Klaviyo SQL brings checkable answers to AI agents

Klaviyo announced SQL in its Data Platform on 14 October 2026, allowing merchants to ask business questions through an AI agent and inspect the query behind the answer.

According to Klaviyo’s release notes, questions can be asked in Composer or through its MCP server from tools such as Claude and ChatGPT. Klaviyo supplies its own metric definitions, including what counts as revenue, rather than leaving each agent to infer them.

The company also says Composer is out of preview, with deeper analysis across customer, marketing and commerce data, scheduled reports and audits, and reusable skills for recurring work.

Official Klaviyo demonstration: SQL in Composer and Claude. Watch on YouTube.

Three useful starting questions

These are our suggested checks for a store preparing its next promotion, not promised outputs or Klaviyo quotations:

• Among customers whose first order was in a specified month, what proportion placed a second order within 90 days?
• How does discount use differ between first-time and repeat buyers over the same period?
• Which products are most commonly bought on a customer’s second order?

Ask for the definition and query alongside each result. A recent cohort may not yet have had 90 days to reorder; a discount comparison alone cannot show whether the discount caused a purchase. Those distinctions affect what action the numbers can support.

Keep review in the workflow

Use one question with a known answer as an initial reconciliation check. Record the date range and metric definition, compare the result with an existing report, then investigate differences before scheduling recurring analyses.

Klaviyo’s Composer getting-started guide is the next step for checking the workflow in your account. Confirm your account’s access and capacity before building an operational process around it; this announcement is not a guarantee that every capability is included in every plan.