What is Cosmos?
A non-technical introduction to the Cosmos AI agent platform — what it does, who it serves, and why it matters
Updated 2026-09-28
Verified 2026-09-28
Cosmos is EY's enterprise AI agent platform — a single foundation for building, running, and managing AI agents that automate complex business tasks. It gives teams the tools to connect AI to real enterprise systems without starting from scratch every time.
Why Cosmos?
Enterprise AI adoption is held back by fragmentation. Every project re-invents authentication, tool integration, model access, and monitoring. Cosmos eliminates that duplication by providing a shared platform that handles the hard infrastructure problems once, so teams can focus on building solutions that deliver business value.
What Does Cosmos Do?
At its core, Cosmos lets you build AI agents that can reason through multi-step problems, connect them to your data and systems, and run them reliably at enterprise scale.
Think of it as a control center for AI-powered automation. Instead of writing one-off scripts or manually integrating AI models, teams use Cosmos to:
- Create AI agents that understand context, plan actions, and carry out tasks across multiple systems
- Connect to enterprise tools — APIs, databases, custom logic, document repositories, and on-premises systems behind corporate firewalls
- Monitor everything — track agent performance, costs, response quality, and usage across the organization
- Manage knowledge — index documents and data so agents can retrieve relevant information when answering questions or making decisions
- Add voice interfaces — deploy agents that users can speak with directly
- Handle sensitive data — detect and redact personally identifiable information in text and images before it reaches AI models
All of this runs on a shared platform with built-in authentication, role-based access control, and organizational boundaries — so different teams and clients can use the same infrastructure securely.
Key Capabilities
| Capability | What It Means |
|---|---|
| Agent Builder | Design AI agents with specific instructions, tools, and guardrails — no deep coding required |
| Tool Integration | Create and share reusable integrations (API calls, database queries, prebuilt custom tools) |
| Multi-Model Support | Use AI models from multiple providers (OpenAI, Google, Anthropic, Azure) with your own API keys |
| Knowledge Base | Upload documents and data so agents can search and reference them in conversations |
| On-Premises Access | Securely connect agents to systems inside corporate networks without opening firewall ports |
| Observability Dashboard | Monitor agent runs, costs, latency, and quality metrics in real time |
| Voice Agents | Deploy agents with real-time speech interfaces for hands-free interaction |
| PII Protection | Automatically detect and mask sensitive personal information before processing |
| Role-Based Access | Control who can build, run, and manage agents at the organization and project level |
| Evals | Test agents against versioned datasets and score live conversations |
Multi-Cloud Deployment
Cosmos runs on both Microsoft Azure and Google Cloud Platform, giving organizations flexibility in where their AI workloads operate. The platform adapts its storage, analytics, and infrastructure layer to match the target cloud — teams get the same experience regardless of the underlying environment.
For local development and demonstrations, Cosmos also runs entirely on a single machine using Docker Compose, making it easy to evaluate and build on without cloud infrastructure.
What's Built on Cosmos?
Several production applications already run on the platform, demonstrating its versatility:
- JumpStart SAP Intelligence — a consulting workbench that helps EY teams assess, plan, and execute SAP and Oracle migrations. It uses AI agents to analyze code quality, generate migration blueprints, and accelerate ABAP development.
- JumpStart AMS — an application management support portal for project-based ticket intake, triage, routing, and SLA tracking.
- JumpStart Workbench — an IDE-integrated agent workbench that brings Cosmos AI capabilities directly into the developer's coding environment.
These are just the first wave. Any team can build new applications on Cosmos using the platform's TypeScript and Python SDKs.
Who Uses Cosmos?
Internal EY teams — technology consultants, developers, and product teams who build AI-powered solutions for clients and internal operations.
Client-facing engagements — consulting teams deploy Cosmos-powered applications during transformation projects, giving clients AI-assisted analysis, migration planning, and ongoing application support.
Platform builders — developers who create new applications on top of Cosmos, leveraging its shared services instead of building authentication, tool management, and agent execution from scratch.
What's Next?
- See Platform Architecture for every service and application in one place
- Read about Core Concepts to understand the building blocks (organizations, agents, tools, apps)
- Start using Cosmos with Getting Started with Cosmos, or set up a development environment with Deploying Locally
For a technical deep-dive into how the platform is structured, see Platform Architecture. For details on the applications built on Cosmos, see Consumer Applications.