Skip to main content
Contact Intelligence is what makes Outband fundamentally different from other communication tools. Instead of treating every email generically, Outband builds a unique AI model for each person you communicate with.

What Is a Contact Model?

A Contact Model is a persistent, evolving AI profile of a specific contact. It captures:
  • Communication style — How they write (formal, casual, terse, detailed)
  • Response patterns — How quickly they typically reply, at what times
  • Intent patterns — What they usually ask for and how
  • Relationship dynamics — Your history, shared context, and rapport
  • Cultural context — Language preferences, honorific levels (for Japanese communication), formality expectations

Viewing Your Contacts

Select Contacts in the sidebar to see all contacts that Outband has built models for. Each contact card shows:
  • Name and email with their primary channel
  • Confidence score — How well Outband understands this contact (based on data volume, recency, and consistency)
  • Trust level — Your current trust setting for this contact
  • Last interaction — When you last communicated
  • Key insights — A brief summary of notable patterns

Exploring a Contact Model

Select any contact to see their full Contact Model:

Overview Tab

  • Communication style summary
  • Typical response time
  • Preferred channels
  • Relationship strength indicator

Insights Tab

  • Common topics you discuss
  • Their decision-making patterns
  • Sensitivity areas (topics they respond strongly to)
  • Cultural context notes

Timeline Tab

  • History of your interactions across all connected channels
  • Shows how the relationship has evolved over time
  • Key moments flagged by the AI

Corrections Tab

  • History of corrections you have made to Outband’s understanding of this contact
  • Each correction improves future interactions

Confidence Scoring

Every Contact Model has a confidence score based on four dimensions:
DimensionWhat it measures
Data VolumeHow many messages Outband has analyzed from this contact
RecencyHow recent the latest interactions are
ConsistencyHow consistent the contact’s patterns are
Context MatchHow well past patterns predict current behavior
A low confidence score does not mean Outband’s analysis is wrong — it means Outband does not have enough data yet. Confidence naturally increases as you communicate more with a contact.

Personal DNA

Your Personal DNA is the Contact Model that represents you. It captures your own communication style, tone, vocabulary, and preferences. To view and refine your Personal DNA:
  1. Go to Settings > Personal DNA
  2. Review Outband’s understanding of your communication style
  3. Make corrections where needed — these directly improve draft quality
Your Personal DNA is used every time Outband generates a draft, ensuring the output sounds like you.

Correcting the AI

When Outband gets something wrong about a contact:
  1. Open the contact’s model
  2. Select Correct next to the insight you want to fix
  3. Provide the correct information
  4. Outband updates the model and uses your correction to improve future analysis
Corrections feed into the RLHF (Reinforcement Learning from Human Feedback) pipeline. Every correction you make improves not just this contact’s model, but Outband’s overall understanding of communication patterns.

Relationship Graph

Select Contacts > Relationship Graph to see a visual map of your communication network:
  • Connections between contacts show shared threads and topics
  • Line thickness indicates communication frequency
  • Color indicates relationship strength and trust level
  • Clusters reveal natural groupings in your network

Best Practices

Your most important contacts should be marked as VIPs from the Contacts page. This ensures their models are built first and they are never auto-handled.
For contacts you communicate with most — your CEO, key clients, direct reports — periodically review their Contact Model to ensure accuracy.
The best time to correct Outband is immediately after you notice an error. The correction is more precise when the context is fresh.

Next Steps