Revolutionizing R&D: Meet Microsoft Discovery’s Agentic AI
Microsoft Discovery has been designed as a dynamic, highly adaptable platform. It empowers researchers to integrate the most up-to-date Microsoft features with their own tools, models, and datasets. Furthermore, it seamlessly fits with a broad array of open-source and partner technologies.
We are pleased to unveil Microsoft Discovery, a next-generation platform for enterprises, at Microsoft Build 2025. It’s built to accelerate research and development (R&D), making breakthroughs happen faster and more effectively.
Microsoft Discovery transforms how research teams harness AI. Scientists and engineers can now streamline every stage of discovery—from forming hypotheses and reasoning with advanced knowledge, right through to running simulations and iteratively improving solutions. Within Microsoft Discovery, you can bring teams together to work alongside specialist AI agents, underpinned by a powerful graph-driven knowledge engine. This winning combination helps drive accurate, large-scale results in record time.
One of the key strengths of Microsoft Discovery lies in its flexibility. Researchers can easily combine Microsoft’s latest AI advances with their preferred datasets and models, alongside valuable resources from partners and the open-source community. The platform is securely built on Microsoft Azure, ensuring robust trust, compliance, clear governance, and full transparency—so users can innovate responsibly, always staying in control.
Our own teams at Microsoft have already used Microsoft Discovery’s advanced AI and high-performance computing (HPC) tools to make remarkable progress. For example, we identified a promising new coolant for datacentre immersion technology in just 200 hours—an effort that could have taken months by traditional means. This rapid development sets a new standard, opening the door to safer, more sustainable breakthroughs across numerous sectors. It showcases the true power of Microsoft Discovery in revolutionising R&D for organisations everywhere.
We’re already partnering with forward-thinking clients in fields like advanced materials, semiconductors, energy, manufacturing, and pharmaceuticals to co-create innovative solutions. Our partner ecosystem is expanding rapidly, with many building exciting new capabilities on Microsoft Discovery. As we tap into AI’s vast potential for R&D, we’re only at the beginning of what’s possible!
The agentic vision for science
At Microsoft, we believe in empowering researchers to drive scientific progress faster than ever. With Microsoft Discovery, our aim is to help R&D professionals push boundaries and transform ideas into reality. R&D presents very unique hurdles, such as:
- Sheer volume and complexity of scientific knowledge, stored in many formats and sources.
- The discovery journey uses varied, highly specialised approaches, making it tough to bridge gaps between domains and disciplines.
- Progress is rarely straightforward—research relies on cycles of evidence, debate, and continuous improvement.
These challenges demand a radical shift. Rather than simply speeding up old methods, we need to rethink how R&D is done from the ground up.
Imagine every researcher equipped with a round-the-clock team of expert AI agents, each dedicated to accelerating breakthroughs. This is the heart of Microsoft Discovery’s approach—agentic R&D that weaves advanced AI through every part of scientific enquiry.
In practice, Microsoft Discovery ensures people and AI agents work hand in hand. Together, they quickly update theories and experiment freely, all within a trustworthy, transparent, and fully governed platform. To deliver on this promise, Microsoft Discovery’s AI agents are designed to:
- Navigate intricate knowledge graphs connecting wide-ranging resources and datasets.
- Master specialist roles, from data analysis to experimental design.
- Gather insights from results, adapting plans as new data emerges.
Introducing Microsoft Discovery
We’re entering a new chapter in R&D with Microsoft Discovery: a platform designed to bring agentic AI to life. By leveraging major enhancements from Microsoft and its partners, we’re transforming how scientific breakthroughs happen.
Graph-based scientific co-reasoning
Large language models (LLMs) have made it possible to speed up particular research tasks, especially gathering information and shaping hypotheses. However, these models may not grasp subtle or contradictory data spread across diverse sources. With Microsoft Discovery, the platform’s knowledge graph technology helps overcome these roadblocks by enabling AI agents to reason within context, linking complex data with clarity.
Troubleshooting Microsoft Discovery
If you run into issues integrating your datasets or AI models into Microsoft Discovery, start by verifying your connection settings and data formats. Always ensure you have the latest updates applied. If you encounter permission errors, check that your Azure configuration allows the right level of access and governance.
For problems with knowledge graph connections, review the documentation about compatible data types. If an AI agent does not respond or if queries return unexpected results, try restarting the workspace and clearing cache. Detailed logs in Microsoft Discovery often highlight errors—consult these logs to pinpoint and resolve issues faster. For more advanced troubleshooting, Microsoft’s official support resources can provide step-by-step guidance tailored to your research environment.
Microsoft Discovery, built around a robust graph-based knowledge engine, goes further than traditional data gathering. Rather than simply offering facts, this solution links both internal data and external scientific research, creating a rich network of insights. The result? Microsoft Discovery delivers a nuanced grasp of conflicting theories, experimental results, and the very foundations beneath scientific disciplines.
This advanced contextual reasoning stands out for its openness. Instead of providing one-size-fits-all answers, Microsoft Discovery keeps experts involved by tracing every source and line of reasoning. This level of transparency in AI cultivates confidence, ensures accountability, and allows experts to verify and adjust the system’s conclusions whenever necessary.
Specialised Discovery Agents for Research
Rather than following rigid, isolated processes, Microsoft Discovery supports ongoing, dynamic research and development. It empowers researchers to steer a customised team of AI agents. These agents, designed through simple natural language, adapt and learn over time. They go beyond logic and can run parts of the scientific research process themselves.
Research teams can quickly form bespoke AI teams, tailored to their workflows and expertise, by defining agents’ roles—such as ‘molecular properties simulation specialist’ or ‘literature review specialist’. You have the freedom to select which tools or models agents use and guide how they interact. This approach is far more adaptable than outdated, highly specific simulation software, and you do not need advanced programming knowledge to drive innovation. If something goes wrong, simply adjust your agent definitions or tool choices and see immediate improvements in outcomes.
This seamless, two-way partnership transforms R&D—agents collaborate not just for you, but alongside you. This means breakthroughs can be reached faster, and your human expertise is amplified by AI’s breadth, attention to detail, and calculation speed.
Central to these processes is Microsoft Copilot. Acting as a scientific assistant, Copilot coordinates your AI agent team according to your instructions. It intelligently matches the right tools, models, and information from your platform catalogue, streamlining complex workflows. If your research tools or workflows are not working as expected, you can use Copilot to quickly reassign tasks or choose alternative models, ensuring the discovery process remains efficient and reliable.
Extensible and Designed for Enterprise
Microsoft Discovery runs on Azure for trusted cloud performance, security, and compliance. It draws upon Microsoft’s reliable infrastructure and core governance features to keep enterprise data safe and research work compliant with all regulations.
The platform welcomes an open ecosystem. Microsoft Discovery combines the best of Microsoft innovation with diverse solutions from users and partners. Your R&D team can easily expand the platform, bringing in their preferred tools and models—ranging from custom code to commercial software and open-source solutions. This flexibility makes onboarding new computational tools a breeze. If you struggle with integrating specific models or your workflows aren’t running as expected, tap into the library’s extensibility to make troubleshooting straightforward. Microsoft Discovery remains prepared for tomorrow as well, supporting advanced technology like quantum computing and embodied AI—ensuring your research is future-ready.
Real Impact: The Discovery of a Novel Non-PFAS Coolant Prototype
Recent advances have helped computational scientists move forward using Microsoft Research’s innovative technologies. This progress led to major achievements, such as finding a new solid-state electrolyte using 70% less lithium with the Department of Energy’s Pacific Northwest National Laboratory (PNNL), and enhancing Unilever’s research speed through powerful simulations. Microsoft Discovery ensures such benefits reach scientists across all fields, not just those with deep technical knowledge.
One of the most promising early applications of Microsoft Discovery is underway at the Pacific Northwest National Laboratory. Scientists leverage the platform’s leading AI and HPC to develop machine learning models for complex chemical separations—an essential step in nuclear research. These advancements help streamline the isolation of radioactive elements after nuclear fission, a process that is both time-sensitive and highly complex. In the future, this technology will allow scientists to spend less time in hazardous environments, while also raising yields and purity for safer, more effective results.
—Scott Godwin, Director, Center for Cloud Computing, Pacific Northwest National Laboratory
By harnessing state-of-the-art AI and high-performance computing tools available through Microsoft Discovery, researchers have managed to discover a groundbreaking, non-PFAS immersion coolant for data centres in just 200 hours. Traditionally, developing these coolants could take years and pose significant environmental risks. If you encounter issues replication this fast pace, double-check your agents’ configuration, ensure your simulation models are optimised, and contact Microsoft support for further troubleshooting tips. With the power of Microsoft Discovery at your fingertips, you can accelerate your research, tackle scientific challenges, and drive meaningful innovation faster than ever before.
Many traditional coolants rely on harmful PFAS-based chemicals, making them no longer viable. As global regulations push to phase out these persistent “forever chemicals”, there’s a strong need for safer, environmentally responsible alternatives in this sector and beyond.
Through an advanced digital discovery workflow, we managed to produce an innovative coolant prototype in less than four months. Ongoing testing shows its essential qualities closely match our AI forecasts, validating the effectiveness of our predictive models. While this development remains experimental, it paves the way for breakthroughs in coolant technology and highlights how Microsoft Discovery, with its powerful combination of high-performance computing (HPC) and tailored AI models, can dramatically speed up and enhance research and development. If you’re facing hurdles in selecting coolants for sustainable operations, Microsoft Discovery offers practical, future-facing solutions—helping troubleshoot compliance issues and anticipate regulatory shifts.
Daniel Pope, founder of Submer—a company that builds sustainable, efficient, and resource-smart datacentres—said:
The pace and depth at which Microsoft Discovery screens molecules is a complete game-changer. What used to require years of trial-and-error experiments in the lab, Microsoft Discovery can now achieve in just a matter of weeks, providing much greater certainty.
A Growing Microsoft Discovery Ecosystem
Microsoft Discovery puts advanced, enterprise-grade tools into the hands of global innovators, delivering real-world results for industries ranging from chemistry and pharmaceuticals to manufacturing and semiconductor design.
A robust community is essential to fully unlock the potential of Microsoft Discovery. That’s why we partner with customers, industry experts, and specialist Microsoft teams, blending our in-house technology with sector-specific skills and leading tools. If you’re implementing Microsoft Discovery, start by gathering your key research challenges—then reach out to collaborators early for the strongest outcomes.
Partner Success: Customers and Internal Collaboration
GSK, renowned for its innovative approach to healthcare, leverages cutting-edge technology through every stage of the medicine and vaccine lifecycle. Their vision is to create ever more effective treatments and immunisations, faster, and on a wider scale.
With Microsoft Discovery, GSK hopes to advance their generative platforms, enhancing their capacity to predict and test new medicines. Such collaboration could revolutionise medicinal chemistry, delivering solutions for patients faster than ever before. If you want to get the most from Microsoft Discovery in healthcare, ensure your data integration is robust and workflows connect effortlessly from early research to production.
The Estée Lauder Companies is famous worldwide for its high-performing skincare, makeup, haircare, and fragrance products—each backed by rigorous research and testing. Eager to push boundaries further, the company is adopting Microsoft Discovery to accelerate new product development and maintain standards of excellence.
Our rich R&D data, carefully built over nearly 80 years, is a major asset. With Microsoft Discovery, we can harness this data for agile innovation and deliver tailored products that truly delight our customers.
—Kosmas Kretsos, PhD, MBA, Vice President, R&D and Innovation Technology, The Estée Lauder Companies
Microsoft is also launching a medical research agent using the same powerful knowledge graph engine found in Microsoft Discovery. This tool sorts through medical journals, distilling key insights for researchers. Built into Azure AI Foundry’s expanding suite, the agent equips specialists to handle complex healthcare needs, such as in cancer care, by providing actionable, trusted guidance. If you’re implementing this tool, check for data compatibility and tailor outputs to your workflow for best results.
Focus Keyphrase Solutions: Domain-Specific Offerings
By merging Microsoft’s and NVIDIA’s strengths in generative AI and scientific computing, we plan to offer Microsoft Discovery alongside NVIDIA ALCHEMI and NVIDIA BioNeMo NIM microservices. This integration enables rapid advancements in materials science and biomedicine. Researchers will benefit from cutting-edge methods for candidate identification, property mapping, and synthetic data generation—all inside Microsoft Discovery’s seamless platform.
Deploying AI agents on NVIDIA-accelerated Azure AI Foundry allows teams to rapidly process and organise huge volumes of scientific data. For those troubleshooting slow R&D cycles, leveraging Microsoft Discovery with these partners can fast-track projects, reduce bottlenecks, and provide more reliable results.
AI is accelerating scientific breakthroughs at an unprecedented speed. By incorporating NVIDIA ALCHEMI and BioNeMo NIM into Microsoft Discovery, scientists can now move from raw data to new discoveries faster and with greater efficiency.
—Dion Harris, Senior Director of Accelerated Data Centre Solutions, NVIDIA
Plans are in place to integrate Synopsys’ industry tools with Microsoft Discovery, making chip design workflows more efficient for hardware and software engineers alike. This combination supports innovation in semiconductor engineering, tackling some of the toughest challenges in tech today. For those troubleshooting design delays or productivity gaps, Microsoft Discovery’s partnership with Synopsys introduces AI-driven strategies to streamline every stage.
Semiconductor development is incredibly demanding. By connecting Synopsys’ AI-powered design solutions with Microsoft Discovery, we’re unlocking new levels of productivity and pushing technology innovation forward at a rapid pace.
—Raja Tabet, Senior Vice President, Engineering Excellence Group, Synopsys
In addition, Microsoft is collaborating with PhysicsX to bring its advanced physics AI foundation models into Microsoft Discovery. This integration will help customers boost automation, optimisation, and performance across both engineering and manufacturing. To avoid common implementation pitfalls, ensure that your data formats align and update model parameters regularly for peak efficiency.
The Microsoft Discovery platform is redefining the role of AI in driving scientific and engineering breakthroughs. It is revolutionising the way complex physical systems are conceived, developed, and managed in leading sectors including aerospace, defence, semiconductors, materials, energy, and automotive. With PhysicsX and Microsoft collaborating, a cutting-edge software foundation is being established to shape the future of engineering innovation.
—Jacomo Corbo, Chief Executive Officer and Cofounder, PhysicsX
Integration support
We are pleased to be teaming up with an expanding list of software integrators such as Accenture and Capgemini. These partnerships are designed to help organisations and their partners roll out tailored Microsoft Discovery platform solutions efficiently and at scale.
By joining forces with Microsoft, we are realising an ambitious AI vision for enterprises focused on deep science. Our strategies for modernising labs, enhanced by the Microsoft Discovery platform, create a dynamic environment for scientific progress. This partnership is paving the way for next-generation labs, allowing researchers to experiment, innovate, and test with increased speed and accuracy.
—Adam Borenstein, Managing Director, Global Laboratory Reinvention Lead, Accenture
We are thrilled to introduce the Microsoft Discovery platform and AI agents to sectors that rely on advanced research and development. These technologies are set to empower expert scientists to accelerate innovation, benefiting industries and society as a whole. Our collaboration taps into Capgemini’s expertise, specialist AI knowledge, and Cambridge Consultants’ leading-edge intellectual property. For clients, this means faster discoveries, improved predictive models, and a competitive edge by harnessing data and AI at scale.
—Roshan Gya, Chief Executive Officer, Capgemini Invent
Ready to take the next steps with Microsoft Discovery?
Discover the many ways the Microsoft Discovery platform can empower scientists and engineers to transform research and development processes. Whether you want to optimise workflow, accelerate innovation, or improve outcomes, Microsoft Discovery offers the advanced tools needed for your success.
¹PFAS definitions are based on criteria provided by the Organisation for Economic Co-operation and Development (OECD, 2021), the U.S. Environmental Protection Agency, and Buck et. al. (2011).
