Team 2, Team Squared: Harvest House

We built an AI-powered system to simplify the customer survey experience, streamline grant access, and help staff manage these data—making it easier to support Harvest House’s non-profit.

Comment

Inspiration

At Harvest House, staff often struggle to locate past grant materials and interpret client feedback due to paper-based forms and language barriers.

  • Drawing from our CS background, we knew unstructured data like the grant filing system is best handled by natural language processing—making this a perfect opportunity to combine AI with practical nonprofit needs.
  • For the customer surveys, in a past nonprofit project, a manual volunteer signup sheet caused confusion and delays. Complex tech tools didn’t help as they overwhelmed staff and volunteers. Replacing it with a simple digital survey solved the issue instantly. For Harvest House, we applied the same "simple-is-better" logic: Google Forms are accessible, familiar, and require no training.

What it does

The solution is two-parts:

1. Grant file storage system We built a self-hosted AI chatbot that retrieves past grant files or relevant information

  • It connects to files stored in Zoho CRM that Harvest House is implementing, so no new systems or storage are needed.
  • The chatbot ensures data security because it is self-hosted
  • Makes it easy for any staff member to find relevant grant content instantly.

2. Surveys We replaced paper-based surveys with a Google Forms–based digital system that:

  • Offers translated fields to accommodate non-English speakers
  • Collects clean, structured data that feeds directly into Google Sheets
  • Customers can run on any tablet, phone, or kiosk

How we built it

Grant Chatbot:

  • Files are stored on the Zoho CRM that Harvest House is already implementing
  • Uses a retrieval-augmented generation (RAG) model to scan and pull past grant files.
  • Any user can prompt the Chatbot for past grant lookups

Survey System:

  • Built using Google Forms
  • Translated by Google Translate
  • Responses flow into Google Sheets.

Challenges we ran into

Issues creating the Chatbot:

Fixed issues after debugging:

  • Inconsistent results: same question gives different answers
  • Search too strict or too loose: misses info or pulls in unrelated content
  • Unreadable PDFs files: look uploaded but contain no usable text

Issue that still exists:

  • Slow updates: the demo chatbot functions slowly

Accomplishments that we're proud of

  • Built a zero to low-cost system
  • Locally-hosted a chatbot with mock data

What's next

  • Survey: Sorts Google Form responses with AI for staff
  • Grants: Generates grant templates in chatbot for future grant writing
  • Overall: Dashboard views of client needs within Zoho CRM

https://docs.google.com/presentation/d/1kGfLeIRLy3rcRoE3T3goQciqRPcFYyW6yzAoiqcupgM/edit?usp=sharing

Built With

Share this project:

Updates