AI4Love
A relationship intelligence platform for non-profit organizations. AI4Love reads supporter activity overnight and shows fundraising teams who needs attention, and why.
- Role
- Founder
- Year
- 2024
- Disciplines
- Product design, Systems design, AI
- Website
- ai4love.ca
Why I built it
I have designed for non-profit organizations for most of my career. The same pattern shows up again and again. Years of relationships sit inside systems built to manage records. The records are there. The relationships are hard to see.
I founded AI4Love in 2024 to make those relationships visible. In July 2025 I wrote a note comparing two ways of thinking about it.
| The old model | The AI4Love model |
|---|---|
| The CRM manages records. | AI4Love manages relationships. |
| Data lives in systems. | Engagement lives in experiences. |
| Fundraising means money. | Generosity means money, time, loyalty and love. |
| Do more with your CRM. | Let the CRM be the CRM. AI4Love handles the connection. |
That instinct is still the product today. AI4Love is a layer that observes, interprets and amplifies generosity without disrupting the systems an organization already has.
The cut that made it a product
The first version tried to do everything. It wanted to write thank-you messages. It wanted to learn an organization’s voice, send sequences and automate outreach.
I cut all of it.
AI4Love observes and surfaces. Staff act. The first version was trying to solve for everything. The product made a decision to be one thing, and to do that one thing well.
The technical unlock
For a long time AI4Love was a database an organization would download and analyze on its own. Useful, but static.
The unlock came from somewhere unrelated. A newsletter post showed a language model working inside a spreadsheet. The post had nothing to do with non-profits. The pattern was the part that mattered. Push data to a language model. Get structured intelligence back.
I applied that pattern to a live supporter database. The first run turned 226 fictional test supporters into 897 structured insights. From that point on, AI4Love was a live intelligence layer instead of a spreadsheet.
How it works
Math first
Every night, rules look for patterns in supporter activity. Scoring, rolling averages, thresholds. This part is not AI. It is arithmetic that anyone can check.
Language second
Only after the math confirms a pattern does a language model write the insight in plain words. The model sees only what it needs for that one supporter. Every number it writes is checked against the data before anyone sees it.
People always
Staff read the insights and decide what to do. AI4Love never contacts a supporter.
The principle behind all three layers is simple. Silence is safer than noise. When the evidence is thin, AI4Love says nothing.
Designing for trust
Non-profit data is personal. Trust had to be designed in, not added at the end.
AI4Love only reads from an organization’s systems. AI4Love never writes back to them. Each organization’s data lives in its own isolated workspace, and an organization can leave with a full export at any time.
The AI4Love Trust Center publishes the details for privacy and IT reviewers. The rule for the Trust Center is strict. A claim goes on the page only when the code enforces it.
What I learned
Designing what a system should not do is the hardest part of building with AI. It is also where trust comes from. I wrote about that here.
Systems design is design. “Generosity means money, time, loyalty and love” became a single stream of participation in the database. Same idea. Different material.
A promise you can verify is worth more than a promise based on trust.
Where it is now
AI4Love is in early pilots with non-profit organizations in Ontario. The analysis runs every night.
