AWS publishes Agentic Data Operations Platform, an open-source reference architecture for AI-driven data engineering

AWS has published the Agentic Data Operations Platform (ADOP), an open-source reference architecture that uses AI agents running on Amazon Bedrock to automate the setup work involved in standing up a new data source.

What's new

AWS's machine learning blog frames the problem directly: "Data engineering teams routinely spend weeks standing up a single new data source... The Agentic Data Operations Platform (ADOP) on AWS is designed to significantly accelerate that timeline." ADOP is not a managed AWS product or a GA service — it is an open-source sample architecture, published on GitHub, that teams can deploy themselves. It chains together agents built on Amazon Bedrock to automate several stages of the data engineering pipeline: extract-transform-load (ETL) work, data quality checks, semantic modeling, and compliance controls.

The pitch is speed: work that would normally take a data engineering team multiple weeks of manual setup — schema mapping, validation logic, documentation, access controls — is handled by agents that carry out the equivalent work in hours instead.

Context

AWS has been building out a family of Bedrock-based agent tooling aimed at specific enterprise workflows rather than general-purpose chat. ADOP follows the same pattern as the Agentic Resource Discovery specification AWS also published this week: rather than shipping a new managed product, AWS is releasing reference architectures and open specifications that show customers how to compose Bedrock agents to solve a concrete operational bottleneck, in this case the slow, manual process of onboarding a new data source into an analytics or AI pipeline.

Why it matters

Data engineering setup work — connecting a new source, validating its schema, documenting it, and wiring in compliance checks — is a common bottleneck that slows down every downstream AI or analytics project relying on that data. By publishing ADOP as an open reference architecture rather than a black-box managed service, AWS is betting that showing customers a concrete, reusable agent pipeline for this work will drive more Bedrock usage than a walled-garden product would, while giving enterprise data teams a starting point they can adapt rather than build from scratch.