Classify Text and Documents — With Zero Training

Aito classifies text using Bayesian inference over your own data. No NLP pipelines, no tokenizer configuration, no model training. Upload a table with labeled examples and predict the label for new entries.

Natural language classification — a question matched to the correct answer with confidence scoring

Try Live Classification

These queries classify real support messages — routing tickets, detecting sentiment, and matching knowledge base answers.

{
  "from": "prompts",
  "where": {
    "prompt": "The package was damaged"
  },
  "predict": "assignee",
  "limit": 1
}

How It Works

1. Upload Labeled Data

Provide a table with text and labels — support tickets with categories, documents with types, messages with sentiment:

{
  "prompt": "The package was damaged",
  "assignee": "shipping-team",
  "sentiment": "negative"
}

2. Classify New Text

Send new text and predict the label:

{
  "from": "prompts",
  "where": {
    "prompt": "I need to change my delivery address"
  },
  "predict": "assignee"
}

3. Get Classification with Confidence

Aito returns the predicted label with a probability score. Set thresholds to auto-route high-confidence classifications and flag uncertain ones for review.

Key capabilities:

• No text preprocessing — Aito tokenizes and indexes text automatically
• Multi-label classification — predict category, priority, sentiment, or any field
• Fuzzy matching — handles typos, abbreviations, and natural language variation
• Instant updates — add new labeled examples and classification improves immediately

Where This Is Used

• Customer Support — ticket routing, response suggestions, sentiment detection
• Process Automation — document classification, email routing
• Accounting & Finance — invoice categorization, expense classification

Get Started

Start for free → Upload labeled text data and start classifying in minutes.

Read the API docs → Full reference for text features, the $match operator, and classification patterns.