How Rippling went AI-native in 6 months with Deep Agents and a 3-layer eval pipeline
LangChain · x · 2026-08-27
LangChain's case study: workforce platform Rippling shipped AI across every product line in 6 months, now serving millions of users.
Architecture: A multi-agent design built on Deep Agents — a supervisor coordinates specialized read, RAG, and action agents — to reason across thousands of tables and hundreds of thousands of fields spanning HR, IT, payroll, and finance, where concepts like "balance" are ambiguous across domains.
Eval pipeline (LangSmith):
- Offline evals: pre-recorded mocks + fixtures run locally on every commit with no external dependencies;
- Post-merge integration evals: 300-400 queries against a full sandbox to validate system health;
- Deploy-blocking evals: 10 critical scenarios against real systems gating every deployment;
- Continuous automated analysis with traces and a semi-automated self-healing loop to catch regressions.
Related event: Ripping Went All-In on AI in Six Months with Deep Agents and Layered Evals(2 posts)→
More from coding & agent
- Optimizing inference on 4090: sub-10ms latency achieved — yacineMTB · 2026-08-27
- Tutorial request: Bulk downloading songs with AI agents and auto-tagging — EnvironmentalTry8353 · 2026-08-27
- Google AI Studio enables two-way GitHub sync with direct commits and one-click deploy — jackwoth · 2026-08-27
- Opinion: Notification systems need rebuilding for agent-primary usage and context linking — andreisavu · 2026-08-27
- OpusClip and Beehiiv integrate to auto-turn videos into newsletters via Claude — azed_ai · 2026-08-27
- Bixbench3: Frontier Agents Score Below 50% in Reproducing Paper Analysis — xeophon · 2026-08-27