Goldilocks Docs
AI Agents

Configuring Persona Behavior

Fine-tune how your AI persona communicates and responds

Persona configuration controls how your AI agent behaves in conversations. These settings affect tone, response style, and escalation behavior.

Accessing Configuration

  1. Go to AI Agents > Personas
  2. Click on a persona
  3. Navigate to the Configuration or Behavior tab

Universal Settings

These settings apply to all persona types.

Tone

Controls the overall communication style:

OptionDescription
FormalProfessional language, no contractions
ProfessionalClear and business-appropriate
FriendlyWarm and conversational
CasualRelaxed and informal
EmpatheticFeelings-first, understanding

Response Length

Controls how detailed responses are:

OptionDescription
ConciseShort, direct answers
StandardBalanced detail
DetailedComprehensive explanations

Greeting Style

How the AI greets contacts:

OptionDescription
Use customer namePersonalized greeting when name is available
Generic greetingStandard welcome
Skip greetingJump straight to helping

Language Complexity

Vocabulary and sentence structure:

OptionDescription
SimpleEasy-to-understand language
StandardGeneral audience
TechnicalIndustry terminology OK

Emoji Usage

Whether emojis appear in responses:

OptionDescription
NoneNo emojis
MinimalOccasional, strategic use
ModerateRegular friendly use

Sign-Off Style

How the AI ends messages:

OptionDescription
FormalProfessional closing
FriendlyWarm closing
NoneNo closing phrase

Support-Specific Settings

These appear for support-focused personas (in the Advanced Settings section).

Empathy Level

How much empathy the AI expresses before providing solutions:

OptionDescription
High (Feelings First)Always acknowledge emotions first
BalancedEmpathy when warranted
Solution-FocusedPrioritize fixing the issue

Escalation Threshold

When to offer human support:

OptionDescription
Patient (Try Harder)Exhaust options before escalating
StandardBalanced escalation timing
Quick HandoffEscalate early when uncertain

Apology Behavior

How the AI handles apologies when things go wrong:

OptionDescription
ProactiveApologize readily
Reactive OnlyApologize when clearly at fault
MinimalRarely apologize

Clarification Style

How the AI handles ambiguous questions:

OptionDescription
Ask FirstClarify before answering
BalancedTry then clarify if needed
Assume and ProceedBest guess, move forward

Follow-up Prompts

Whether the AI suggests next steps after answering:

OptionDescription
AlwaysAlways suggest next steps
When RelevantSuggest next steps when helpful
NeverLet customer lead

Policy Stance

How strictly to enforce documented policies:

OptionDescription
Flexible (Customer-First)Bend rules when reasonable
BalancedFollow policy with judgment
Strict (Policy-First)Enforce policies consistently

Sales-Specific Settings

These appear for sales-focused personas (in the Advanced Settings section).

Recommendation Style

How product recommendations are made:

OptionDescription
ConsultativeAsk needs, then recommend
DirectRecommend immediately
PassiveOnly recommend when asked

Upsell Approach

How to handle upselling:

OptionDescription
SubtleGentle, natural mentions
ModerateClear but not pushy
DirectActively promote upgrades

Urgency Language

Use of urgency in messaging:

OptionDescription
NoneNo urgency tactics
SoftFactual availability info
StrongEmphasize limited time/stock

Discount Behavior

How the AI handles discounts and promotions:

OptionDescription
Never MentionDon't discuss discounts
On RequestOnly when customer asks
ProactiveActively share deals

Escalation Settings

Configure how escalation works for this persona.

Escalation Email

Where escalation requests are sent:

support@yourcompany.com

Include Transcript

Whether to include conversation history in escalation emails:

  • Yes - Full transcript included
  • No - Just the escalation request

Widget Text Customization

Customize widget messages for this persona:

Greeting

Initial message shown:

Hi there! 👋 How can I help you today?

Resolution Prompt

Shown after providing an answer:

Did that answer your question?

Resume Message

When returning to a conversation:

Welcome back! Would you like to continue where we left off?

Escalation Offer

When offering human support:

Would you like me to connect you with a support specialist?

Input Placeholder

Chat input placeholder text:

Type your question...

Saving Changes

After configuring:

  1. Review all settings
  2. Click Save
  3. Test in the widget preview

Changes apply immediately. Test in preview before expecting them in production.

Best Practices

Match Your Brand

Configuration should reflect your brand voice:

  • Formal brand? Formal or professional tone, no emojis
  • Playful brand? Friendly tone, minimal or moderate emojis

Consider Your Audience

  • B2B? More professional
  • B2C? Can be more casual
  • Technical product? Higher language complexity

Test Different Combinations

Try different configurations and monitor results:

  • Resolution rates
  • Contact satisfaction
  • Escalation rates

Document Your Choices

Note why you chose specific settings. Helps when reviewing later.