Team 9- Fundraise

AI tool for nonprofits that auto-summarizes meetings, extracts tasks, identifies recurring issues across transcripts, and suggests practical solutions—so no follow-ups or problems are missed.

Comment

💡 FundRaise Insight: AI Meeting Memory & Action Engine for Nonprofits

FundRaise Insight is an AI-powered "meeting memory with action" system designed to ensure nonprofit teams never lose critical follow-ups and are guided toward higher-leverage strategic decisions.

The Inspiration

Nonprofits meet constantly, but key details and follow-ups frequently slip through the cracks. We set out to build a system that auto-captures notes, summarizes them, turns action items into tasks, and then learns across meetings to surface organization-level issues with practical, ready-to-run solutions.

The goal: Fewer missed opportunities, faster execution, and a living knowledge base that compounds value after every meeting.

Key Features: What It Does

FundRaise Insight automatically transforms fragmented conversations into organized tasks and strategic insights.

  1. Automated Ingestion: Ingests meeting notes/transcripts automatically from Nylas Notetaker via webhooks, or supports manual uploads.
  2. Summary & Task Extraction: Generates concise meeting summaries and extracts actionable items instantly.
  3. Task Management: Stores action items as first-class Tasks (title, assignee, due date, status, priority) and displays them in a clean, intuitive Tasks dashboard.
  4. Cross-Meeting Insight Agent: Analyzes 50–60+ transcripts to rank the top recurring organizational issues (with a $0–100$ score, rationale, occurrence count, and date range).
  5. Brainstorm Agent: On-demand, proposes $2–3$ practical solution ideas per issue, each with expected impact and concrete next steps.
  6. Real-Time Updates: The entire pipeline updates automatically and incrementally when new transcripts arrive via webhook—no manual reprocessing is required.

| Automatic Insights Update | backend/src/services/transcriptProcessingService.ts | The core service that triggers the incremental insight update immediately after a new transcript is processed. | | Insights API Surface | backend/src/routes/insights.ts | The REST API layer exposing endpoints for full analysis, incremental updates, listing, and requesting brainstormed solutions. |

Technical Deep Dive: How We Built It

Tech Stack

Component Technology Role
Frontend Next.js 13 (Pages), MUI, Axios User Interface, data fetching, optimistic updates.
Backend Express.js, MongoDB (Mongoose) REST API, data persistence, business logic.
AI/LLM OpenAI API Server-side structured JSON responses for summaries, tasks, and insights/solutions.
Automation Nylas Notetaker + Webhooks Automatic, real-time ingestion trigger.

Data Model and Services

  • Models: Transcript, Task, Insight, Solution (+ InsightContext for analysis summaries).
  • AI Services:
    • Transcript pipeline to generate summary + action items.
    • InsightAgentService for initial full analysis and incremental updates.
    • BrainstormAgentService for solutions with expected impact and next steps.
  • API: Clean REST endpoints for Tasks CRUD, Insight analysis, listing, and solution generation.
  • Webhooks & Automation: The Nylas webhook handler verifies signatures, persists the transcript, triggers the AI processing pipeline, and initiates the incremental Insight update.

Frontend UX Highlights

  • Tasks board with optimistic updates for quick interactions.
  • Insights page featuring ranked issues, clear rationale, date spans, and a visible "💡 Generate Solutions" button.
  • Clear loading states, alerts, and SSE status during AI processing.

Challenges and Accomplishments

Challenges We Ran Into

  1. Webhook Signature Verification: We implemented express.raw middleware on the webhook route to access the raw request body, which is required to correctly validate Nylas signatures before JSON parsing.
  2. Reliable AI Output Parsing: We enforced structured JSON response formats where possible and added robust fallback parsing/validation for scenarios like task due dates and bulleted lists.
  3. Scaling Analysis: To ensure speed and cost-efficiency, we developed an incremental analysis path that compares a new transcript to existing insights rather than re-analyzing the full history every time.

Accomplishments We're Proud Of

  • End-to-End Automation: A fully automated pipeline: Notetaker $\rightarrow$ Transcript $\rightarrow$ Summary + Tasks $\rightarrow$ Insights $\rightarrow$ Solutions.
  • Cross-Meeting Intelligence: Surfacing the top organization-level issues with transparent rationale, not just anecdotal evidence.
  • Actionable Solutions: Delivering practical solutions complete with expected impact and concrete next steps that staff can immediately act upon.
  • Clean Architecture: Clear separation of models and services, coupled with indexed queries and solution caching for scalability.

Key Learnings

  • For AI in production, the workflow "parse, validate, fallback" is essential for robustness and reliability.
  • Incremental analysis provides massive gains in both API cost and latency compared to repeated full data scans.
  • Small UX affordances (optimistic updates, status spinners, clear alerts) dramatically improve user trust and clarity during long AI processing steps.

What’s Next for FundRaise Insight

Our immediate roadmap focuses on making the generated insights even more actionable:

  1. One-Click Action: Turn solutions into ready-to-assign tasks with one click (adding owners and due dates automatically).
  2. Proactive Notifications: Implement notifications and reminders (email/Slack) for high-priority insights or overdue tasks.
  3. Enhanced Search: Add rich semantic search across transcripts, tasks, and insights.
  4. Security & Compliance: Introduce role-based access controls and tighter PII safeguards (masking/redaction at the prompt boundary).

Built With

Share this project:

Updates