AI is helping funders evaluate more ideas more fairly

The distance between a world-changing innovation and its funding often comes down to four minutes—the average time a human reviewer tends to spend on an initial grant application. In those four minutes, reviewers must assess alignment, eligibility, innovation potential, and team capacity, all while maintaining consistency across thousands of applications.

It’s an impossible ask that leads to an impossible choice: either slow down and review fewer ideas or speed up and risk missing transformative ones. At MIT Solve, we’ve spent a year exploring a third option: teaching AI to handle the repetitive parts of review so humans can invest real time where judgment matters most.

WHY AI, AND WHY NOW

In 2025, Solve received nearly 3,000 applications to our Global Challenges. Even a cursory four-minute review per application would add up to 25 full working days. Like many mission-driven organizations, we don’t want to trade rigor for speed. We want both.

That led us to a core question many funders are now asking:

“How can AI help us evaluate more opportunities, more fairly and more efficiently, without compromising judgment or values?”

To answer this question, we partnered with researchers from Harvard Business School, the University of Washington, and ESSEC Business School to study how AI could support early-stage grant review, one of the most time-intensive and high-volume stages of the funding lifecycle.

WHAT WE TESTED AND WHAT WE LEARNED

The research team developed an AI system (based on GPT-4o mini) to support application screening and tested it across reviewers with varying levels of experience. The goal was to understand where AI adds value and where it doesn’t.

Three insights stood out:

1. AI performs best on objective criteria. The system reliably assessed baseline eligibility and alignment with funding priorities, identifying whether applications met requirements or fit clearly defined geographic or programmatic focus areas.

2. AI is more helpful to less experienced reviewers. Less experienced reviewers made more consistent decisions when supported by AI insights, while experienced reviewers used AI selectively as a secondary input.

3. The biggest gain was standardization at scale. AI made judgments more consistent across reviewers, regardless of their experience, creating a stronger foundation for the second level of review and human decision-making.

HOW THIS TRANSLATES INTO REAL-WORLD IMPACT

At Solve, the first stage of our review process focuses on filtering out incomplete, ineligible, or weak-fit applications, freeing human reviewers to spend more time on the most promising ideas.

We designed our AI tool with humans firmly in the loop, focused on the repetitive, pattern-based nature of initial screening that makes it uniquely suited for AI augmentation. The tool:

1. Screens out applications with no realistic path forward.
2. Supports reviewers with a passing probability score, a clear recommendation (Pass, Fail, or Review), and a transparent explanation.

When the 2025 application cycle closed with 2,901 submissions, the system categorized them as follows: 43% Pass; 16% Fail; and 41% Review. That meant our team could focus deeply on just 41% of the applications—cutting total screening time down to ten days—while maintaining confidence in the quality of the results.

THE BIGGER TAKEAWAY FOR PHILANTHROPY

Every hour saved during the early stages of evaluation is an hour redirected toward the higher-value work that humans excel at: engaging more deeply with innovators and getting bold, under-resourced ideas one step closer to funding.

Our early results show strong alignment between AI-supported screening and human judgment. More importantly, they demonstrate that it’s possible to design AI systems that respect nuance, preserve accountability, and scale decision-making responsibly.

The philanthropic sector processes millions of applications annually, with acceptance rates often below 5%. If we’re going to reject 95% of ideas, we owe applicants—especially those historically excluded from funding—a genuine review. Dividing responsibility, with humans making decisions and AI eliminating rote review, makes it that much more possible at scale. It’s a practical step toward the thoroughness our missions demand.

 Hala Hanna is the executive director and Pooja Wagh is the director of operations and impact at MIT Solve.

View the full article