Microsoft Discovery Platform Brings Agentic AI to Scientific Research
Microsoft has officially launched its Discovery Platform, marking a significant advancement in the deployment of AI for scientific research. This service is now available in a production-ready environment, specifically tailored to meet the needs of scientists and researchers looking to harness AI agents in their work. The platform is built to streamline processes like data analysis, hypothesis generation, experimentation, and knowledge management, utilizing a suite of specialized AI agents.
Integrating AI in Scientific Processes
The general availability (GA) announcement at Microsoft’s Build 2026 event underscores the company’s commitment to embedding agentic AI across its services, reflecting the increasing trend of using AI to propel scientific and industrial innovation. The Microsoft Discovery Platform is underpinned by a graph-based knowledge engine, designed to interlink proprietary research data with external scientific information. This setup empowers AI agents to interpret complex relationships, assess varying outcomes, and bolster iterative research processes.
In conjunction with this launch, Microsoft has introduced a preview of the Microsoft Discovery app. This on-premises desktop application targets researchers, students, academic labs, and scientific teams who might not yet be prepared for full-scale enterprise deployment.
[Click on image for larger view.]
Figure 1. Welcome screen for the Microsoft Discovery app (currently in preview). The emphasis, consistent with recent AI developments at Microsoft, is on creating agents that go beyond answering questions. These agents are designed to plan, reason, utilize tools, and navigate multi-step processes. This approach is particularly applicable to research and development, where repeated cycles of hypothesis, experimentation, validation, and verification are standard.
Core Features and Governance
The Microsoft Discovery Engine is at the core of the platform, supporting the fundamental cycle of scientific work. It facilitates teams in transitioning from evidence to hypotheses, then executing, analyzing, and iterating on them. Microsoft describes this version as a “production-ready platform” for research and development environments.
For enterprise IT and research organizations, governance remains a pivotal concern. Microsoft positions Discovery as a system capable of integrating institutional knowledge, domain-specific data, simulation tools, and external scientific information while ensuring that results remain verifiable and workflows reproducible. The platform emphasizes retaining human judgment as central to research decisions.
The preview version of the Microsoft Discovery app aims to lower entry barriers. Available for download on GitHub, it is designed for use with a GitHub Copilot account. This app provides smaller teams with the opportunity to engage in literature research, hypothesis building, scientific reasoning, and iterative experimentation before transitioning their work to the broader Microsoft Discovery platform.
Real-World Applications and Use Cases
Microsoft has highlighted several early applications by research institutions and industry partners. For instance, Yale Engineering utilized the Discovery Engine for their work on small molecule design for grid-scale aqueous organic redox flow batteries. Associate Professor David Kwabi noted the synergy between human-led experiments and AI’s capacity to explore vast chemical design spaces.
Collaborations with Pacific Northwest National Laboratory focus on energy storage and biosystems engineering, including self-driving scientific workflows connecting AI agents with laboratory automation. Ginkgo Bioworks is partnering with Microsoft for biological discovery, using specialized agents for dataset analysis, hypothesis generation, and experiment design.
Commercial and industrial applications are also emerging. BHP employs Discovery to investigate advanced copper leaching methods, while Syensqo leverages agentic AI for developing next-generation heat transfer fluids for semiconductor manufacturing. Meanwhile, GSK is exploring Discovery for drug development workflows.
Expanding Microsoft’s Agent Strategy
This announcement aligns with Microsoft’s broader strategy of expanding its agentic capabilities across platforms like Azure, Microsoft Foundry, GitHub, and Microsoft 365. With Discovery, Microsoft targets a niche but potentially high-value audience: organizations whose research processes are costly, data-heavy, and subject to rigorous regulatory or scientific scrutiny.
Microsoft Discovery is now generally available, with the Microsoft Discovery app currently in preview. Microsoft has indicated that preview features may evolve before the final release. For more information, visit the official Microsoft blog.
“`
