Fundraising is communication.

Communication is psychology.

Psychology is trait-driven.

And now… fundraising is also AI-driven.

But don’t take that from a blog post. Take it from MIT.

They ran a field experiment — not a lab simulation, not some clever survey. A real-world test with 2,310 participants creating real ad content for a real think tank. 11,138 ads were produced and served to real people. That’s ~5 million impressions worth of actual behavioral outcomes.

• Half the teams were human-human.
• Half were human + AI.

And the AI wasn’t just a passive assistant — it was a full collaborator, with randomly assigned personality traits.

What happened? Let’s just say if you’re clinging to the “AI’s not ready yet” line, you might want to sit down.

• Human-AI teams were 60% more productive per worker.  They cranked out more content, faster.

The Human-AI teams also had more back and forth communication with less of it being fluff and far fewer edits on copy.  So more colloborative “think” time = fewer edits.

• The ad copy? Better.  Human-AI teams generated stronger copy than human-only teams — confirmed by separate human and AI evaluators.
• Images? Here, humans still won. But only because the image-generation models in the study were, by today’s standards, ancient history.

As of a few days ago, OpenAI launched its new image generator which obliterates the old visual limits. That creative edge humans held? It’s gone — or about to be.

Here’s where it gets behavioral science-y. 

The researchers didn’t just compare “AI vs. no AI.” They randomized the AI’s personality traits — open, conscientious, agreeable — and matched them with different human teammates.

Sound familiar? It should. It’s the same trait-matching principle we preach in message design.

• Conscientious humans + open AI partners → higher image quality
• Extroverted humans + conscientious AI agents → everything got worse: copy, images, clicks

AI wasn’t just a tool. It was a teammate. And fit mattered. A lot.

When trait fit improved, performance improved.
When it clashed, outcomes cratered.

This is exactly the premise of our identity- and trait-based segmentation model: Fit drives results — whether it’s message-to-donor or teammate-to-human.

Field-Tested: CPCs Don’t Lie

This wasn’t theory. The ads were live. MIT ran them in-market, with real click-through rate (CTR) and cost-per-click (CPC) metrics.

The winning combo?
AI-written copy + human-created images.

Let’s say that again: AI-written messaging performed better.
Even if your gut says “humans are better at creative,” the behavior says otherwise. And the image gap is already being erased.

How Did Humans Feel About It?

Here’s another behavioral angle most people miss:

MIT surveyed participants post-task. They didn’t just want to know what worked, they wanted to know how it felt to work with a robot.

Participants rated their experience on teamwork quality and how their perceptions of collaboration changed.

• Key insight? Many people had positive experiences working with AI.
• But — and this is critical — that experience varied by trait fit.

In other words, it’s not just about whether you use AI. It’s which AI, with which person, in which context. Sound like message design again?

Final Thought: AI Is the Worst It Will Ever Be Today

The AI in this study was already outperforming humans in key areas — and it was outdated before the ink dried on the paper.

With tools like GPT-4o now generating hyper-realistic, context-aware images, the one domain where humans had an edge has closed. And fast.

AI is on a compounding improvement curve.
Humans? We’re evolutionary. No judgment — just biology.

So What’s the Play?

You don’t need to replace your fundraising team.  But if you’re not augmenting it with AI — and aligning for fit — you’re working at a disadvantage.

The future is:

• Strategy guided by behavioral science
• Messages matched to motivation
• And collaborators (human or not) chosen for trait fit, not just résumé bullets

The best teammate you’ll ever have might not need coffee breaks. But it definitely needs the right prompt.

Kevin