Nonprofits and impact organizations are using AI to tackle economic inequality with remarkable breadth, from eviction prevention to wage negotiation to helping farmers fine-tune planting schedules.

There’s just one problem: Technology is moving faster than philanthropy can keep up. And that risks the entire sector being left behind. 

When large language models went mainstream three years ago, few predicted this transformation. The ambition in the social sector today is broader and the applications are more sophisticated, all while our funding system and vision have struggled to keep up.

In GitLab Foundation’s AI for Economic Opportunity Fund, we surface 50 to 100 fundable projects every cycle and can support roughly 15. Meanwhile, the 10 largest venture-backed AI funding rounds in 2025 alone totaled $84B. The imbalance between private versus public goods investment is painful for those of us who care about social impact.

The patterns emerging from this moment should give philanthropic funders both a clearer sense of what’s possible and the confidence to move faster and deploy more. What’s apparent, though, is that technology continues to outpace institutions’ abilities to adapt.

Five Emerging Trends

At GitLab Foundation, we’ve had an unusual window into this shift, reviewing more than 1,400 applications and funding more than 50 projects over the past three years. Five trends are emerging: 

  • AI implementation is growing exponentially, but the thinking around it is too small. Applications for our fund doubled year-over-year. The ideas are getting broader and bolder. Yet, most potential solutions are still designed for thousands, not millions of beneficiaries. The ambition to scale isn’t always matching the breadth of ideas.
  • We’re still fitting AI into a human-shaped hole. The next frontier isn’t just better tools; it’s reinventing and building AI-native mission-driven organizations. What would it look like to reimagine our work with greater AI-human augmentation and radically different cost structures, delivering more services to more people at lower cost?
  • AI can now fix the thorniest legacy systems, cost-effectively. AI’s first wave in the social sector focused on building new applications. Now we’re seeing its potential inside institutions themselves. Universities, benefits agencies, and municipal governments can rapidly improve the high-friction processes that have blocked opportunity for decades. One proposal through the Foundation for California Community Colleges will help the 2.1M students in their system identify high-wage career pathways, increasing annual earnings by $12K-$20K per year.
  • New approaches only really work if the underlying data and systems are useful. Builders know a front-end product is only as good as the data behind it. The rule is simple: Garbage in, garbage out. The best projects aren’t just deploying AI, they’re investing in aggregating, cleaning, and filling gaps in key data. MIT Media Lab, for example, is using a range of data sources to identify 18M workers at risk of job loss and proactively steer them toward upskilling programs.
  • Funding isn’t keeping pace with change, with a lot of structural constraints holding it back. Many promising concepts we see are stuck in the pilot phase. That’s not because the proposals are weak, but because funders haven’t built the internal capacity to support them or evaluate them. Gaps in technical expertise, a low risk tolerance, and internal bureaucracy all slow capital down. 

The Pace Mismatch

In 2020, at the height of the pandemic, roughly 16M workers who lost their jobs had their unemployment benefits delayed for months because of outdated, Eisenhower-era computer systems.

Just two years ago, a botched rollout to a new Free Application for Federal Student Aid (FAFSA) process led to a 9% decline in applications and a 6 to 8% drop in college enrollment.  

When we move slowly and don’t do our best to support new technology, we don’t just wait. We quash opportunity. 

We need to actively think about speed as an area for intentional investment and focus. When we do, we can be proactive and fix creaky systems before they break and turn promising pilots into solutions at scale. Scholarship America, for instance, is using AI to automate and proactively award financial aid to underserved students, radically streamlining a process that has long been a barrier rather than a bridge.

When we see a solution that works in one region or geography we should seize the opportunity, because we know what happens when we don’t. People wait and opportunity lags. 

Campus Evolve has helped tens of thousands of college students with career navigation by creating an AI guide that brings together a range of data on salaries, credentials, and workforce trends in Washington state. With follow-on funding, it will now scale those efforts to millions of learners in 2027, with plans to expand to more states.

What Needs to Change

This moment calls for a shift in how philanthropy operates. We have a need for speed, diversification, and collaboration. 

Here’s how we address it:

First, we need to recognize our problem. Funding cycles are too long, solutions are moving along quickly, and technology is moving fastest of all. We need to get capital to market faster and embrace more iterative approaches. 

Second, we need funding to go beyond the application layer. That means investing in shared infrastructure—data systems, standards, and platforms—while continuing to support point solutions that demonstrate what’s possible. We can work alongside institutions to establish common rules and data-sharing agreements that make interoperability possible.

That also means expanding our theories of change. The most durable solutions can look unglamorous from the outside, like an AI tool that helps workers negotiate wages, or one that redesigns the intake process at a public benefits office, rather than the next consumer-facing app. We limit the opportunity for collective impact when we don’t zoom out. This is one of the most promising and under-resourced areas in AI-backed solutions right now.

Third, funders need to lead by example, working together, pooling efforts, and strategically aligning work to nurture the best ideas to their full potential. We should be willing to strategically collaborate and place multiple bets across the capital spectrum, from early prototyping all the way to scale.

At GitLab Foundation, we’ve helped unlock new funding for our best-performing and most promising grantees by getting other funders in the room. By going first, and taking on some of the risk that comes with it, we’ve also made it easier to say yes. Greater coordination in the field would benefit us all in terms of higher long-term impact.

Finally, we need to reframe how we think about this moment. The risks AI poses to workers are real and deserve serious attention, and we shouldn’t paper over it. The opportunity to dramatically expand economic mobility is also unprecedented, though, and treating AI as purely a threat risks squandering its potential to create opportunity. 

The Choice in Front of Us

When I look across the field, the scale of ambition among social entrepreneurs is unlike anything I’ve seen. Our job now is to move with the same velocity as this generation of builders.

The data makes one thing clear: In this environment caution is not neutral. In a world where tens of billions of dollars are being invested in private sector solutions, anything less than boldness is a decision to fall behind. 

Ellie Bertani is the CEO of GitLab Foundation and a former executive at Wells Fargo and Walmart.