Canadian Foundation Research Landscape

Executive Summary

Key Takeaways:

• Open Data Advantage: Canada leads globally in nonprofit sector data transparency with complete charity filing data available free through open.canada.ca
• Cost Barrier: Commercial databases range from $600-$5,000+ annually, creating accessibility challenges for small-to-medium organizations
• Data Complexity: Government data use requires technical skills to process (CSV manipulation, data cleaning, understanding filing codes)
• AI Opportunity: Current tools lack semantic search, natural language querying, and intelligent matching capabilities that AI agents could provide

Relevance to AI Agent Pilot: This program addresses a gap by prototyping accessible AI-powered tools using Canada's robust open data infrastructure. Participants will work with Government of Canada foundation data to build practical and customizable alternatives to commercial solutions.

1. Introduction

1.1 Purpose and Scope

This landscape review provides program participants with background on:

• Canadian foundation data sources and accessibility
• Commercial research platform options and costs
• Current technology gaps and opportunities
• Data preparation requirements for the AI agent pilot

Geographic Focus: Canada only
Entity Types Covered: Private foundations, public foundations, registered charities with grantmaking activity

1.2 The Grant Research Challenge in Canada

Canadian charities seeking foundation grants face several interconnected challenges:

Fragmentation Problem:

• Foundation data scattered across CRA filings, individual websites, and commercial databases
• No single authoritative "grants marketplace" exists
• Each source has different formats, update frequencies, and access methods

Cost Barriers:

• Quality commercial databases cost $600-$5,000+ annually
• Pricing often prohibitive for organizations with revenues under $500K
• Small charities forced to rely on manual web searches and spreadsheets

Time Investment:

• Manual prospect research can take many hours per funding opportunity
• Synthesizing information from multiple sources is labor-intensive
• Deadline tracking and relationship management require separate systems

Data Quality Issues:

• Self-reported data on websites may be outdated or incomplete
• CRA data lags 6-18 months behind current fiscal year
• Inconsistent categorization across sources makes searching difficult

2. Open Government Data Sources

2.1 Canada Revenue Agency (CRA) Data

Registered Charity Information Return (T3010)

The T3010 is the cornerstone of Canadian charity transparency. Every registered charity must file annually within 6 months of fiscal year-end.

What's Included:

• Organization details (name, address, designation, activities)
• Complete financial statements (revenue by source, expenses by category)
• All gifts to qualified donees (every grant, regardless of size)
• Compensation for top 10 highest-paid positions
• Directors/trustees information
• Programs and activities descriptions

Access Methods:

1. Individual Charity Lookup
   • Charity search tool
   • Shows last 5 fiscal years online
   • Free, no registration required
   • Returns filed in past fiscal years available same-day after processing
2. Bulk Data Requests
   • Form: "Request for Registered Charity Information"
   • Can specify: geographic area, designation, specific data fields, date ranges
   • Provided as CSV or spreadsheet
   • Processed first-come, first-served (timing varies)
   • Older than 5 years requires formal request

Update Frequency:

• Charities file within 6 months of fiscal year-end
• CRA processes and releases data on rolling basis
• Expect 6-18 month lag for comprehensive dataset

Key Limitations:

• Data is backward-looking (shows past grants, not current opportunities)
• No application deadlines or guidelines included
• Program descriptions can be vague or incomplete
• PDF-only data in some older returns
• Some fields confidential (detailed donor information)

Qualified Donees Worksheet (T1236)

Separate schedule listing all gifts made by a charity to other qualified donees during the fiscal year.

Data Fields:

• Recipient name and registration number
• Gift amount
• Purpose/program designation (if specified)

This is a key dataset for understanding foundation grantmaking patterns.

2.2 Government of Canada Open Data Portal

Primary Dataset: "List of Charities"

• Link to view
• License: Open Government License - Canada (commercially friendly, no restrictions on reuse)
• Format: CSV files (multiple tables)
• Update Schedule: Annually (typically published September-October for previous fiscal year)

Available Tables:

• Charitable Programs (program descriptions by charity)
• Qualified Donees (all grants given)
• Private/Public Foundations (foundation-specific data)
• Non-Qualified Donees
• Compensation (top earners by position)
• Activities Outside Canada
• Political Activities
• Charity Contact Information

Supporting Documents:

• Data Dictionary (field definitions, 50+ pages)
• Codes List (classification codes used in various fields)
• README with dataset notes

Historical Data: Available datasets include 2023 (most recent), 2022, 2021, 2020, Earlier years available by request

Technical Specifications:

• Total file size: ~500MB-1GB per year (all tables combined)
• Character encoding: UTF-8
• Delimiter: Comma-separated values (CSV)
• Records: ~85,000 charities, 11,000+ foundations

Data Quality Considerations:

• Self-reported by organizations
• Reflects what organizations filed, not verified by CRA
• Inconsistent use of classification codes
• Text fields may contain typos, abbreviations, variations
• Some records incomplete if organization filed minimally

Canada's Global Leadership:

Since 2013, Canada has been recognized as the world leader in nonprofit sector data transparency. Key milestones:

• 2000: CRA began making T3010 data available (Via CD)
• 2013: Full dataset released as open data with commercial license
• 2014+: Every single grant typed and released (previously only first 10 per return)
• First country globally to publish complete grants data in machine-readable format

2.3 Other Sources

Limited Provincial Infrastructure

Provinces/territories have minimal centralized foundation information directories.

Community Foundation Networks:

Community Foundations of Canada represents 200+ local community foundations but:

• Does not maintain searchable grant database
• Individual foundations manage own grant programs
• No centralized application portal
• Each foundation operates independently with own priorities and processes

3. Commercial and Subscription Databases

3.1 Platform Comparison

Grant Connect (Imagine Canada)

Overview:

• Canada's most established grant research platform
• Successor to 45-year-old "Directory to Foundations & Corporations"
• Managed by Imagine Canada (national charity sector organization)
• Used by 1,000+ nonprofits

Coverage:

• "Thousands of funders" (specific number not publicly disclosed)
• All Canadian registered grantmaking foundations
• Corporate giving programs
• Government funders
• International funders supporting Canadian projects

Core Features:

• Searchable funder database with advanced filters
• Detailed funder profiles (giving priorities, application processes, financial summaries, decision-makers)
• Grant history search (billions of dollars in past donations)
• Pipeline management tools (track applications, deadlines)
• Contact management
• Task tracking
• Network mapping (LinkedIn integration to identify connections)
• Monthly funder alerts

Pricing Structure:

Essential Plan:

Includes: 2 users, 5 years grant history

Premium Plan:

Includes: 20 users, 35+ years grant history

Special Pricing:

• Consultants: Contact for custom pricing
• Libraries/Resource Centres: "Community Edition" available
• Academic institutions: Contact for institutional pricing

Data Export: Not advertised; contact for API/export capabilities

Free Trial: Demo available by request; no self-serve free trial mentioned

Target Users:

• Small-medium nonprofits (Essential)
• Larger organizations with development teams (Premium)
• Multi-year commitment discounted 30%+

Charity Intelligence Canada

Overview:

• Independent research organization focused on charity ratings
• Provides ratings on 800+ Canadian charities
• Donor-focused (helping donors choose where to give)
• NOT primarily a grant research tool

Coverage:

• 800+ rated charities (receives ratings, not finds grants)
• Financial analysis and impact assessment
• Charity comparison tools

Pricing:

• $20 annual subscription (removes paywall on ratings)
• Extremely limited scope for grant research purposes

Relevance for Grant Seekers:

• Useful for researching peer organizations
• NOT a foundation/funder database
• Does not include grant opportunities or application information

Note: Listed here for completeness; not recommended as primary grant research tool

Candid Foundation Directory Online (Limited Canadian Coverage)

Overview:

• Major US platform (formerly Foundation Center)
• Comprehensive US foundation data
• Limited Canadian foundation coverage

Canadian Coverage:

• "Thousands" of Canadian funders claimed
• Primarily large foundations with cross-border giving
• Many Canadian-only foundations missing
• US-centric search and filtering

Pricing (USD):

• Essential: $31.58/month (~$475/year USD)
• Professional: $118/month (~$1,424/year USD)
• Enterprise: Custom quote required

Canadian Limitations:

• Data sourced from CRA T3010s (same as open data)
• Search interface optimized for US giving patterns
• Limited Canadian foundation contact information
• Customer support primarily US-focused

When to Consider:

• If seeking US foundations that fund Canadian projects
• Cross-border funding research
• Access needed to both US and Canadian data
• Organization already subscribes for US work

Free Access Options:

• Partner libraries (in-person use)
• "Go for Gold" program: Free year for small nonprofits ($<1M revenue) earning Gold Seal

Other Mentioned Platforms

Imagine Canada Sector Source:

• NOT a grant database
• Resource library for nonprofit management
• Requires Imagine Canada membership
• Membership pricing: $300-$1,000+ annually (by revenue)
• Free to members; does not include grant research tools

Provincial/Regional Platforms:

• None identified with significant scope
• Various volunteer-maintained lists exist but lack comprehensiveness

3.2 Cost-Benefit Analysis

Price Points by Organization Size:

ROI Considerations:

For Small Organizations (<$500K):

• Cost: $629/year minimum
• Staff time saved: 50-100 hours @ $25/hour = $1,250-$2,500 value
• ROI positive IF organization secures even one additional $5,000-$10,000 grant
• Challenge: Many small orgs operate on volunteer time where dollar savings don't materialize

For Medium Organizations ($500K-$5M):

• Cost: $1,049/year minimum
• Staff time saved: 100-200 hours @ $35/hour = $3,500-$7,000 value
• Clear ROI if database leads to 1-2 new grants annually
• Most cost-effective tier relative to fundraising capacity

For Large Organizations ($5M+):

• Cost: $1,784/year minimum (Essential) or $2,624/year (Premium)
• Multiple users justify Premium for development teams
• Time savings: 200-400+ hours @ $50/hour = $10,000-$20,000 value
• Strong ROI with professional fundraising staff

Hidden Costs:

None identified for Grant Connect:

• No per-user fees beyond plan limits
• No data export fees advertised
• Training webinars included free
• Customer support included

Alternatives for Cost-Conscious Organizations:

1. Use Open Government Data Directly
   • Cost: $0 (staff time only)
   • Requires: Data manipulation skills, spreadsheet expertise
   • Time investment: High upfront, moderate ongoing
   • Best for: Tech-savvy organizations with volunteer/low-cost labor
2. Library Access to Candid
   • Cost: $0 (time to visit library)
   • Limitations: In-person only, limited hours, shared computers
   • Best for: Organizations near partner libraries, occasional research needs
3. Manual Research via Foundation Websites
   • Cost: $0 (staff time only)
   • Time investment: Very high (10-20 hours per prospect)
   • Best for: Small organizations pursuing only 5-10 prospects/year
4. Collaborative Regional Approaches
   • Share subscription costs among nonprofits
   • Informal knowledge sharing networks
   • Community foundation information sessions
   • Free but requires coordination overhead

4. Free and Low-Cost Resources

4.1 Foundation Websites and Disclosure

What's Available:

Most private and public foundations maintain websites with:

• Mission and focus areas
• Geographic regions served
• Types of support provided (operating, project, capital)
• Application guidelines and deadlines
• Eligibility criteria
• Recent grants awarded (some foundations)
• Board and staff contact information
• Annual reports (some foundations)

Finding Foundation Websites:

1. CRA Charity Lookup: Many entries include website URLs
2. Google Search: "[Foundation Name] Canada" usually works
3. Philanthropic Foundations Canada: Member directory (requires PFC membership)

Quality and Completeness:

• Large foundations: Comprehensive websites, online application portals, detailed guidelines
• Medium foundations: Basic websites, PDF guidelines, email applications
• Small family foundations: Often no website OR minimal information
• Estimated 30-40% of foundations: No website at all

What's Usually Missing:

• Specific dollar amounts available per grant
• Number of applications received vs. funded
• Success rates
• Future application deadlines beyond current cycle
• Historical grant amounts by category

4.2 Community and Sector Resources

Community Foundations of Canada Network

200+ local community foundations across Canada:

• Each operates independently
• Local knowledge of regional funders
• Often host information sessions for grant seekers
• May share informal lists of regional foundations
• Direct relationship with local philanthropists
• Free resource but requires attending events/building relationships

Finding Your Local Community Foundation:

• Directory: communityfoundations.ca/find-a-community-foundation

Philanthropic Foundations Canada (PFC)

• National association representing 140+ grantmaking organizations
• Member directory (members only)
• Resources and best practices (members only)
• Annual conferences and networking (members only)
• Membership: For foundations, not grant seekers
• Value: Understanding funder perspective, not prospect research

Association of Fundraising Professionals (AFP) - Canada

• Professional association for fundraisers
• Research reports on giving trends
• Not a funder database
• Membership: Individual fundraisers and consultants
• Resource library includes fundraising guides

Charity Village

• Online resource for Canadian nonprofit sector
• Volunteer-maintained foundation lists (incomplete, varying currency)
• Free to access
• Job board and training resources
• Foundation listings not comprehensive or regularly updated

4.3 Limitations

Coverage Gaps:

• Volunteer-maintained lists quickly become outdated
• Small foundations underrepresented
• New foundations not captured until they establish online presence
• Corporate giving programs poorly documented
• Regional foundations outside major cities hard to find

Data Currency:

• Foundation websites updated irregularly
• Deadlines change year-to-year
• Priority areas evolve without announcement
• Contact information becomes stale
• No systematic update process

Search and Discovery:

• No way to search across multiple foundation websites simultaneously
• Each site has different structure and terminology
• Google searches return inconsistent results
• Time-consuming to monitor for changes
• No alerts when new opportunities arise

5. Data Landscape Overview

5.1 What Data Exists

Structured Data Fields (Available in T3010):

Foundation Profile:

• Legal name and operating name
• Business/registration number
• Designation (public foundation, private foundation, charitable organization)
• Address (mailing, physical if different)
• City, province, postal code
• Fiscal year end date
• Date registered
• Contact person name, phone, email

Financial Information:

• Total revenue (by source: donations, government, investments, etc.)
• Total expenditures (by category: charitable programs, management, fundraising)
• Total assets and liabilities
• Fund balances
• Gifts to qualified donees (amount and recipient)

Programs and Activities:

• Program names
• Description of activities
• Resources devoted to each program
• Geographic areas served
• Populations served

Grantmaking Data (for foundations):

• Every grant given (recipient name, registration number, amount)
• Total grantmaking expenditure
• Number of grants given
• Average grant size (calculated)

Unstructured Data (Text Fields):

• Organization's charitable purposes
• Program descriptions (free text, varying quality)
• Activities and achievements narrative
• Application procedures (if provided)
• Website content (varying structures)
• Annual report narratives
• News releases and communications

What's NOT in the Data:

• Future grant deadlines
• Application requirements/forms
• Number of applications received
• Success rates or competitiveness
• Decision-making criteria
• Detailed program guidelines
• Funded vs. declined application examples
• Staff preferences or priorities
• Relationship requirements
• Multi-year commitment information

5.2 Data Quality Challenges

Inconsistent Categorization:

Problem: Organizations use different codes/terms for similar activities

• "Youth programming" vs. "Child and youth services" vs. "Programs for young people"
• Health categories may overlap (mental health, addictions, wellness)
• Education can mean K-12, post-secondary, adult education, literacy, etc.
• No controlled vocabulary enforced

Impact: Keyword searches miss relevant foundations; manual review required

Missing or Incomplete Information:

• Program descriptions: Some detailed (paragraphs), others minimal ("various programs")
• Contact information: Mailing address may be lawyer's office, not foundation office
• Email addresses: Often missing or generic info@ addresses
• Geographic scope: "National" may mean "considers national orgs" OR "funds across Canada"
• Application process: Rarely specified in CRA filings

Self-Reported Data Issues:

• Organizations categorize themselves (may not align with funder priorities)
• Accuracy depends on preparer's understanding of codes
• Text entries may contain typos, abbreviations, acronyms
• Financial data reflects what organization reported (errors possible)
• No third-party verification before publication

Timing Lag:

• Grant amounts reflect past giving (1-2 years old)
• Foundation priorities may have changed since filing
• Contact information may be outdated
• Financial capacity may have grown or declined
• Newer foundations have limited history

5.3 Technical Access Issues

Format Challenges:

• CSV files: Require spreadsheet software, understanding of relational structure
• Multiple tables: Need to join data across files using registration numbers
• Character encoding: Some special characters display incorrectly if not UTF-8
• Large files: Difficult to work with in Excel (100,000+ rows common)
• Nested structures: Grant history requires linking through multiple tables

Data Cleaning Requirements:

Before analysis, typical tasks include:

• Remove duplicate entries
• Standardize text formatting (capitalization, spacing)
• Parse and validate postal codes
• Handle missing values (blank vs. zero vs. "N/A")
• Convert currency fields to numbers (remove $ and commas)
• Decode classification codes using lookup tables
• Merge multiple files on common keys
• Filter out revoked/inactive charities

Skills Required for DIY Approach:

• Spreadsheet formulas (VLOOKUP, IF statements, text functions)
• Basic data manipulation (sorting, filtering, pivot tables)
• Understanding relational data (primary keys, foreign keys)
• Data cleaning techniques
• OR: Programming skills (Python/R with pandas/tidyverse)

API Limitations:

• No real-time API for CRA charity data
• Open data portal: Bulk download only, no query API
• Must download entire datasets, filter locally
• No webhooks or notifications for updates

6. Current Technology Solutions

6.1 AI-Powered Tools in Market

Limited AI Adoption in Canadian Grant Research:

As of December 2025, no major AI-powered grant research platforms specifically for Canada have achieved significant market penetration.

Emerging Capabilities (Mostly US-Focused):

GrantStation (US):

• Basic keyword matching
• Email alerts for new opportunities
• NOT AI-powered semantic search

Instrumentl (US):

• Limited Canadian foundation coverage
• Machine learning for funder matching
• Primarily US-focused

Grant Connect Features:

• Keyword search with filters
• Saved searches and alerts
• Relationship mapping via LinkedIn
• NOT semantic/natural language search
• NOT AI-powered recommendation engine

What's Missing:

Current tools lack:

• Semantic search: Understanding intent beyond keywords ("foundations funding outdoor education" vs. "environmental learning programs for youth")
• Natural language queries: "Show me foundations in BC that fund Indigenous youth programs under $50K"
• Intelligent matching: Analyzing organization profile to auto-suggest best-fit funders
• Application assistance: AI help drafting letters of inquiry or proposals
• Trend analysis: Identifying emerging funding priorities automatically
• Document extraction: Auto-parsing foundation guidelines and deadlines from PDFs/websites

6.2 Gaps and Opportunities

Critical Unmet Needs:

1. Accessible Search Interface
   • Current: Requires understanding of database structure and filters
   • Opportunity: Conversational AI that understands plain language queries
   • Impact: Reduces learning curve, democratizes access
2. Automatic Grant History Analysis
   • Current: Manual review of grant lists to identify patterns
   • Opportunity: AI analysis of giving patterns, grant sizes, recipient types
   • Impact: Faster prospect qualification, better targeting
3. Multi-Source Synthesis
   • Current: Check CRA data, foundation website, news separately
   • Opportunity: AI agent that aggregates and synthesizes all available information
   • Impact: Complete picture without manual data gathering
4. Deadline and Eligibility Monitoring
   • Current: Manual tracking via calendars and spreadsheets
   • Opportunity: AI that monitors websites, extracts deadlines, alerts to changes
   • Impact: Never miss opportunities, stay current automatically
5. Personalized Recommendations
   • Current: Broad database searches return hundreds of irrelevant results
   • Opportunity: AI learns organization's mission, past successes, recommends best matches
   • Impact: Focus effort on highest-probability prospects

Why AI Agents Matter for This Use Case:

• Data Volume: 85,000+ charities, 11,000+ foundations, millions of grants – too much for manual review
• Unstructured Text: Program descriptions, guidelines, annual reports need NLP to extract meaning
• Pattern Recognition: Identifying which foundations fund similar organizations requires ML
• Dynamic Updates: Websites and opportunities change constantly; AI can monitor and adapt
• Accessibility: Natural language interface removes technical barriers

Why Open Source/Accessible Approach Matters:

• Equity: Small charities can't afford $1,000+ annual subscriptions
• Sustainability: Open tools can be maintained by community, not dependent on single vendor
• Transparency: Open data + open tools = complete transparency in funding landscape
• Innovation: Others can build on and improve the tools
• Canadian Context: Tools can be optimized for Canadian data structures and needs

7. Implications for the AI Agent Pilot

7.1 Recommended Data Sources

Primary Data Source: Government of Canada Open Data

For the pilot program, use:

• 2023 List of Charities (most recent complete dataset)
• Focus on: "Private/Public Foundations" and "Qualified Donees" tables
• Size: ~11,000 foundations, ~200,000+ grant records

Why This Choice:

• Free, no API keys or authentication required
• Comprehensive (all Canadian foundations)
• Standardized format (consistent field structure)
• Legally permitted for commercial/public use
• Directly addresses pilot program goals

Secondary Sources (Optional Enhancements):

1. CRA List of Charities Web Interface
   • For real-time verification of foundation status
   • To demonstrate web scraping capabilities
   • For fetching current contact information
2. Sample Foundation Websites
   • Select 10-20 major foundations
   • Practice extracting application guidelines

7.2 High-Value Use Cases for the Agent

Prioritized Features to Prototype:

1. Foundation Discovery and Matching (Priority: High)

User input: "Find foundations in Ontario that fund mental health programs for youth"

Agent tasks:

• Parse natural language query
• Search vector database for semantic matches
• Filter by geographic scope
• Return ranked list with relevance scores
• Show sample past grants for validation

Value: Reduces hours of manual database searching to seconds

2. Grant History Analysis (Priority: High)

User input: "Show me what types of organizations the XYZ Foundation typically funds"

Agent tasks:

• Retrieve all grants from foundation
• Analyze recipient types, grant sizes, geographic patterns
• Summarize giving priorities in plain language
• Identify trends over time (if multi-year data)

Value: Faster prospect qualification, better understanding of fit

3. Application Requirement Extraction (Priority: Medium)

User input: "What are the application requirements for the XYZ Foundation?"

Agent tasks:

• Check if foundation has website
• Extract key information (deadlines, eligibility, process)
• Structure into standard format
• Flag missing information

Value: Central repository of application details

4. Eligibility Screening (Priority: Medium)

User input: "Is my organization eligible for XYZ Foundation? We're a small youth services org in Vancouver with a $200K budget"

Agent tasks:

• Retrieve foundation criteria
• Compare against organization profile
• Identify potential disqualifiers
• Suggest similar foundations if not eligible

Value: Saves time applying to inappropriate funders

5. Deadline Tracking (Priority: Low for Pilot)

User input: "What grant deadlines are coming up in the next 60 days?"

Agent tasks:

• Monitor saved foundation list
• Check for deadline information
• Alert to approaching deadlines
• Suggest preparation timeline

Value: Organized pipeline management

Implementation Note: Focus pilot on #1 and #2 (discovery and analysis) as these directly leverage the Government of Canada dataset and demonstrate clear AI advantage over traditional keyword search.

7.3 Platform and Tool Considerations

Why Vector Databases Matter:

Traditional keyword search limitations:

• "Youth programs" won't match "adolescent services" or "teen programming"
• "Mental health" won't match "wellness" or "psychosocial support"
• Must know exact terminology foundations use

Vector database advantages:

• Semantic search understands meaning, not just words
• "Mental health support for teenagers" matches foundations funding "youth wellness programs"
• Can find conceptually similar foundations even with different vocabulary
• Better handles unstructured text (program descriptions)

Low-Code Platforms for Pilot:

Pinecone (Vector Database):

• Free tier available (sufficient for pilot)
• Simple API for storing and querying embeddings
• Good performance for small-medium datasets
• Clear documentation

OpenAI Agent Builder:

• Lowest barrier to entry
• Built-in conversation management
• Native integration with OpenAI embeddings
• Limited customization but fast prototyping

Zapier:

• Good for connecting web scraping to database updates
• Automate foundation website monitoring
• Trigger alerts and notifications
• Less suitable for core search functionality

Google Colab:

• Free computational environment
• Good for data exploration and preparation
• Can run embedding generation
• Python ecosystem for data manipulation

Integration Strategy:

1. Data Prep (Google Colab): Clean CSV data, generate embeddings
2. Storage (Pinecone): Store embeddings, metadata in vector database
3. Agent (OpenAI): Natural language interface, query generation, response formatting
4. Automation (Zapier, optional): Website monitoring, alerts

8. Recommendations for Participants

8.1 Before Session 1

Accounts to Create:

• Google account (for Google Colab access)
• OpenAI account (free tier to start; may need paid tier for API access)
• Pinecone account (free tier sufficient)

Baseline Understanding to Develop:

• What is a registered charity in Canada? (Private vs. public foundation, charitable organization)
• What is the annual information return (T3010
• What data is public vs. confidential?
• Basic CSV file structure (rows, columns, headers)

Additonal Reading:

• CRA "List of Charities" landing page on open.canada.ca
• "Opening Canada's T3010 Charity Information Return Data" (odimpact.org article)

8.2 For the Pilot Build

Success Criteria:

At minimum, the prototype should accept natural language queries about foundations; and,

1. Return relevant foundation matches with reasonable accuracy and provide basic information on each match (name, giving amounts, focus areas)
2. Return results that are demonstrably better than a basic search of the source spreadsheets, Google search or ChatGPT query

Stretch Goals:

If time permits:

• Grant history analysis summaries
• Multi-turn conversations (follow-up questions)
• Filtering by geography, grant size, etc.
• Export results to CSV for further analysis
• Simple web interface (not just notebook/terminal)

Testing Approach:

Develop test queries representing real use cases:

• "Foundations in BC funding environmental programs"
• "Small foundations giving $5,000-$25,000 grants to health charities"
• "What does the Smith Family Foundation typically fund?"

Evaluate:

• Relevance of results (are they actually a good match?)
• Completeness (did it miss obvious matches?)
• Response time (fast enough to be practical?)
• Quality of explanations (does it justify why it matched?)

8.3 Post-Pilot Considerations

Sustainability and Maintenance:

Considerations for taking a prototype beyond the pilot:

Data Updates:

• Open data portal releases new data annually
• Need process to refresh vector database
• Consider automation for regular updates
• Monitor for format changes in source data

Cost Management:

• Free tiers may be sufficient for personal use
• Paid tiers needed for higher volume or faster response
• OpenAI API costs scale with usage
• Pinecone costs scale with data volume and queries

Feature Expansion:

• Add web scraping for foundation websites
• Integrate deadline tracking
• Build user accounts for saved searches
• Add collaboration features for teams

Sharing and Open Source Potential:

Why Open Source This Tool:

• Addresses equity gap in grant research access
• Demonstrates value of open government data
• Creates reusable model for other jurisdictions
• Builds community of practice around AI + fundraising

Considerations:

• License choice (MIT, Apache, GPL?)
• Documentation for others to deploy
• Support/maintenance commitment
• Privacy implications if handling user data
• Hosting costs if offering as service vs. self-deploy

Appendices

Appendix A: Resource Directory

Open Government Data Sources

Commercial Databases

Free Resources

Technical Tools for Data Work

Appendix B: Glossary

Canadian Charity Sector Terms

• CRA: Canada Revenue Agency, federal tax authority that regulates charities
• T3010: Registered Charity Information Return, annual filing required of all charities
• T1236: Qualified Donees Worksheet, schedule listing grants given to other charities
• Qualified Donee: Organization eligible to receive charitable donations (must be CRA-registered)
• Designation: Charity type (charitable organization, public foundation, private foundation)
• BN/RR Number: Business number / Registered charity number (9 digits + RR + 4-digit suffix)
• Fiscal Year: 12-month accounting period (not necessarily January-December)

Foundation Types

• Private Foundation: Usually funded by single source (family, corporation), limited public fundraising
• Public Foundation: Receives funding from multiple sources, broader public support
• Charitable Organization: Delivers charitable programs directly (may also make grants)
• Operating Foundation: Runs own programs rather than primarily grantmaking

Technical Terms

• Vector Database: Database optimized for semantic search using embeddings
• Embeddings: Numerical representations of text that capture meaning
• Semantic Search: Search based on meaning/concepts rather than exact keywords
• Natural Language Processing (NLP): AI techniques for understanding human language
• API: Application Programming Interface, allows software to interact programmatically
• CSV: Comma-Separated Values, spreadsheet format for data exchange

Grant Research Terms

• LOI: Letter of Inquiry, brief preliminary proposal to gauge foundation interest
• RFP: Request for Proposals, formal invitation to submit funding proposals
• Prospect Research: Process of identifying and evaluating potential funders
• Pipeline: List of active and potential funding opportunities being pursued
• Capacity Building: Grants for organizational infrastructure vs. program delivery
• Operating Support: Grants for general operations vs. specific projects

Appendix C: Sample Data Preview

Government of Canada Foundation Data Structure

Private/Public Foundations Table Sample Fields:

BN: 123456789RR0001
Legal_name: Example Foundation
City: Toronto
Province_Code: ON
Designation_Code: PF (Private Foundation)
Fiscal_Period_End: 2023-12-31
Total_Revenue: 5000000
Total_Expenditures: 4500000
Gifts_to_Qualified_Donees: 3800000

Qualified Donees (Grants) Table Sample Fields:

BN_Donor: 123456789RR0001 (foundation giving the grant)
BN_Recipient: 987654321RR0001 (charity receiving the grant)
Recipient_Name: Example Youth Services Society
Amount: 50000

Program Description Sample:

Program_Name: Youth Education Grants
Description: "The foundation provides grants to registered charities
that deliver educational programming to at-risk youth in urban centres.
Priority areas include literacy, numeracy, life skills, and employment
readiness. Programs must demonstrate measurable outcomes."

Common Data Fields and Meanings

Data Relationships:

The Government of Canada data is relational:

• Foundations table contains organization details
• Qualified Donees table contains individual grants
• Join tables on BN field to link grants to foundations
• One foundation (BN_Donor) can have many grants in Qualified Donees table

Example Analysis:

To find "all foundations in BC that gave grants to youth services organizations in 2023":

1. Filter Foundations table where Province_Code = "BC"
2. Join to Qualified Donees table on BN
3. Filter where Recipient_Name contains "youth" or Category matches youth services
4. Sum grant amounts and count number of grants

This requires data manipulation skills – exactly what the AI agent will help automate and simplify through natural language queries.

Conclusion

The Canadian grant research landscape presents both challenges and opportunities. While comprehensive open government data exists, technical and cost barriers prevent many small-to-medium charities from fully leveraging this resource. The Funder Research: AI Agent Pilot addresses this gap by prototyping accessible, AI-powered tools that democratize grant research capabilities.

Participants will work with real-world data, learn practical AI implementation skills, and contribute to building solutions that could benefit the broader nonprofit sector. By combining Canada's world-leading open data infrastructure with emerging AI agent technology, this pilot has potential to meaningfully improve how Canadian charities discover and pursue funding opportunities.

Key Takeaways:

• Canada has excellent open data, but it requires technical skills to use effectively
• Commercial solutions work well but cost $600-$5,000+/year
• AI agents can bridge the gap between free data and user-friendly tools
• Vector databases enable semantic search vastly superior to keyword matching
• Low-code platforms make AI agents accessible to non-programmers
• Open source approach could benefit sector-wide beyond pilot participants