(Photo by iStock/miniseries)
Seven years ago, Becky Straw co-founded The Adventure Project, a nonprofit that works to create jobs in Africa. She was not a computer programmer, so she paid a developer $7,000 to create a website using the application Webflow. For years, she wanted to embed a customized map feature on the homepage using Mapbox to show where the organization was making progress, but she didn’t have the funds to afford more development. Earlier this year, funding to hire a developer became irrelevant. Straw didn’t know any more about website developing and coding than she did seven years earlier, but Anthropic’s AI assistant Claude did, and she used it to create the code she needed to embed Mapbox herself.
For decades, nonprofits have been beholden to commercial software companies and developers—and more recently, software as a service (SaaS) companies—to manage and support infrastructure essential to core operations. This created expensive trade-offs, including vendor lock-in, price increases, limited customization, dependence on commercial roadmaps, and occasional break down, like when the fundraising company Flipcause went bankrupt and left organizations scrambling to replace it.
AI-powered coding tools are challenging the stranglehold that software companies have traditionally had over nonprofits. Some people are calling this moment the SaaS-pocalypse. I call it Strategic Independence Day.
Vibe Coding
Straw is now “vibe coding,” which is when a non-developer uses a tool like Claude or ChatGPT and other apps to build or modify products according to their own unique needs. This democratization of coding has profound implications for nonprofits that have traditionally hired experts and licensed software to create websites, fundraise, analyze data, and manage finances.
For instance, instead of licensing an expensive customer relations management tool like Salesforce, a person can describe the tool they want, including how it should look and operate, to a generative AI or an AI coding app like Lovable, and then let it do the coding work to create exactly the spreadsheet they need to track projects and contacts. It’s the difference between buying a suit off the rack to having one tailor made.
Highly Tailored Tools
Developers are also using AI in ways that promote nonprofit efficiency and strategic independence. Aaron Suzuki, founder of the AI consultancy Prowess Consulting and a board member of the Blue Devils Drum & Bugle Corps (BDPA), offers an example. BDPA is a musical performing organization whose primary fundraising mechanism is hosting bingo games—a financially complicated operation with thousands of daily transactions to manage. Suzuki’s consulting firm, Prowess, used “agentic development”—where AI-powered tools write code, run tests, find and fix errors, and sometimes make decisions about how to implement a feature, with human oversight—to create a custom AI agent called Bingo Daily. The agent reads daily transaction spreadsheets, organizes and structures the data for automated entry into QuickBooks, and flags anomalies such as duplicates and unusually large or small transactions.
Built in less than 20 hours, the system has already reduced BDPA’s outsourced accounting costs by as much as 25 percent per month, and timely visibility into the operation’s financial performance allows leadership to close the monthly financial reports, spot trends, and make adjustments more quickly.
Another example comes from GiveDirectly, a global nonprofit that provides direct cash assistance to people in poverty around the world, where staff often have to decide very quickly whether the organization can mount an effective response to a crisis. To support this process, GiveDirectly built an AI-powered, scenario-planning tool that brings together the multiple variables GiveDirectly must consider, including the scale of need, available funding, operational capacity, access to affected communities, and whether cash assistance is appropriate based on previous organizational experience. The tool doesn’t make the decision, but it helps staff evaluate the options systematically based on past experiences.
This is not to suggest that organizations will or should stop using commercial software products, but these examples show how they can decide which systems are mission critical and worth controlling, and build and customize them independently.
Improved Platform Quality
AI-powered coding is also ushering in a new era for nonprofit technology providers. Over the past few decades, as technology became an enormous driver of economic growth, conventional wisdom dictated that only commercial technology companies had the know-how and resources to create quality products. That made sense when it took millions of dollars and many engineers to make and maintain products, but AI is erasing that barrier. Now, smaller engineering teams are using agentic coding to create products that only much larger teams could have developed in the past.
For example, using agentic coding at Every.org, a nonprofit giving platform I lead, has contributed to a 10x increase in nonprofit donations without increasing engineering headcount. Every.org is entirely supported by donations, but because of AI, its engineering capacity and development speed can match much larger commercial companies. The organization is meanwhile adding capacity in customer service and partner outreach to make sure that nonprofits and donors always have access to human support.
Other nonprofit tech providers are using agentic coding to build and sustain sophisticated, mission-critical software at commercial scale. These include Mifos X, a spinoff of the Grameen Foundation’s Mifos product that provides free, open-source financial software for microfinance efforts globally. OpenMRS is another open-source effort that provides free software to store and protect medical records. It replaces paper records and currently supports more than 22 million patients worldwide, including the treatment of people with HIV, malaria, tuberculosis, and maternal health issues.
Social good isn’t a market segment or a quaint notion; it’s an essential part of a healthy democracy and society. We need a technology ecosystem dedicated to social good not to profits.
Getting Started
Nonprofits interested in how AI can support greater technological independence can take three steps to get started. The first thing to do is set ethical and human-centered standards, and put policies in place to guide new efforts. For example, AI shouldn’t take over the management of internal operations at any organization. Organizations need to help staff understand when and how they should and shouldn’t use AI-powered code and remember that human judgment is essential to running ethical, empathetic organizations. Humans should test and review everything before implementation.
Second, organizations should identify their most critical infrastructure—the systems or products essential to achieving their mission. Maybe it’s client communications for a social welfare organization or a hospital workflow that literally impacts life and death. These systems are worth investing in, and nonprofits should develop and customize them for their own use.
From there, organizations should determine whether they want to use vibe coding to create a new iteration of a core product themselves, using generative AI platforms or coding apps, or whether they need outside help from a company proficient in agentic coding that can customize applications for them. Either way, it’s important to adopt an iterative process—to start with one part of infrastructure, and learn and build over time. There is no need to do this quickly, but it’s important to do it well.
Nonprofits no longer need to accept the trade-offs of using commercial software. The tools to build and customize applications will continue to change and evolve rapidly over the next few years, but the opportunity to claim strategic independence is here, right now.
Read more stories by Allison Fine.