Impact Intelligence · The Foundation Knowledge Graph

Foundation strategy research in hours, not weeks.

Subscription-priced like a 990 database. Research-grade like a $150K bespoke engagement. Without either's trade-off.

See it in action below — a complete veteran-services landscape, assembled end-to-end from a foundation's strategy. The same pipeline runs for any strategy, any geography, any intervention your foundation funds.

The problem

The current approach doesn't scale.

Nonprofit landscape research is hard, slow, and expensive. Every program officer doing strategy work faces the same three options today — and each one has trade-offs they can't live with.

01 · DIY
Trawl 990s and websites by hand
Scan 990s, click through websites, build spreadsheets, chase who-knows-who calls. Weeks to months per strategy — and you still miss half the landscape. Doesn't survive a strategy shift.
02 · Subscription databases
Candid, Cause IQ, GuideStar
$1K–$20K per year. Structured 990 data, flat. No program-level distinction, no strategy-aware classification. Fine for lookup, not for landscape research.
03 · Bespoke research
Candid Custom, Bridgespan, FSG
$50K–$150K per engagement. Months to deliver. Ages the moment your strategy shifts — and you pay again to refresh.

What's actually needed: the bespoke-research output — specific, strategy-aware, explainable — at the subscription price. That's the gap. That's what we built.

The solution

AI-enabled, graph-first landscape research.

A knowledge graph assembled from IRS 990s, nonprofit websites, foundation reports, academic research, and federal rosters — queryable by strategy intent, not by keyword. Here's the 50-second explainer.

Preview
The Foundation Knowledge Graph
From strategy to landscape in hours, not weeks.
Explainer is being produced — see the interactive walkthrough below for a live version of the same pipeline, running against a real foundation strategy.
What's in the graph
1.27M grant edges · $88B traced · 718K EIN-anchored orgs · 52K funders · 1.93M IRS BMF rows (all 50 states) · 21,573 graph routers · 500+ research papers · 18 intervention taxonomies with pre-computed gap analysis for every state.
How it works
Graph-first navigation narrows the universe before any classifier runs. Enrichment (website, 990 narrative, report parsing, fact extraction) builds durable org state. Funder-walk discovery surfaces orgs the seed missed. Every signal writes durable value back to the graph.
The classification model
Three buckets, not a flat label: primary veteran org (identity is veteran), broad parent + program (veteran work is program-scoped inside a non-veteran org), adjacent signal (in the neighborhood but evidence doesn't support promotion).
Why it's better than the three boxes
Every inclusion is explainable. Every edge has provenance. Every org is EIN-anchored. Every candidate is re-enrichable when your strategy shifts. Subscription-priced delivery, bespoke-research output.
What this enables

Not a dashboard. A research layer.

The graph is the substrate. These are the things a foundation can do on top of it.

Strategy workshops
A program officer uploads a strategic plan. The system proposes a classification scheme with bespoke prompts, runs it against the narrowed universe, and produces an explainable landscape the team can iterate on — with the officer in the loop at every step.
Gap analysis
Where the money isn't going, relative to where the work — and the need — actually is. Pre-computed gap tables for all 50 states across 18 interventions, queryable by state, metro, or population.
Portfolio health
A funder's own giving history through the lens of their own strategy. Which grants map to which program areas? Where is the portfolio concentrated or thin? What are peer funders supporting that you aren't?
New-grantee discovery
Expansion beyond the names a program officer already knows. The funder → fundee loop surfaces candidates through actual grant flows, not semantic similarity. Every candidate is explainable and re-enters enrichment before promotion.
The demo

Live: Chicago veteran services, assembled from the warehouse.

Watch the graph build itself — from the intervention hub, through structural classification, enrichment, program extraction, funder flows, and loop-discovered misses. Every node and edge is a real warehouse record.

↓ Interact below ↓
Source: IRS 990 grant_edges + schedule_i_grants + website enrichment  |  Impact Intelligence Platform