AI-Generated Datasets

AI-generated datasets — formerly known as AI Fields — are composable datasets generated from your warehouse data using AI. They transform hard-to-activate signals — product affinities, behavioral patterns, weather readiness, predicted intent — into structured fields you can use in segments, personalization, and campaigns.

🟦

Core AI · Included with your Simon platform.

📘

A dataset is a dataset

Datasets built with AI show up right alongside the datasets your data team builds by hand — the same field library, the same refresh schedules, the same dependency tracking, and the same availability in segments, journeys, flows, and channel actions. And a single request isn't limited to one field: ask for a set of related signals and the resulting dataset carries all of them.

Getting started

You can create AI-generated datasets through the Chat by describing what you want in natural language, or from within a Project's workflow.

1. Describe the attribute — or set of attributes — you want (e.g., "Score each contact on their likelihood to purchase winter apparel")
2. The agent analyzes your warehouse and proposes the dataset logic
3. Review the proposal — including logic explanation, SQL, and sample values
4. Approve it to create the dataset
5. Choose an inference schedule: one-time or recurring at a frequency you set

📘

Created through conversation

AI-generated datasets are built through the Chat interface. Describe what you want in plain language — no SQL or data engineering required.

What AI-generated datasets do

• Extract insights from unstructured data — Turn product reviews, catalog metadata, and behavioral logs into actionable attributes
• Predict customer behavior — Build inference models that score contacts on likelihood to purchase, churn risk, or category affinity
• Produce one field or many — A single request can generate a dataset with multiple related fields (e.g., a category affinity plus a confidence score plus a recommended product)
• Stay fresh automatically — AI-generated datasets support one-time and recurring inference schedules, so computed attributes update on a cadence you choose
• Integrate into your existing workflows — Use the fields anywhere you use contact attributes: segments, journeys, flows, and channel actions

Field details

Every field in an AI-generated dataset gets the same field details page as any other connected field. Open a field from the Schema Builder to see:

From the field details page you can also act on what you see: Create segment with this field, or Ask the agent to dig into it further.

Managing AI-generated datasets

AI-generated datasets don't live in a separate index — find and manage them alongside all your other connected data:

• Schema Builder — every connected table and dataset in one searchable list, including AI-generated ones
• Data Hub — browse, search, and enrich fields across all your datasets
• Chat — ask the agent to update a dataset's logic or delete one you no longer need; you review and approve every change before it's applied

The "Show Only AI-Generated Content" filter on the content index page helps you find templates and content blocks that use AI-generated fields.

Inference models

AI-generated datasets can be powered by inference models that predict customer behavior based on historical patterns. Inference models support:

• One-time inference — Run the model once to generate field values
• Recurring inference — Set a schedule to keep predictions fresh as new data arrives

When updating an inference model, a SQL comparison view shows what changed between the previous and updated versions.

Availability

AI-generated datasets are available to all AI-enabled organizations.