How can marketers create complex customer segments using natural language?

Summary

• Describe segments in plain English instead of writing SQL or complex logic; AI translates your intent into queries against customer data.
• Governed segmentation: segments respect Unity Catalog permissions, so each marketer sees only the customer data they're authorized to query.
• Instant feedback on segment size before launch, letting you refine your criteria without technical back-and-forth.

How can marketers create complex customer segments using natural language?
Marketing teams can build complex customer segments by describing who they want to reach in plain English, and an AI agent translates that intent into data queries against your unified customer records. This removes the friction of learning SQL or waiting for data teams, letting marketers move faster while keeping data governance intact through role-based access controls.

What is natural language segmentation?

Natural language segmentation uses AI to interpret a marketer's informal description of a customer group and automatically build a segment based on the underlying data. Instead of writing complex SQL queries with multiple conditions, a marketer can describe their target segment in plain language like high-value customers in a specific region who spent significantly in recent months. The agent understands the intent, queries the customer data model, and returns the segment with a member count and preview.

How to build segments with natural language

The capability requires these building blocks:

Building block	Why it matters	Implementation
Unified customer data model	The agent needs a single source of truth for customer attributes and behaviors	Stored in a CDP layer that normalizes customer data from multiple sources
AI agent trained on your schema	The agent must understand your business terminology and data structure	Agents require a data dictionary or schema metadata so they can map intent to actual column names and logic
Role-based access control	Marketers should only segment customers they're authorized to see	Governed through the same identity system that restricts database queries
Natural language to query translation	The agent converts plain language into deterministic queries	Requires semantic understanding of business concepts and syntax validation

How Databricks approaches this

CustomerLake serves as the unified customer data layer. Genie Agents understand your customer schema and can translate natural language requests into queries. A marketer opens the Genie Agent in Databricks Apps, types a segment description, and the agent returns a query, member count preview, and can save the segment back to the CDP for activation. Because both the agent and the data are governed by Unity Catalog, the marketer's queries respect their data permissions automatically. You can also use Agent Framework to build a custom agent with your own business logic layered on top.

Use cases

Retail: A marketer wants to target high-value customers who've been inactive. They describe it in plain English, get back a preview of the segment, and can immediately activate a win-back campaign.
SaaS: Marketing needs to identify customers on premium plans who haven't logged in recently. The agent builds the segment, and they can automatically enroll them in a re-engagement drip.
Financial services: Compliance and marketing collaborate to identify high-net-worth customers with recent engagement. The agent respects role-based data access so compliance can review the logic before activation.

FAQs

Does the agent understand complex Boolean logic?

Yes. You can describe OR, AND, and NOT conditions with qualitative terms. The agent maps your intent to the right query syntax.

What if the agent misunderstands my description?

You can iterate. The agent shows you the query it built, and you can refine your description or edit the query directly if you have SQL access.

How does role-based access control work with natural language segments?

The agent uses the same permissions framework as the database. If you're not authorized to see a customer attribute (like revenue data), the agent won't include it in the segment, even if you mention it.

Can I save segments and reuse them?

Yes. Once built, segments can be persisted in your CDP and used for campaign activation, lookalike modeling, or reporting.