Create AI Labeling

What is AI Labeling ?

AI Labeling automatically applies labels to incoming emails and messages by reading and understanding the intent of each message — no manual tagging needed.

Steps

1. Go to Settings

2. Click on the AI Labeling icon

   

3. You will see a toggle:

   • Enable/disable AI labeling.

   Turn it ON.

   1. Select an AI Provider

4. Select an AI Provider

   An AI Provider is the service that supplies the intelligence behind AI Labeling.

   You are not creating an AI — you are connecting an existing AI model to your inbox. AI Labeling is the feature; the AI Provider is what actually understands and classifies the messages.

   OpenAI Requires an OpenAI API key. Recommended model: GPT-4o Mini — best for fast, accurate, and cost-efficient labeling.

   OpenRouter Requires an OpenRouter API key. Recommended models: Claude 3.5 Sonnet or Mistral Large — optimized for intent detection and classification.

   

5. Select the AI Model

   • OpenAI → Select GPT-4o Mini (recommended)

   • OpenRouter → Select Claude 3.5 Sonnet or Mistral Large

   These models are optimized for intent detection and classification.

6. Choose What You Want to Label

   Toggle the options based on your needs:

   Label Old Replies

   Keep this toggle ON to apply AI labels to existing messages already in your inbox — not just new ones going forward.

   Relabel Ongoing Conversations

   Keep this toggle ON to update labels automatically as conversations evolve. Useful when a prospect's intent changes over time.

7. Use Your Custom Prompt (Optional)

   You can define your own labeling rules and logic by creating a custom prompt. This step is completely optional — you can toggle it off if you don't need it.

   Master Inbox already comes with a default labeling prompt that works out of the box for standard email workflows. You only need a custom prompt if your workflow has specific requirements — such as LinkedIn outreach or specialized labeling logic.

   The Mental Model

   We have already added here our mental model: this is the way you think about and organize your labeling rules. Essentially, it’s the framework you use to decide which messages get which labels, helping MasterInbox automatically categorize conversations in the way that makes sense for your team.

When to use a custom prompt:

• You want labels applied based on your internal workflows

• Your agency uses custom terminology (e.g. Qualified, Demo Booked, Not ICP)

• You're running LinkedIn outreach with different intent patterns

• Different clients need different labeling logic

  Add a Custom Labeling Prompt (Optional)

  To get accurate labeling, your prompt needs to cover five things clearly. The clearer your instructions, the more accurate the labeling will be.

  The Task

  What should the AI do? Example: Analyze each incoming message and assign ONE clear, concise label based on the sender's intent.

  Allowed Labels

  List the exact labels the AI is permitted to use. It must not invent new ones outside this list.

  The Rules

  One label only. No explanations. No new labels invented. If intent is unclear, choose the closest reasonable match. Output only the label text.

  Context to Consider

  Message content, tone and urgency, whether the sender is new or part of an existing conversation, and any questions, requests, or actions being asked.

  Out Format

  Label only. No punctuation. No extra text. Clean output that maps directly to your label list.

  AI Labeling Prompt Example

  You are an AI assistant responsible for labeling inbox conversations.

  Your task:

  Analyze each incoming message and assign ONE clear, concise label based on the sender’s intent.

  Rules:

  • Choose the single most relevant label.

  • Labels must be short (1–3 words).

  • Do not invent new labels outside the allowed categories.

  • If the intent is unclear, choose the closest reasonable label.

  • Do NOT explain your reasoning.

  • Output ONLY the label text.

  Use these labels:

  • Lead

  • Customer

  • Support

  • Billing

  • Sales

  • Cold Outreach

  • Partnership

  • Spam

  • Follow Up

  • Internal

  • Feedback

  • Other

  Context to consider:

  • Message content

  • Tone and urgency

  • Whether the sender is new or part of an existing conversation

  • Questions, requests, or actions being asked

  Output format:

  Label only. No punctuation. No extra text.

  

  Once saved, AI Labeling will read every incoming message, understand its intent, apply the correct label automatically, and keep your inbox organized — without any manual work.

Related: How to Create a Label — Learn how to manually create and manage labels in your workspace.