What challenges did GSK face before AI implementation?

R&D processes in science and engineering are slow, complex, and highly iterative. Scientific knowledge is distributed, nuanced, and siloed across domains. Simulating and validating new materials (such as sustainable datacenter coolants) can take years with traditional methods. Connecting data, tools, and knowledge across domains is difficult for research teams lacking deep computational expertise. Compliance and trust in AI systems is required for enterprise-scale adoption.

What was the impact of the AI solution?

Reduced the time to discover a sustainable, non-PFAS datacenter coolant from years to about 200 hours (plus under 4 months for lab prototype synthesis). Enabled faster development cycles at GSK for medicinal chemistry and at The Estée Lauder Companies for product innovation. Brings capabilities for rapid parallel prediction, testing, and workflow automation to R&D teams. Accelerates semiconductor and material science breakthroughs through third-party integrations.

Source

MicrosoftExpandedEvidence: Medium45/100

Global Innovators Accelerate Scientific Discovery with AI-Driven Agents

Use case typeLife sciences innovationUpdated Jun 13, 2026

Microsoft Discovery is a new enterprise AI agent platform designed to transform research and development (R&D) across sectors. By orchestrating specialized AI agents with a sophisticated graph-based knowledge engine, researchers and scientists can reason contextually through complex data, simulate experiments, and iterate research plans faster than ever before. Built on Azure High Performance Computing (HPC) and Azure AI Foundry, it emphasizes trust, compliance, and extensibility. Real-world impact is demonstrated by Microsoft researchers rapidly developing a sustainable immersion datacenter coolant, a process that normally takes years, in just 200 hours. GSK aims to accelerate medicinal chemistry, while The Estée Lauder Companies focuses on faster product innovation. The platform integrates partner technology from NVIDIA, Synopsys, and PhysicsX to enable advances in pharma, materials, semiconductor design, and industrial engineering. Strategic alliances with Accenture and Capgemini are helping scale deployments. Microsoft Discovery is positioned as a future-proof system for solving the most challenging R&D problems using AI agents and Microsoft’s secure cloud foundation.

Organization: GSK
Industry: Pharma
Location: United States
Published: May 2025

Reported outcomes

Time: Less than 200 hours

Time & speed

Planned next steps

The source says this outcome is planned: Accelerated research and iteration cycles.

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Time: 200 hours decrease

azure.microsoft.comMay 19, 2025Blog postInferred claimMedium evidence strength

Reduced the time to discover a sustainable, non-PFAS datacenter coolant from years to about 200 hours (plus under 4 months for lab prototype synthesis).

Last evidence check: Jul 22, 2026

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: GSK, The Estée Lauder Companies
Provider: Microsoft
Maturity: Unknown
Linked source: azure.microsoft.com

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 3 of 4

1AI research agent
2materials simulation
3knowledge graph

R&D processes in science and engineering are slow, complex, and highly iterative.
Scientific knowledge is distributed, nuanced, and siloed across domains.
Simulating and validating new materials (such as sustainable datacenter coolants) can take years with traditional methods.
Connecting data, tools, and knowledge across domains is difficult for research teams lacking deep computational expertise.
Compliance and trust in AI systems is required for enterprise-scale adoption.

Microsoft Discovery orchestrates domain-specialized AI agents (e.g., for simulation, literature review) using a graph-based knowledge engine.
Built on Azure HPC, Azure AI Foundry, and Microsoft Copilot for trusted and compliant infrastructure.
Researchers can build and define custom agents to fit specific R&D workflows.
Partnerships extend the platform with NVIDIA (AI/GPUs), Synopsys (semiconductor design), and PhysicsX (physics models).
Consulting partners Accenture and Capgemini scale adoption and co-innovation.

Technologies

Azure AI Azure HPC Azure OpenAI Copilot

Reduced the time to discover a sustainable, non-PFAS datacenter coolant from years to about 200 hours (plus under 4 months for lab prototype synthesis).
Enabled faster development cycles at GSK for medicinal chemistry and at The Estée Lauder Companies for product innovation.
Brings capabilities for rapid parallel prediction, testing, and workflow automation to R&D teams.
Accelerates semiconductor and material science breakthroughs through third-party integrations.

Architecture

Microsoft Discovery is built on Azure HPC and Azure AI Foundry, using a graph-based knowledge engine to orchestrate teams of domain-specific AI agents. Copilot acts as a scientific assistant, overseeing workflow orchestration and integrating customer/partner models and datasets. Extensibility is achieved via modular integration with open-source, partner, and custom solutions, with enterprise governance controls. Integrations with NVIDIA ALCHEMI and BioNeMo NIM, Synopsys, and PhysicsX expand the ecosystem for specialized scientific and engineering domains.

Sources & evidence2

Evidence: Medium45/100Evidence strength

Customer explicitly identified
Quantified outcome available
Multiple corroborating sources available
Recent evidence check available
Last evidence check: Jul 22, 2026.

ExpandedExpanded

The same organization appears in newer AI deployment evidence.

Same organization re-documented as recently as 2026.
Cited source last checked Jun 12, 2026 — ok (0/2 broken).

Measures whether this deployment's public evidence persists — not whether the system is still in production.

Type: Blog PostPublished: May 19, 2025Publisher: azure.microsoft.comEvidence: VendorConfidence: Medium

Primary source Source 2

AI-generated summary. Verify important details with the linked sources before relying on this case.

Global Innovators Accelerate Scientific Discovery with AI-Driven Agents

Planned next steps

Use case focus

Challenges

Solution

Reported impact

Comments

Similar cases