Boon.town
TurboTax for grants — an org knowledge vault with AI grant writing on top
The Problem
The “Accidental Grant Writer” — an ED at a sub-$500K nonprofit who never wanted this job — needs grants to survive but can’t afford the $5,000 consultant. Generic AI produces slop funders catch immediately. Meanwhile, the real asset — past proposals, financials, impact reports — sits in Google Drive, never connected to the next application.
What It Does
Upload your org’s documents and the Vault extracts structured data from 27 document types automatically. A readiness gauge shows how prepared you are.
Paste a grant guidelines PDF — the system pulls out funder requirements, deadlines, evaluation criteria, and compliance rules. You confirm before generation starts.
The AI drafts submission-ready proposals grounded in your actual org context: language from proposals that won, financials that match this funder, impact data from two years ago.
Every deliverable is scored across 6 dimensions — NOFO Coverage, Budget Coherence, Voice Match, Compliance, Funder Alignment, and Org Readiness — before you see it, with a full intake-to-submission workflow and operator review gate before anything goes out.
What the AI Does
Before generation starts, the system pre-fetches the full vault context — giving the model complete access to your org’s structured data in a single pass. It parses funder requirements, maps them to your deliverables, builds budget narratives from your actual financials, and assembles everything from verified context.
Every competitor has the same foundation models. The advantage is your org’s specific context: language from proposals that won, financial patterns that match this funder, stats buried in impact reports from two years ago. TurboTax for grants — it knows your situation and walks you through it.
“Every competitor has the same foundation models. The vault is the moat — structured knowledge from your org’s actual history, financials, and impact data, not a prompt template.”
Key Design Decisions
Vault as moat, not the model
The system ingests your org’s documents and structures them into a queryable knowledge base with versioning, funder-specific variants, and a full audit trail.
Every competitor has the same foundation models. The org’s structured context is the only defensible advantage. A prompt-only approach produces generic output that funders catch immediately.
Agent-native architecture
AI capabilities are defined as markdown prompt files with a tool-based interface — not hardcoded generation chains. The agent can read/write vault data, process documents, draft proposals, and manage workflow state.
Prompt files can be iterated without deploys. Tool-based architecture means the agent’s capabilities grow with the product, not with engineering cycles.
Document intelligence pipeline
Uploaded documents are automatically classified and processed through a multi-step pipeline that figures out what each file is and extracts the relevant data — no manual tagging required.
Nonprofits don’t know what type each document is. The system has to figure it out automatically, or the vault never gets populated.
Per-deliverable pricing
$75 per Letter of Inquiry with Stripe integration. First project free for new organizations. You pay for outputs, not access.
Grant consultants charge $5,000. Per-deliverable pricing de-risks the first purchase and aligns incentives — the tool earns money when it delivers value.
Wise Caregiver design system
Literata (serif) + DM Sans (sans) typography. Teal, cream, terracotta, and sage color palette. The UI is designed to feel calm and competent, not flashy.
Grant writers are stressed and time-poor. The "Wise Caregiver" brand archetype guided every design decision — from color palette to component density to empty states.
What Happened
I designed and built a production SaaS from scratch — solo. Payments, multi-tenancy, AI agent architecture, document processing, and a real pricing model, all shipped and running. The product spans four surfaces: a dashboard with project tracking and vault readiness, an onboarding flow with EIN lookup, a vault browser for reviewing extracted data, and a project workspace where AI drafting and quality scores come together.
The “Accidental Grant Writer” framing and four validated pain points from early user conversations guided the product architecture. The TurboTax-for-grants positioning captures what it actually does.
Measuring Success
What I’m measuring once early users are onboarded.
Vault completeness rate
What percentage of structured data do orgs fill before their first generation
Document processing success rate
Percentage of uploads successfully classified and extracted without manual intervention
Generation quality scores
Average across 6 scoring dimensions before delivery
Time-to-first-draft
How fast the system produces a complete LOI from grant guidelines upload
Proposal acceptance rate
The metric that matters most to the user — does the grant get funded
Ready for more?
Prompt Theory
A Chrome extension for saving, organizing, and injecting prompts into ChatGPT, Claude, and Gemini with semantic search.
Vouch
Automated design governance using AI agents and deterministic rules. Scans live web pages against hand-authored design rules and surfaces drift.
SG Resale: Agent-Native Retail
An AI-powered resale shopping experience — browse, discover, and purchase secondhand goods through an agent-native retail interface powered by Google Gemini.
SG-art: Forced choice over stated preference
A pair-wise preference quiz over MoMA's collection. Ten rounds of forced choice between two works produces a six-axis taste profile and a personalized museum guide — no filters to set first, no sliders to adjust.