Classify Node

The Classify node uses AI to sort records into categories you define. It's the AI option in the Functions group of the element panel (the sparkle icon). Give it a set of records and a set of categories, and it assigns each record to the category it best matches — useful for tagging free-text such as maintenance notes (Mechanical, Electrical, Process), support tickets, or product feedback.

This is a runtime node

Classify runs as part of flow execution — your data is sent to the configured AI provider when the Flow runs. It is distinct from FactoryThread's editor-time assistants, Copilot and Formula Assist, which help you author a Flow and add no AI step when the Flow runs.

How it works

Classify is embedding-based, not prompt-based. At run time it:

1. Embeds each record's description text via your OpenAI connection.
2. Embeds each category (its name and description) the same way.
3. Assigns every record to the closest-matching category by semantic similarity, and adds that category to the record.

Because it compares meaning rather than keywords, a note that says "bearing is overheating" can land in a Mechanical category even though it shares no words with the category name.

Inputs

The Classify node takes two inputs — connect both before configuring it.

Input 1 — Records to Classify

The data you want to categorize (for example, an Entity of work-order notes). Map two fields:

Field	Purpose
Identifier Field	A unique identifier for each record, so results can be tied back to the source row.
Description Field	The text content used to determine the category.

Input 2 — Category Definitions

The list of categories to choose from — typically another source such as an Entity or a small reference table. Map two fields:

Field	Purpose
Category Name Field	The category label that gets assigned (e.g. Mechanical).
Category Description Field	A short description of what belongs in the category. This is optional but improves accuracy, because the description is embedded along with the name.

Configuration

Setting	Description
AI Connection	The OpenAI connection used for embedding-based classification.
Records to Classify	The Identifier and Description fields from input 1 (above).
Category Definitions	The Category Name and Category Description fields from input 2 (above).

Until both inputs are connected, the panel shows a reminder that the corresponding fields aren't available yet.

Prerequisites

Configure an OpenAI connection in Connections first — OpenAI is the supported AI provider, and the same connection supplies the embeddings Classify relies on.

Example

Tag maintenance notes by discipline:

1. Input 1 (Records to Classify): an Entity of work orders. Map workOrderId as the Identifier and notes as the Description.
2. Input 2 (Category Definitions): a small list with a name and a description — e.g. Mechanical / "pumps, bearings, seals, vibration", Electrical / "wiring, motors, drives, sensors", Process / "flow, temperature, pressure setpoints".
3. Pick your OpenAI connection.
4. On each run, every work order is assigned the closest-matching discipline, ready for a downstream Group By or dashboard.

Next steps

• AI Connections — configure the OpenAI provider
• Function Nodes — the rest of the Functions group
• Building Flows — put nodes together end to end