Skip to content

AI Field Intelligence

AI Field Intelligence makes your fields smarter. Instead of accepting whatever someone types, AI can ask follow-up questions, validate against rules, and extract information from documents.

Without AI: Someone types “car accident” in your incident description field. You have to follow up manually to get details.

With AI: The AI asks “Where and when did this happen? Were there injuries? What’s the damage?” until it has a complete picture, automatically.


AI asks follow-up questions until it gets a complete answer.

Best for:

  • Description fields
  • Complex information
  • Qualitative data

AI validates answers against specific rules.

Best for:

  • Format requirements
  • Specific criteria
  • Structured data

AI reads uploaded documents and extracts information.

Best for:

  • Resumes → Name, experience, skills
  • IDs → Name, DOB, address
  • Invoices → Amount, dates, line items

Socratic Negotiation is perfect for fields where you need complete, detailed answers.

This works best with:

  • Textarea fields (long descriptions)
  • Text fields (shorter responses)
  1. Click the field to edit it
  2. Find AI Intelligence in the settings panel
  3. Click to expand

Toggle Socratic Negotiation to On.

Tell the AI what a “complete” answer looks like. Be specific!

Example goals:

FieldGoal
Accident Description”Include: date/time, location, parties involved, what happened, damage to vehicles, injuries, and any witnesses”
Project Requirements”Include: project goals, target audience, desired features, timeline, and budget constraints”
Job Experience”Include: job title, company name, duration, main responsibilities, and key achievements”
Support Issue”Include: what happened, when it started, steps to reproduce, what they’ve already tried, and error messages”

How persistent should the AI be?

LevelBehavior
LowAsks 1-2 follow-ups, then accepts
MediumAsks until most criteria are met
HighInsists on complete information

Tip: Start with Medium. Adjust based on feedback.

Field: “Describe the accident”
Goal: “Include date/time, location, parties involved, what happened, damage, and witnesses”

IntakeBot: Please describe what happened.
User: I was in a car accident.
AI: I need a few more details. When and where did the accident happen?
User: Yesterday afternoon on Main Street.
AI: What exactly happened, and what damage was there?
User: A truck ran a red light and hit my driver's side door.
My door is dented and won't open properly.
AI: Were there any other parties, injuries, or witnesses?
User: No injuries, just me and the truck driver.
A pedestrian saw it happen.
AI: Thank you, I have all the details I need.

The AI continues until it has a complete picture, no manual follow-up needed.


Strict Validation ensures answers meet specific criteria before accepting them.

Works with any field type.

  1. Open AI Intelligence settings
  2. Toggle Strict Validation to On

Tell the AI what makes an answer valid.

Example rules:

FieldValidation Rule
Policy Number”Must be format: two letters, dash, six digits, dash, two letters (e.g., AB-123456-CD)“
Phone Number”Must be a valid US phone number with area code”
Date”Must be within the last 30 days”
Email”Must be a company email, not personal (no gmail, yahoo, hotmail)“
URL”Must be a working website URL starting with https”

Field: “Enter your policy number”
Rule: “Format: XX-123456-YY”

IntakeBot: What's your policy number?
User: 123456
AI: Policy numbers should be in the format XX-123456-YY
(two letters, dash, six digits, dash, two letters).
Please check and try again.
User: AB-123456-CD
AI: Got it, thank you!

Let AI read uploaded documents and pre-fill fields.

  1. Someone uploads a document (resume, ID, invoice)
  2. AI analyzes the document using Vision
  3. Relevant fields are pre-filled
  4. User confirms or edits the extracted data
  1. Add a File field to your IntakeBot
  2. Configure allowed types (PDF, images, etc.)
  3. Give it a descriptive label like “Upload your resume”

Add the fields where you want extracted data to go:

  • Name (text)
  • Email (email)
  • Phone (phone)
  • Years of Experience (number)
  • Skills (multi-select or text)
  1. On each target field, open AI Intelligence
  2. Enable Extract from documents
  3. Specify what to look for

Example extraction prompts:

FieldExtraction Prompt
Name”Extract the person’s full name”
Email”Extract the email address”
Years of Experience”Calculate total years of work experience”
Skills”Extract technical skills and tools mentioned”
  1. User uploads their resume PDF
  2. AI analyzes: “John Smith, [email protected], 8 years experience in marketing, skills: SEO, Google Ads, Analytics”
  3. Name, email, experience, and skills fields are pre-filled
  4. User reviews and confirms

Result: What used to require manual data entry now happens automatically.


Some information should never be processed by AI.

Mark fields as Private when they contain:

  • Social Security Numbers
  • Medical information
  • Financial account numbers
  • Passwords or credentials
  • Other sensitive PII

How to enable:

  1. Edit the field
  2. Open Privacy settings
  3. Enable Private Field

What happens:

  • Data is stored separately
  • AI never sees or processes it
  • Only workspace admins can view
  • Field is marked with a lock icon
Data TypeUse Private?
Name, email, phoneNo (standard PII is fine)
Social Security NumberYes
Medical conditionsYes
Credit card numbersYes
Login credentialsYes
Salary informationConsider it

Important: Private fields cannot use AI Intelligence features (no Socratic mode, no extraction). This is by design, the AI never sees this data.


Field: “Describe the incident”

Socratic Goal:

“A complete incident description includes: date and time, location (address or intersection), what happened, parties involved (names if known), damage to property, injuries if any, witness information, and police report number if available.”

Strictness: High (insurance claims need details)

Field: “Tell us about your project”

Socratic Goal:

“Include: what you’re trying to accomplish, who the target audience is, what success looks like, timeline expectations, and any constraints or requirements.”

Strictness: Medium (creative projects can be more flexible)

Field: “Describe your issue”

Socratic Goal:

“Include: what’s happening, when it started, steps to reproduce, what you’ve already tried, and any error messages you’ve seen.”

Strictness: Medium (balance detail with user frustration)

Field: “Why are you interested in this role?”

Socratic Goal:

“Response should mention: specific interest in the company or mission, relevant experience, and what they hope to contribute.”

Strictness: Low (don’t interrogate candidates)


❌ “Get a complete answer”
✅ “Include: date, location, parties involved, damage assessment, and witnesses”

  • High: Legal, insurance, compliance
  • Medium: Business intake, support
  • Low: Feedback, casual inquiries

Before publishing, test your AI settings with realistic answers, both complete and incomplete, to see how the AI responds.

Simple fields (email, phone, dropdown selections) don’t need AI intelligence. Reserve it for fields where incomplete answers are actually a problem.

Use conditional fields to show AI-enabled fields only when needed. No point using AI credits on a follow-up question that wasn’t triggered.


AI Intelligence uses your AI credits:

  • Each Socratic exchange uses credits
  • Vision document extraction uses credits
  • Strict validation uses fewer credits

Monitor usage in SettingsUsage.


  • Verify Socratic Negotiation is enabled
  • Check that you’ve defined a goal
  • Ensure the field type supports AI (text/textarea)
  • Verify you have AI credits remaining
  • Lower the strictness level
  • Simplify your goal (fewer required elements)
  • Add “nice to have” vs “required” distinction in goal
  • Check file type is supported
  • Ensure document is readable (not scanned poorly)
  • Verify extraction is enabled on target fields
  • Check Vision credits balance