At Microsoft Build 2025, Microsoft is launching Microsoft Discovery, a new corporate Agentic platform designed to speed up research and development (R&D).
What is Microsoft Discovery?
A new business agentic platform called Microsoft Discovery was created especially to speed up research and development (R&D). Its highly expandable architecture, which was unveiled at Microsoft Build 2025, enables researchers to incorporate partner and open-source solutions, Microsoft breakthroughs, and their own models, tools, and datasets.
Microsoft Discovery’s main objective is to empower scientists and engineers with AI capabilities to revolutionise the entire discovery process. From advanced knowledge reasoning and hypothesis formulation to experimental simulation and iterative learning, this transition seeks to encompass it all. With the use of a graph-based knowledge engine and a group of specialist AI agents, the platform allows academics to work together to produce accurate, fast, and large-scale scientific results.
Microsoft Discovery is based on an agentic R&D paradigm. This new paradigm aims to radically alter the way that R&D is conducted, not just speed up the same tests. It imagines a future in which all researchers can work together with an industrious group of clever, cooperative AI agents dedicated to rapid discovery. This entails integrating AI into the scientific approach at every level.
This platform tackles the following particular issues that R&D faces:
- The body of scientific knowledge is extensive, complex, and dispersed.
- It is challenging to make connections between many disciplines since the discovery process is dynamic, varied, and requires a number of highly specialist techniques and tasks.
- Scientific knowledge develops via evidence, discussion, and improvement; research and development is iterative and rarely produces straightforward, definitive solutions.
Scientific AI agents in Microsoft Discovery must be able to reason across a convoluted and contextual graph that connects all knowledge sources in order to realise this agentic vision.
- Focus on a variety of tasks and domains.
- Take lessons from findings and modify entire study strategy as necessary.
Among Microsoft Discovery’s primary elements and functionalities are:
Graph-based scientific co-reasoning: The contextual understanding required for deep reasoning over dispersed, complex, or even contradicting scientific data is frequently lacking in Large Language Models (LLMs), despite the fact that they can expedite activities such as information retrieval and hypothesis development. The foundation of Microsoft Discovery is a robust graph-based knowledge engine that creates intricate graphs of relationships between external scientific research and proprietary data, enabling a thorough comprehension of competing theories, a range of experimental findings, and underlying presumptions across disciplines. With thorough source tracking and reasoning, this transparent reasoning keeps the expert informed and enables them to verify, comprehend, or modify each step.
Specialized discovery agents: The platform employs an ongoing, iterative R&D cycle in place of compartmentalized pipelines, in which researchers lead and coordinate a group of specialized AI agents that are capable of learning and adapting. Both domain knowledge and process logic are captured by these agents, which are defined using natural language. R&D teams can encode their methods and experience to create custom AI teams that are in line with their procedures and knowledge. Compared to hard-coding behaviours in conventional digital simulation tools, this method offers greater flexibility. ‘Molecular properties simulation specialist’ or ‘literature review specialist’ are two examples, and users can recommend which models or tools agents should use or develop. By working together, these agents enhance human creativity.
Microsoft Copilot as orchestrator: Microsoft Copilot, a scientific AI assistant, is at the heart of cooperation. In response to researcher cues, Copilot coordinates the specialised agents. By establishing end-to-end workflows that integrate cutting-edge AI and HPC simulations, it can determine which agents to use and is aware of the tools, models, and knowledge bases in a customer’s catalogue.
Adaptable and enterprise-ready: Microsoft Discovery is based on Azure services and infrastructure, utilizing Azure’s governance, compliance, and trust controls. It facilitates an open environment by fusing partner and customer solutions with Microsoft’s innovations. By contributing their own tools, models, and knowledge bases whether proprietary, open-source, or commercial R&D teams can expand the platform. With developments like embodied AI and dependable quantum computing, the platform is built to be future-proof.
The salient features of multiple instances of Microsoft Discovery’s practical influence:
- Instead of taking months or years, Microsoft researchers used the platform to find a new coolant prototype with promising qualities for immersion cooling in data centers in roughly 200 hours. The global movement to outlaw these “forever chemicals” is addressed by this prototype, which is non-PFAS. In less than four months, the digital discovery was successfully synthesized, and the initial attributes matched AI forecasts.
- Through cooperation with the Pacific Northwest National Laboratory (PNNL) of the Department of Energy, a novel solid-state electrolyte candidate that uses 70% less lithium was discovered. In an effort to decrease the amount of time spent in dangerous radioactive conditions and enhance yields and purity, PNNL is also utilizing Microsoft Discovery’s capabilities to advance machine learning models that forecast and optimize difficult chemical separations, especially in nuclear science.
- Unilever uses the technology for quick computer simulations that speed up scientific advancements.
Microsoft is building a platform ecosystem with customers, partners, other Microsoft organisations, and global entrepreneurs. Customers co-innovate in manufacturing, medical, silicon design, energy, chemistry, and materials. The following clients are specifically mentioned:
- GSK: Seeking a potential collaboration to enhance their generative platforms for testing and prediction in order to accelerate the development of novel medications and revolutionise medicinal chemistry.
- The Estée Lauder Companies: They are eager to use their exclusive R&D data, which has been gathered over almost 80 years, to speed up product development.
You can also read NetApp Azure Files: Scalable Risk Modelling And EDA Storage
The following partners are delivering domain-specific offerings:
NVIDIA: NVIDIA ALCHEMI and NVIDIA BioNeMo NIM microservices will be integrated to speed up advancements in the life sciences and materials sciences by supplying cutting-edge inference capabilities and expediting the construction of AI models. Their innovations will make it possible to efficiently process vast amounts of scientific data.
Synopsys: Aiming to re-engineer chip design workflows, boost engineering efficiency, and spur innovation, the company intends to integrate its industry solutions to speed semiconductor engineering.
PhysicsX: In order to enable automation, optimisation, and performance across engineering and manufacturing in a variety of sophisticated industries, they intend to incorporate their physics AI base models.
Custom platform deployments are being scaled with the assistance of software integrators such as Accenture and Capgemini. By using their industrial experience and AI skills, they hope to revolutionise labs and increase productivity in R&D-intensive industries.
You can also read Benefits Of Azure Active Directory: Secure Your Organisation
In order to improve information retrieval and synthesise insights from reliable medical publications, Microsoft is also releasing a medical research agent that uses the same graph-based knowledge engine as Microsoft Discovery. The goal of this agent, which is a component of a larger group of specialist agents in the Azure AI Foundry healthcare agent orchestrator code sample, is to provide practical, evidence-based advice for intricate, multidisciplinary healthcare workflows, including cancer treatment.
Built on Azure’s safe basis, Microsoft Discovery is positioned as a revolutionary platform that uses specialised agents managed by Copilot, a graph-based knowledge engine, and agentic AI to significantly speed up and enhance the R&D process across a range of sectors. It seeks to enable more scientists, not only those with extensive computational knowledge, to access sophisticated computational R&D.
You can also read 365 Windows, Azure Virtual Desktop for Modern Virtualization
