Microsoft Fabric agentic AI just took a significant leap forward. In January 2026, Microsoft announced the acquisition of Osmos, a Seattle-based startup that automates data engineering using AI agents. For organisations already invested in Microsoft Fabric, or those considering the platform, this development is worth paying close attention to.
At Data-Driven AI, as a Microsoft Fabric Featured Partner and one of only a few companies in Australia to hold the Microsoft Advanced Specialisation in Analytics, we have been watching this acquisition closely. Here is what it means for your data platform, your team and your roadmap.
Osmos was founded in 2019 by former Google and Microsoft engineers. The company focused on a problem that every data team knows well: too much time spent preparing data, and not enough time analysing it.
The Osmos platform used agentic AI to automate the most manual and repetitive parts of data engineering. Its two core products were the AI Data Wrangler, which normalised messy, unstructured data such as nested JSON files, irregular CSVs and PDFs, and the AI Data Engineer, which generated production-ready PySpark code for building pipelines.
Before the acquisition, Osmos had already launched as a native app within Microsoft Fabric. The results spoke for themselves. According to Microsoft Senior Director of Product Roy Hasson, using Osmos on top of Fabric Spark reduced development and maintenance efforts by more than 50 per cent.
Microsoft acquired Osmos to bring those capabilities natively into Fabric at scale, integrated directly with OneLake. You can read the official Microsoft acquisition announcement for the full details.
The Osmos team has now joined Microsoft’s Fabric engineering organisation. The standalone Osmos products are being phased out as the technology is absorbed directly into the Fabric platform.
What this means in practice is that autonomous data engineering capabilities, already proven in production environments, will become a core part of the Microsoft Fabric experience. Rather than relying on third-party tools or writing extensive pipeline code by hand, data teams will increasingly be able to delegate ingestion, transformation and schema evolution tasks to AI agents.
Microsoft’s Corporate Vice President of Azure Data Analytics, Bogdan Crivat, described the acquisition as a step toward a future where autonomous AI agents work alongside people, helping reduce operational overhead and making it easier to connect, prepare, analyse and share data across the organisation.
For Fabric users, this signals a clear direction: the platform is moving from human-assisted workflows to AI-led, human-approved automation.
OneLake sits at the heart of Microsoft Fabric as a unified data lake. Osmos technology is designed to work directly within OneLake, turning raw, fragmented data from customers, partners and suppliers into analytics-ready assets, without manual transformation work.
The AI Data Engineer agent interprets requirements, writes code, validates outputs and deploys to production tables pending human approval. This keeps humans firmly in control while removing the most time-consuming steps from the data engineering workflow.
Industry analysts have noted that this addresses one of Fabric’s key growth areas. As the platform expands with new databases and deeper OneLake interoperability, the bottleneck shifts from data access to data readiness. Osmos fills that gap directly. InfoWorld’s analysis of the acquisition provides further context on what this means for enterprise data teams.
For Australian government agencies and enterprises already using Microsoft Fabric, the Osmos acquisition has practical implications that go beyond the technology itself.
Data teams that currently spend significant time on manual ETL work, pipeline maintenance and schema mapping will have access to AI agents that can handle much of that groundwork automatically. This frees engineers to focus on architecture, data quality, governance and extracting strategic insights from their data.
For organisations subject to Australian Government data policies and security requirements, the fact that this automation runs natively within Fabric and OneLake, under existing governance controls, is an important consideration. There is no additional data leaving the environment to a third-party service.
If you are responsible for a Microsoft Fabric Advanced Analytics platform today, this is the right time to begin planning for how agentic capabilities will integrate into your existing architecture and governance framework.
If your organisation has not yet developed a clear strategy for agentic AI within your data platform, the Osmos acquisition is a signal to start that conversation now.
The direction Microsoft is taking with Fabric is clear. Autonomous data engineering is not a distant future state. It is being built into the platform today. Organisations that wait to engage with this shift will find themselves behind, not just in technology adoption, but in the efficiency and cost advantages that early movers will gain.
Data-Driven AI specialises in Agentic AI and custom agent development on Azure. We have hands-on experience deploying agentic AI workflows for government and enterprise clients across Australia. We are also customer zero, meaning we use our own AI agents internally to run our business, giving us a genuine understanding of both the opportunity and the practical realities.
Whether you are building a new Fabric platform or evolving an existing one, our team can help you develop a data strategy and AI governance framework that positions your organisation for the next generation of autonomous data engineering.|
Ready to Build Smarter with Microsoft Fabric Agentic AI?
Data-Driven AI is a Microsoft Fabric Featured Partner and one of Australia’s leading specialists in Agentic AI on Azure. We help government agencies and enterprises design, build and optimise Fabric platforms ready for the future of autonomous data engineering.
Explore our services: Fabric Advanced Analytics | Agentic AI Services Book a Free Discovery Call at data-driven.com
