Drafts the campaign brief, routes it to you to validate, and improves from every edit and outcome.
You ask in plain English. Based on the request, it picks its role — no mode switch, no setup:
Reads every source — product usage, engagement, support, billing — and maps each one onto four kinds of data:
Applies your rules to segment each account and pick the play — and cites the rule as the reason, not just a dashboard:
Drafts the campaign brief — every field traceable to its source — routes it to you to validate, and improves from every edit and validation.
No data engineering. Generate a realistic sandbox, tune your lifecycle stages and messaging, and start chatting with your customers today. Connect real feeds when you're ready.
Scoped tenants, curated read-only views, nightly rule runs, and human validation. The model only ever sees segment summaries — never customer rows — so it deploys in your tenancy with almost nothing to review.
See how it deploys →Your · Small business · —
each one is its own model, data, and chat
For large teams, turnbrief deploys in your own cloud. Your data never leaves — and neither do the prompts, because the model only ever sees aggregates.
The pipeline and app run in your VPC against your warehouse (Snowflake / BigQuery / Postgres). The model runs through Bedrock or Vertex in your own account. Sign-in is your SSO/SCIM.
Every prompt is a segment summary — {segment, size, median MRR, state mix} — never a customer row. The security review that usually stalls in-VPC AI doesn't apply: there are no records to approve sending.
A forward-deployed specialist maps your source tables onto the model, defines your lifecycle stages and first segments from your own CRM taxonomy, then hands the definitions to your admin to own.