Cloudflare launches unified data platform + AI agent; Town Lake + Skipper reduce analytics latency

Cloudflare detailed Town Lake, its internal unified data platform, and Skipper, an AI-powered natural-language analytics agent, designed to unify access to siloed operational, billing, security, and business data. Town Lake runs on a lakehouse architecture using Apache Trino, Iceberg, Cloudflare R2 object storage, and DataHub for metadata, enabling a single query to join data across Postgres, ClickHouse, and Iceberg tables without data movement. Cloudflare processes over 1 billion events per second across 330+ cities; the platform replaces fragmented discovery across Postgres, ClickHouse, Kafka, BigQuery, and object storage.

Skipper translates plain-English questions into validated SQL queries using metadata, schema, lineage, and runtime inspection. Billing workloads account for 53% of all Town Lake queries (~91,760 queries from 324 employees in measurement period), with tasks once requiring complex SQL now completed in seconds. Design emphasizes governance: new datasets default to inaccessible until automated scanning + human review via internal service 'Skimmer' (combining classification + AI-based analysis for PII detection) clears access. Cloudflare plans deeper integration with internal chat, ticketing, and dev workflows.

For architects: this is a template for enterprise analytics infrastructure where AI agents and lakehouses converge. The 'default closed' governance model is notable—it inverts typical data access patterns and flips burden to curators rather than requesters. Watch for this pattern spreading in SaaS observability stacks, and for vendors who couple lakehouse + agentic SQL to compete in the internal analytics layer.