in Responses.

Do not accept a conversational success claim without the response. Compare Conversations and Responses.

4. Check the shared result

AI answers use the same table columns, saved queries, and custom.<key> workflow conditions as a visual form. You do not need an AI-only mapping layer.

Extend the integration

• Connect different forms to service-specific landing sections.
• Use Select and Multi Select values as operating categories and table filters.
• Measure final responses as conversion. Do not count only chat starts.
• Use custom.consultation_type in a workflow to assign an owner, send a notice, or change state.

Change the question in Custom fields. Use Prompt only to improve how AI completes the form goal.

Check the result

• The AI CTA opens a conversation with the selected public form as its goal.
• Questions, options, required rules, and conditions match the visual form.
• One Submitted response appears after customer confirmation.
• Customer and reservation values use the same owners as a visual submission.
• A conversation retry does not duplicate a customer, reservation, or response.