What counts as "interaction data"?
Labelf is built to analyze customer interactions — support tickets, phone call transcripts, chat conversations, emails, and similar text-based exchanges. Each interaction typically includes the conversation text along with metadata like timestamps, agent names, contact IDs, resolution status, and any other fields your source system tracks.
How data gets into Labelf
There are several ways to connect your data to Labelf:
Integrations — Labelf offers direct integrations with popular customer service platforms. These integrations pull your interaction data automatically and keep it in sync. Currently supported integrations include Zendesk, Freshdesk, Freshservice, and Intercom. For setup instructions, contact your Labelf administrator or reach out to support@labelf.ai.
File uploads — You can upload interaction data from CSV or Excel files. This is useful for one-time imports, testing, or when your data comes from a system without a direct integration.
API — For more advanced setups, you can use the Labelf API to push data programmatically. This is ideal for custom pipelines, scheduled imports, or connecting systems that don't have a built-in integration.
Your workspace administrator will typically set up the data connection during onboarding. If you're unsure how your data is connected, check with your admin or contact Labelf support.
Understanding datasets
Once your data is in Labelf, it's organized into datasets. A dataset is a collection of interactions — think of it as a table where each row is an interaction (or a message within an interaction) and each column holds a specific piece of information.
Common columns you'll see in a dataset include:
Utterance / message text — The actual content of the interaction
Contact ID / ticket ID — A unique identifier linking messages to a conversation
Agent — The support agent who handled the interaction
Timestamps — When the interaction happened
Metadata fields — Any additional data from your source system (e.g., department, priority, language, resolution status)
Your workspace may have one or more datasets, depending on how your data sources are set up. You can switch between datasets when using Search or creating models.
Where datasets appear in the platform
Once connected, your datasets are used throughout Labelf:
Search — Select which dataset to explore using the dataset selector in the top-right corner of the Search page. You can search, filter, and browse through all the interactions in that dataset.
Models — When creating a classification model, the first step is selecting which dataset the model should be trained on. The model will learn from the labels and interaction data in that dataset.
Dashboards — Charts and reports in your dashboards pull from your dataset to visualize trends, volumes, and other metrics.
Next steps
Key Concepts: Datasets, Models, Labels — Learn about the building blocks of Labelf
Creating a Classification Model — Build your first AI model
Need help?
If you need assistance connecting your data or have questions about supported integrations, reach out to us at support@labelf.ai or use the chat widget in the bottom-right corner of the screen.