Why Artificial Intelligence Doesn’t Work for Nonprofits (and How to Fix This)
Artificial intelligence can be a powerful tool for nonprofits. So why aren’t more organizations using it successfully? Over a year ago, my company launched one of the first platforms focused on helping nonprofits leverage AI to fundraise better, faster and more affordably. Since then, over 100 nonprofits have used our platform, and we’ve had the opportunity to observe first-hand how nonprofits apply AI to their efforts.
Our conclusion: Oftentimes, AI does not work for nonprofits. Not because of the people involved, but because of the challenges inherent in adopting new technologies. Below are the most common obstacles to AI in nonprofits and our recommendations on how to fix these problems.
1. AI isn’t always better or faster. Doing something with AI doesn’t always mean the results are superior to other ways of achieving the same outcome. Take, for example, the identification of major donors. There are many ways that AI can be used to help identify major donors: improving wealth screening, predicting the best means of making an ask, predicting which contacts look most like current major donors. However, there are other ways to accomplish each of these tasks besides AI, and without the hassle and expense of AI. AI has to be better or faster than these alternatives otherwise there’s no reason to adopt AI.
2. AI is often too much work. AI often requires already overworked nonprofit staff to do more work, both to get the recommendations from AI, and then to find a way to implement them. In particular, AI platforms that require significant setup and then require users to “pull” information from platforms will see slower adoption, even if they provide better or faster results once setup.
3. AI requires too much trust. Prioritizing a prospect or making important fundraising decisions based simply on a model generated score or label without any understanding of how the model works requires a nonprofit staff member to truly trust the AI — and perhaps put their credibility on the line. That’s often too much of an ask for a nonprofit.
4. AI often isn’t affordable. Standalone AI platforms are often one-size-fits-all with extensive capabilities, most of which are not used by the average customer. Yet, customers are often charged the same regardless of what capabilities they need and use. These fees can make AI only affordable to those nonprofits with larger budgets and the need for AI to address a variety of use cases.
5. AI often overlooks adoption. AI solutions can produce some incredibly interesting and complex insights that no other methods could uncover. However, there is a difference between interesting and actionable. The question of “So What? What Next?” is a one that is overlooked by the vast majority of AI platforms. Without knowing how to action a specific insight, ROI from AI cannot be achieved.
There are several ways to overcome these hurdles. Here are some methods that we’ve found:
1. Use AI for those fundraising applications where it can be clearly better and faster than existing methods. AI for nonprofits should address clearly defined problems where it provides value that existing methods can’t or where it has an overwhelmingly clear advantage to existing techniques. When in doubt, the simplest solution is best.
2. AI should push information to users. Rather than requiring users to “pull” insights and recommendations from a platform, AI should push actionable information to users when they need it, through existing communication channels, and in a format that integrates with existing nonprofit workflows.
3. AI should enable better decision making, not replace decision making. AI should not tell nonprofits what to do. Instead, AI should provide useful analytics, insights and predictions that help nonprofit professionals make better decisions. The future of fundraising isn’t AI, but Human-AI teams with each partner doing what they do best.
4. Adoption and AI should go hand in hand. AI should be actionable, not only interesting. A concrete adoption plan should be included in every AI project that nonprofits undertake. Adoption plans should integrate into an existing workflow and have an easily measurable ROI.
5. AI should be priced based on usage, not access. AI should be available to users priced based on what they use, and not based on merely their access to AI systems. This means AI will either have to be standalone platforms that provide usage based pricing or be integrated into existing data platforms as add-on capabilities.
Although there are many obstacles to AI, it is safe to say the future of the nonprofit sector lies in the ability to move beyond traditional fundraising methods. AI, when used properly and integrated seamlessly, can help nonprofits leverage the data they have, uncover new insights, and continue growing into the future.
Kisa Brostrom is vice president of data for boodleAI. A full-stack data professional, Kisa specializes in building data teams and infrastructure through advocacy and education. She most recently served as senior data scientist at PatientPop Inc. and has also held senior roles across organizations in aerospace, media and marketing. Kisa brings energy and insight to the democratization of data for use in dynamic business platforms that are easy to understand, apply and solve problems with.