Fivetran

Rippling launches Data Cloud to measure AI ROI

Rippling today launched Rippling Data Cloud, a product that embeds analytics and organisational context inside its human-capital-management platform to surface workforce insights — including which employees generate value from AI tool spend. CEO Parker Conrad demonstrated dashboards that combine data from sources such as Anthropic usage logs, GitHub pull requests, Salesforce tickets and internal performance ratings to identify over‑spend and under‑performance, and to enforce spending limits or alerts. Rippling also announced a Business Banking product with high-yield checking and same‑day payroll. The company says roughly 560 companies use the AI features, generating about $5–7 million of new monthly revenue; the base SKU with Rippling AI is around $20 per month with usage-based charges for heavy consumption. Conrad said Rippling has shifted much usage from Anthropic to OpenAI’s 5.5 model and reiterated the company is not planning an IPO soon.

Data Engineering Harness: AI-Driven Next Decade

The article argues the Modern Data Stack's decoupling improved capabilities but created operational complexity that traps data engineers in tool management. The author proposes a new layer — the Data Engineering Harness — that exposes engineering capabilities (ingestion, CDC, orchestration, observability, governance) as callable, auditable skills for LLMs and agentic systems (e.g., Codex, Claude Code). Harnesses provide engineering boundaries, observability UIs for human review, and memory/skills to make AI-generated pipelines production-ready. The piece cites WhaleStudio's Harness Suite as an example and reports a demo where a MySQL-to-Snowflake ETL pipeline was automated with Codex and WhaleStudio in about 10 minutes. The article frames future data engineers as governors of harnessed capabilities rather than manual tool operators.

Best Practices for Building a Data Analytics Platform

This technical guide outlines practical best practices for designing and building a scalable data analytics platform. It defines four analytics maturity levels (descriptive, diagnostic, predictive, prescriptive) and a five-layer architecture (data ingestion, storage/data warehouse, transformation, business intelligence, security & compliance). The article recommends modern patterns such as ELT, modular/multi-tenant architectures for SaaS, and a technology stack centered on Python and SQL for data work plus Node.js and TypeScript/React for application layers. It highlights essential features—scalable ingestion, governance (RBAC, lineage, audit), high-performance querying, extensibility (APIs/SDKs), and tailored visualization/UX. A Seedium case study (AllClinics) describes using asynchronous Python ingestion, Google BigQuery, Docker/Kubernetes orchestration, and an interactive React front end to consolidate large healthcare datasets (millions of procedures across thousands of hospitals). The piece also recommends testing, cloud deployment (AWS/GCP/Azure) and establishing a central metrics system.