How Is AI Really Working in the Social Impact Sector? New Blackbaud Institute Research Has Answers

Kate Averett Anderson
Kate Averett Anderson Blackbaud Employee
Third Anniversary Kudos 3 Name Dropper Participant
in Community News
Community-Blog-Header-1280x720.jpg

AI is showing up everywhere in the social impact sector—from fundraising and communications to operations and data analysis. But one question keeps coming up in conversations with leaders and practitioners alike: Is AI actually helping yet? 

To find out, the Blackbaud Institute surveyed thousands of social impact professionals and donors to understand how AI is really being used across organizations like yours—and where teams are still getting stuck. The result is our newest research report, Bridging the AI Effectiveness Gap.  

What we found 

The research shows that while AI adoption is widespread, confidence in its impact is far less common. Many organizations are experimenting with AI tools, but fewer feel they’re seeing clear results or return on investment.  

Across the data, four consistent gaps emerged: 

  • The Effectiveness Gap: Only about 1 in 3 organizations say they’re using AI effectively today.  
  • The Transparency Gap: Donors are open to AI, but 76% expect transparency, and only 26% feel they see it today.  
  • The Infrastructure Gap: AI use is often fragmented, with adoption outpacing organizational readiness.  
  • The Data Readiness Gap: Fewer than 20% of organizations rate their data health as excellent, limiting AI’s potential impact.  

 These gaps aren’t about lack of effort or interest. They reflect the real conditions teams are navigating as AI becomes part of everyday work. 

 

How this report can help 

The Bridging the AI Effectiveness Gap report introduces a clear framework for understanding these challenges and shows what actually changes when organizations address them intentionally. When teams bridge these gaps, AI begins to deliver clearer ROI, better productivity, and stronger trust—internally and with constituents.  

Whether you’re leading AI decisions, using AI tools day to day, or still figuring out what responsible adoption looks like, this research is designed to reflect real experiences across the sector and provide clarity on what helps AI truly work. 

 

👉 Read the full report and let us know what resonates with your experience. 

Comments

  • Teddi Taylor
    Teddi Taylor
    Second Anniversary Kudos 1 Name Dropper Facilitator 1

    I would say I am surprised to see such a positive sentiment towards AI, but we're clearly being sold on something. I was most intrigued by the 45% gap in donor vs professionals' perspective on ethical environmental impacts of AI. Whenever I hear Blackbaud (and anyone else trying to sell me on AI) talk about AI, they never talk about the ethical and environmental impacts of AI. I don't know how we can be "agents for good" when the AI data centers are having enormous negative environmental impacts on energy use (companies securing sweetheart deals with energy companies, increasing demand and passing higher costs onto consumers), water use (cooling their data centers with potable water, leading to water shortages for humans), greenhouse gas emissions (accelerating climate change), the documented cognitive decline in heavy gen-AI users, the mental health effects of gen-AI use, and the copyright infringements of AI data training. While Blackbaud claims to secure our data and that our org's data isn't use to train their models, how did their models "learn" without any access to data? Until these real concerns are addressed with the seriousness they deserve, widespread adoption of AI is unethical and antithetical to the mission of most charitable organizations.

  • Isaac Comer
    Isaac Comer
    Tenth Anniversary Kudos 1 Name Dropper Participant

    I have a lot of thoughts on "AI". (My apologies that they are out-of-order.)

    Also I have some wonderings about how the "Bridging the AI Effectiveness Gap" report was made. It says it was put together by "Blackbaud Institute". But that page has a link to Blackbaud.com so it is clearly not some sort of independent entity. It is just some arm of Blackbaud.

    From skimming the report, the one thing I definitely agree with is the need for Transparency— not just with donors, but with everything having to do with AI generally.

    I'm really curious to know how the "Institute" compiled the report. In the report I see an eyes-glaze-over number of percentages, but not much (anything?) in the way of concrete, absolute numbers.

    Speaking of gaps, I feel like the way "AI" has been presented and sold to the public leaves the whole of the "AI" industry with a large Credibility Gap.

    ———
    For what it is worth: I do suspect there is a place in the future for some kind of substantive use to come out of all of this— some of the coding harnesses show promise for example. Bug finding is useful. Helping coders code is (potentially) useful. Although even there, there is the risk that too much reliance on AI will cause a coder's skill to rust.

    If there is to be a real future for "AI", it is probably primarily in local models that are small enough to run on more reasonable hardware and therefore don't require ludicrous amounts of electricity and water cooling.

    ———

    But as the previous commenter alluded to, the AI industry has a lot to answer for. For example the insane push for a seemingly infinite number of seemingly infinitely large datacenters being built, and using up all of the world's supply of capital to try to push it out. Or the fact that much of the data for the initial training of these things was just straight up stolen. (Many number of ongoing lawsuits against OpenAI and Nvidia at least. There may be one or more against Anthropic too.)

    ———
    When it comes to work, I will use "AI" only when I deem it to be actually useful or when management explicitly requires it. I am not going to use it out of some nebulous FOMO. So far I've used Co-Pilot about six or eight times for very small "asks". It's been useful maybe four or five of those times. I'm also pretty sure chatbots specifically have a weird curve where the more you use it the less you get out of it. On some level, you're basically just talking to yourself with a math-based random word generator in between.

    (Yes I know— it seems like so much more sometimes. Do a search for "Clever Hans" and what people thought the horse was doing vs what the horse was actually doing. But if you do that search, you might not want to trust the AI generated summary. Because, again, those things are deeply unreliable. My belief is that pure LLMs (no harness) always will be, because LLMs are not genies. They are just sophisticated blobs of numbers representing words* and how they relate to each other. Sophisticated matrix multiplication plus a good algorithm (made by humans btw) allow traversal of this blob of numbers allowing next word prediction. The fact that it feels like magic is down to our own human psychology and the fundamental nature of Language itself rather than anything particularly magical about the model.)

    As for any gains that are to be had by doing things with "AI" — I suspect as people are required to pay the true cost-per-token, people's infatuation with AI as a magic everything-solver will diminish.

    ————————————-
    * the numbers actually represent "tokens" not "words". I don't want to get into the technicals on "tokens". Suffice a token is roughly equivalent to 3/4s for a word and it is the unit of measure for cloud-based LLM usage.

Categories