AI Commons

End users access basic AI chat functionality from AI Chat Providers and interact with AI Agents using conversational UIs (Teams, Email, TDX, & supported integrations). No custom websites / apps needed.

Initial Tiering Approach for Chat Providers:

• Free Tier: MS Copilot, no cost to staff/faculty/students, low-to-moderate risk data
• Premium Tier: ChatGPT EDU, per-user licensing to staff/faculty, up to high-risk data

AI Agent Studio

Low-Code / No-Code AI Agent Building platform.

IT Staff and Non-IT Power Users (Agent Builders) can build their own AI Agents and automations. CIT Applied AI Team can build and maintain AI Agents and Automations for campus customers as part of a separate service offering (Managed AI Agents Service).

AI Gateway

Central hub for AI model and Agent governance, access (API Keys), monitoring, and cost-tracking.

• Technical end users build their AI-powered applications and agents around the AI Gateway to access approved AI models.
• Developers access models safely with AI coding assistants.
• Agent Builders configure their Agents to access approved AI Models.
• End Users access MCP Agents via AI Gateway.

Managed AI Agents Service

The AI Platform components are built with a “self-service first” mindset. Campus units can be onboarded onto the Platform offerings and get charged only for what they use in a metered capacity to recover our central costs (model token usage, number of requests, number of workflow executions).

That said, a high-level service offering will be developed on top of the AI Platform that will build and optionally maintain AI Agents within the AI Platform supported service offering and features (AI Agent Studio & AI Gateway).

What the service will do

• Accept proposals from campus units to have initial AI Agent POCs built by the Innovation Lab for no charge (we want to encourage innovation without monetary roadblocks)
• Accept fees to build AI Agents from the ground up
• Accept fees to add features to existing AI Agents
• Accept fees to provide ongoing maintenance and support for existing AI Agents

What the service won’t do

• Build or maintain custom frontends or backend APIs that cannot be built within the AI Agent Studio
• Build or maintain complex databases or backend data storage, especially not for high-risk / regulated data
• Other CIT units (Custom Dev, DBAs, Cloud Team, etc.) may provide these services, if necessary

Current Usage Metrics

AI Gateway

We have onboarded over 80 teams to the AI Gateway across all of Cornell’s colleges and campuses. This includes teams at WCM, Cornell Tech and Cooperative Extension Offices. We onboard an average of 2-3 more teams every week.

Marty has been starting discussions with various campus IT directors on plans to offer them governance and insights into their own sub-organization’s AI usage. Ideally, we will work to allow them to self-service manage users and teams within their own orgs.

SandboxAI has also been fully run using the AI Gateway (700-800 monthly active users).

AI Agent Studio

Current 25+ AI Innovation Lab projects are being focused within the AI Agent Studio by the various tech leads. This is giving our team insight into the advantages and limitations of the service offering. This is informing what our high-level service offering will be able to do and how much work is involved in building and maintaining different types of agentic workflows.

We are still working on enterprise licensing with the vendor before we onboard campus customers to the Agent Studio.