Our system supports AI assistants. You can continue to use our built-in reporting and display features, but if necessary you can use AI assistants to provide additional application, grant and scholarship review, summary, analysis and visualization.
A cloud AI assistant pulls data across the Internet from the grant management system's server and processes it on its own servers elsewhere. A self-hosted open-source LLM runs the opposite way - the AI model lives directly on the grant management system's server, so all the data access and processing happens there too.
The two also differ in reasoning capability, resource availability, and maturity, since the model running in the cloud isn't the same as the model running on-premise.
This means there are different advantages and disadvantages to each approach. These include...
Our AI features are built on the Model Context Protocol (MCP) and OAuth 2.1 for security, open standards already adopted by major AI providers. We've launched with Claude first, and we're adding support for additional assistants as their own capabilities and self-service connector options mature.
The AI assistant resides externally to our system, it is not built-in to our dashboard. This means the AI assistant not only has access to your data, but any data available on the Internet.
Once connected, your AI assistant can help you review applications and grants, analyze evaluations, summarize reports, and answer questions using your actual account data - no more copying and pasting information by hand. Ask about programs, grant cycles, applications, agreements, payments reports, evaluations and users, and your assistant can export the results into a Word document, PDF, or Excel spreadsheet. The assistant can read data such as applications, letters of intent, reports, recommendations, evaluations, and users, though it can't change applicant or configuration data, and uploaded documents aren't yet viewable or downloadable.
Today, only grantmaker account owners and administrators can connect an AI assistant. Evaluators and applicants don't have AI access, although we plan to expand availability to evaluators in the future.
If you find yourself asking your AI assistant to do the same thing repeatedly, just ask it to save that as a reusable prompt. Plain language like "save this as a prompt" is all it takes - no special commands needed.
Connections use OAuth 2.1 security standards, so your username and password are never shared with the assistant - and no one else, including other users or anyone on the Internet, can access your data through the connection. You decide what's shared, the assistant can never exceed the permissions you've approved, and you can revoke access immediately, at any time, for any reason. Just as with your own login, you're responsible for what the assistant does in your account, since it only ever has the access you've granted it.