Create AI labelling

AI Labeling automatically reads incoming conversations, understands the sender's intent, and applies the most appropriate label without requiring manual review.

What is AI Labeling?

AI Labeling automatically analyzes incoming conversations and assigns labels based on the intent of each message.

Instead of manually reviewing every reply, AI Labeling categorizes conversations as they arrive, helping your team prioritize responses, organize conversations, and trigger downstream workflows.

Configure AI Labeling

Step 1: Open AI Labeling

1. Go to Settings.

2. Select AI Labeling.

   

Step 2: Enable AI Labeling

Turn AI Labeling ON.

AI labeling requires the following:

• AI Labeling enabled

• A connected AI provider

• A valid API Key

• A supported AI model

Without all four, conversations cannot be labeled automatically.

Step 3: Connect an AI Provider

Select your preferred AI provider.

Enter the corresponding API Key.

Once validated, you'll be able to choose one of the supported AI models.

Step 4: Select an AI Model

Choose the model you want MasterInbox to use for conversation classification.

Each provider offers multiple models with different performance and pricing characteristics.

For most workspaces, we recommend:

• GPT-5 Mini (OpenAI)

• Claude Sonnet 4.6 (Anthropic)

• Claude Sonnet 4.6 (OpenRouter)

These models provide excellent intent recognition while remaining cost-effective.

Use a Custom OpenAI-Compatible Endpoint

Under the OpenAI provider configuration, turn on Use custom endpoint to connect an OpenAI-compatible service such as an Azure OpenAI deployment or proxy.

1. Enter the Custom Base URL for the endpoint.

2. Enter its Custom Endpoint API Key.

3. Enter the Custom Endpoint Model name provided by that service.

4. Click Test Connection to validate the endpoint.

The settings explain that the default OpenAI API key is used automatically if the custom endpoint is unreachable. Ensure your regular OpenAI configuration is valid if you want that fallback.

Copy AI Labeling Settings to Other Workspaces

If you manage multiple workspaces, you can copy your AI Labeling configuration instead of configuring each workspace manually.

How to Copy AI Labeling Settings

1. Go to Settings → AI Labeling.

2. Click Copy AI Label Settings in the right corner.

3. Select one or more destination workspaces.

4. Click Copy.

Your current AI labeling configuration, including provider, model, labeling options, default labels, and custom prompt (if configured), will be copied to the selected workspaces.

AI Labeling Dashboard

The AI Labeling dashboard shows Emails Labelled, Pending, and Failed counts, plus Total Tokens. Use the date filters Last 7 Days, Last 10 Days, or Last 30 Days to review labeling activity.

If conversations fail to process, resolve the underlying provider issue, then use Requeue Failed Labeling to retry failed conversations.

Configure Labeling Options

After connecting your AI provider, configure how AI Labeling should behave.

Label Old Replies

When enabled, MasterInbox automatically labels conversations that already exist in your workspace.

This is useful when enabling AI Labeling on an existing workspace with historical conversations.

If disabled, only new incoming conversations will be labeled.

Relabel Ongoing Conversations

When enabled, MasterInbox reevaluates ongoing conversations as new incoming replies arrive. A later reply may indicate a different intent from an earlier message.

Only Relabel Conversations with These Labels

Optionally select the existing labels whose conversations should be eligible for relabeling. This lets you limit reevaluation to conversations with the selected labels.

The separate Single Label per Prospect setting in Labels & Properties controls whether a prospect can retain multiple labels.

Choose Your Labeling Method

MasterInbox supports two approaches to AI Labeling.

Default Labels

The default labeling system uses MasterInbox's built-in AI prompt together with the labels you enable.

No prompt engineering is required.

Simply select the labels you want available, and AI Labeling automatically classifies conversations.

This is the recommended option for most workspaces.

Available Default Labels

MasterInbox includes the following built-in labels:

When using Default Labels, AI Labeling reads the entire conversation, understands the sender's intent, and applies the single most appropriate label.

You can enable only the labels that are relevant to your workflow. In the example settings shown above, Referral and Sales Pitch are available but unchecked; you can enable them if needed.

View and Edit the Default Prompt

When Use your custom Prompt is off, the Default Prompt section displays the built-in classification instructions. Click Edit to modify them for your workspace. The interface explains that editing the built-in prompt makes it your own custom prompt.

The displayed prompt covers both email and LinkedIn messages. It instructs the AI to judge the sender's intent rather than message formality, avoid assuming short LinkedIn replies are unclassifiable, and distinguish connection acknowledgments from actual interest. It also differentiates sales pitches, wrong-person replies, and referrals, and uses ordered classification rules.

How the Default Prompt Interprets Intent

When running on default labels, the system reads the intent of each message and assigns the closest matching label automatically. Here's how it works in practice:

Custom Prompt

Turning on Use your custom Prompt replaces the default labeling setup with your custom instructions. Leave it off to use the built-in default labels and prompt.

Enable Custom Prompt if you want complete control over how conversations are labeled.

Instead of using MasterInbox's default prompt, you can write your own instructions for the AI.

Custom prompts are ideal when:

• Your team uses custom label names.

• You have unique qualification criteria.

• You want more specific classification rules.

• Different clients require different labeling logic.

Using a Custom Prompt

MasterInbox includes a built-in AI prompt that works well for most email workflows.

However, if your team uses custom labels, follows unique qualification rules, or needs more control over how conversations are classified, you can enable Custom Prompt and define your own labeling instructions.

When a custom prompt is enabled, your instructions replace the default labeling prompt.

Example Custom Prompt

You are responsible for labeling incoming conversations.

Analyze every incoming message and assign exactly one label based on the sender's intent.

Available labels:

Interested
Meeting Request
Information Request
Not Interested
Wrong Person
Follow-up in Future
Automated Response

Rules:

- Choose only one label.
- Never invent new labels.
- If multiple labels apply, choose the strongest intent.
- Return only the label name.
- Do not explain your reasoning.

Requeue Failed Labeling

If AI Labeling fails to process conversations because of a temporary billing issue, you don't need to manually relabel your conversations.

Use Requeue Failed Labeling to retry only the conversations that previously failed.

Common reasons include:

• OpenAI billing issues

• API quota limits

• Temporary provider outages

• Recoverable API errors

After resolving the issue with your AI provider, click Requeue Failed Labeling.

MasterInbox will automatically retry labeling only the failed conversations.

Previously labeled conversations are not processed again.