Inspiration

I've worked in the non-profit and higher education space for most of my career and funding (or the search for funding) is one of the most time-consuming tasks a research organization can face. I wanted to leverage the freely available data from grants.gov but present it to a non-profit or research university staff member in a manner that supports their needs and existing expertise.

What it does

There are two main pieces to this project -- the first is a Mulesoft implementation that retrieves data from Grants.gov based on keyword search. The second is a Salesforce Agentforce implementation that helps a user:

  • find a series of grant opportunities based on a keyword search
  • find a series of grant opportunities based on keywords derived from the existing research
  • find details about a specific grant
  • save the details of a specific grant to an opportunity record tied to their contact record

How we built it

We combined the Mulesoft topics for Agentforce feature set with the Agent Studio, Flow and Prompt Builder tools in Salesforce.

Challenges we ran into

Understanding how to design an API's returned data to best meet the needs of an Agent was a challenge, as was determining how to best summarize a sample research paper in a few keywords. Working with a freely available API meant keeping things simple in terms of search demands.

Accomplishments that we're proud of

Leveraging Prompt Builder to summarize the research records created in the demo org felt like a lightning strike moment. Learning how to draft an API specification that would return data in the format best designed to meet the needs of an Agentforce agent was also a great experience.

What we learned

I've learned quite a bit about what an Agent can do once it successfully receives data -- Agents are great at recontextualizing something later in a conversation if they've properly ingested it earlier. One example is the project's use of numerical 'ids' for the grant opportunities. I was concerned that the agent starting remove these numbers from the output of the chat to the user -- but upon asking for details of a specific opportunity, the Agent used the appropriate id.

What's next for Test Project

There is ample room for growth for this project, particularly if it were connected to a grant data API with additional filtering capabilities that could more precisely match the research expertise of the user.

Built With

Share this project:

Updates