How would you measure the success of Atlan AI, the generative AI copilot inside Atlan?

Introduction

Atlan AI is the set of generative AI features inside Atlan, first announced in 2023 and expanded since. It helps users write and improve asset descriptions, search the catalogue in natural language, explain SQL and lineage, and translate questions into queries. Atlan has since added context agents that enrich metadata automatically, and an MCP server that serves governed context to external AI agents. For this question, we focus on Atlan AI as the copilot inside the product.

Measuring an AI copilot in a B2B product is harder than it looks. Usage alone can be misleading: people try AI features out of curiosity, and AI-generated content can be wrong. We need metrics that capture whether Atlan AI actually saves time and improves the quality of metadata and answers.

Step 1

Product Context

Atlan AI capabilities include:

• AI-generated descriptions for tables, columns and dashboards.
• Natural-language search ("Which table has customer churn by month?").
• SQL explanation and generation.
• Summaries of lineage and asset context.
• Context agents that propose descriptions, READMEs and other enrichment in bulk.

Key stakeholders:

• Data stewards and engineers who document assets.
• Analysts and business users who search and ask questions.
• Governance leads who care about metadata coverage and accuracy.
• Atlan, which wants AI to drive adoption, retention and expansion.
• Customer security teams, which care about how AI processes metadata.

User flow example:

1. A steward opens an undocumented table.
2. Atlan AI suggests a description and column descriptions.
3. The steward edits and accepts.
4. An analyst later searches in natural language and finds the table thanks to the description.

Product lifecycle stage: Growth. Core features exist; the focus is quality, trust and expanding to agentic workflows.