What AI4Love Does With Donor Data

AI4Love reads your supporter data from your source systems, normalizes it into one bounded working base per organization, and writes the insights it generates into that same base. It does not aggregate or share your donor data across organizations. AI4Love maintains one bounded working base per organization to run its nightly analysis, held in AI4Love's Airtable account and isolated from every other organization's. It holds only the normalized activity records and insights the nightly analysis requires. It exists while your service is active plus a 90-day exit window, then it is deleted; a full export or immediate deletion is available on request at any time. A nightly sync to a file your organization owns is designed and available to be provisioned per organization; it is not yet running in production, and we will describe it in the present tense only once it is.

What Data AI4Love Reads

AI4Love processes supporter engagement data:

Donation history: Gift amounts, dates, channels, and campaign associations
Volunteer activity: Hours logged, roles, and event participation
Event attendance: Which events supporters attended and when
Communication records: Email opens, clicks, and engagement history

What Data AI4Love Does NOT Read

AI4Love does not require and does not process:

Protected health information (PHI)
Financial account numbers, credit card data, or banking details
Government-issued identifiers (SIN, SSN, driver's licence)
Biometric data
Passwords or authentication credentials from your source systems

What AI4Love Writes

AI4Love writes only to fields it owns in your Airtable base:

AI-generated insight records: Headline, pattern detected, domain, and recommendation
Enrichment fields: Postal-code-level demographic segments from Environics (PRIZM, giving propensity)

AI4Love does not write to source-of-record fields (donation amounts, contact details, engagement history). This restriction is enforced via application-level logic and monitored through audit logs.

How Data Flows Through the System

Nightly Agent Processing

Seven agents run each night to analyze your supporter data. Each agent applies deterministic mathematical rules (RFM scoring, trend calculations, threshold triggers) to identify patterns. Only after a pattern is confirmed does the system call the LLM to generate readable insight text.

The LLM receives a scoped prompt containing only the specific fields needed for that pattern — not the full supporter record. For example, an at-risk detection prompt includes the supporter's name, days since last activity, lifetime giving, and activity trend. It does not include email, phone, address, or payment information.

MCP Queries

When staff query supporter data through an AI assistant, the MCP server applies allow-list field filtering. Only explicitly approved engagement fields are returned. All other fields are blocked by default, including custom fields added after onboarding.

Source System Integrations

Platform	Direction	What AI4Love Accesses
Blackbaud RE NXT	Read-only	Constituent profiles, gift history, events
Mailchimp	Read-only	Contact lists, campaign activity
Environics	Read-only	Postal-code-level demographic segments

AI4Love never writes back to any source system. Data flows in one direction only.

Data Minimization

AI4Love applies allow-list field filtering at two boundaries:

Agent prompts: Each agent's prompt template declares exactly which fields it needs. Only those fields are included. Adding a new field to Airtable does not automatically expose it to the LLM.
MCP responses: Before passing data to an AI assistant, the MCP server filters responses to include only approved engagement fields.

This is deterministic and rule-based, not an AI judgment call.

Sub-Processors

Provider	Training on Your Data	Retention
Anthropic (Claude)	No — excluded by API terms	Up to 30 days, then deleted
OpenAI (ChatGPT)	No — excluded by API policy	Up to 30 days, then deleted

Neither provider uses your supporter data to train their models. Both are listed as sub-processors in AI4Love's Data Processing Agreement.

Data Retention

Layer	Retention
Airtable (supporter records, insights)	Indefinite — owned by your organization
Application memory	Released when serverless function completes (under 30 seconds)
Application logs	Request metadata only (no PII) — 1 hour on Vercel Pro
LLM providers	Up to 30 days for trust and safety
OAuth tokens (Nango)	Until revoked
API keys (Doppler)	Until rotated or removed

AI4Love does not store supporter data outside Airtable. Processing is transient only.

For full technical details, see the AI4Love Trust Center — Data Handling.

Common Questions

Does AI4Love sell or share our donor data? No. AI4Love does not aggregate data across organizations. Your insights come from your relationships only. No shared model learns from your donors to benefit another organization.

What happens to our data if we cancel? Nothing changes in your source systems. OAuth tokens are destroyed and API keys removed. The working base enters a 90-day exit window during which you can request reactivation, a full export (including every insight generated for you), or immediate deletion; after 90 days it is deleted.

Can AI4Love access individual payment information? No. AI4Love reads donation amounts and dates but never accesses credit card numbers, bank accounts, or payment processing details.

Does the AI learn from our donor data over time? No. Neither Anthropic nor OpenAI trains on API inputs. AI4Love does not fine-tune models on your data. Each nightly analysis runs fresh against your current data.

Can we control which fields AI4Love can see? Yes. Allow-list field filtering is deterministic and code-defined. Fields not on the allow-list are never exposed to the LLM, even if they exist in your base. Custom configurations are available during onboarding.