How to Use AI Without Losing Your Organization’s Soul

A friend recently hosted a casual dinner and one of the guests, a development director, shared a story that silenced the table. She had just found out her executive director used AI to write her performance review.

Someone asked how she knew that for certain. She laughed. “She left the prompt at the top: ‘Here are a few options that match the warm but direct message to a major gifts fundraiser about low performing Q4 results.’”

The “major gifts fundraiser?” Yep, that was her.

Later she told me she’d been reading her boss’s emails differently for months, scanning them for a change in tone. She said there was a cadence in her words that felt “performed and off base.” The prompt just confirmed what she already suspected: That AI, not her boss, had written the emails. 

What nobody at that dinner table named, but everyone felt, was the power of asymmetry. A leader used a shortcut on a direct report’s performance review, skipping the hard thinking her employee deserved while expecting the employee to respond with full human attention. That asymmetry is now everywhere and will erode organizational culture faster than any single technology.

We are living in a low-trust content era where staff quietly use ChatGPT and Claude to draft donor stewardship letters, summarize board meeting notes, write grant proposals, and respond to colleagues’ work. Most people recognize it when their colleagues don’t really read what they write or send critiques that are copy and pasted from ChatGPT. This leaves staff wondering why they should craft thoughtful responses when they don’t receive the same in return.

It’s Personal 

The implicit contract of nonprofit work — we sacrifice salary for meaning — depends on the belief that the people around you are showing up as themselves and with care. These are the organizations where a handwritten note to a major donor still represents a deliberate act of care and where a committed CEO knows the name of the volunteer who pitches in every Saturday. 

So when a fundraiser suspects the board chair’s “personal” thank you note was generated in 30 seconds, the gesture registers as dishonest. This may be expected from corporations, but it seems antithetical to the nature of nonprofit work. AI-assisted communication doesn’t exactly violate the nonprofit contract, but it certainly complicates it. And the problem isn’t only about the AI tool itself, but what the tool reveals about the people using it.

In my own consulting work and in conversations with peers across the sector, the same pattern keeps surfacing: Organizations blame the model or the platform for failing to articulate their messaging or structure. But AI exposes organizations that never had a clear strategy to begin with. 

If your nonprofit doesn’t know what it believes or what it’s trying to change, AI can’t invent that for you. It just averages the internet. Every AI-assisted email and chatbot-drafted stewardship note sounds the same: competent and forgettable unless it is tied to a distinct brand voice. 

The same goes for internal operations. You can’t encode your logic if you don’t know your logic. The vagueness that many organizations have long hidden behind has nowhere to go. Consider the executive director in the opening story. She didn’t produce a careless review because the tool failed. She produced one because she hadn’t done the thinking the tool required to create a thoughtful response. The AI just made the absence of that thinking visible.

Most of the fallible AI adoption I’ve observed in nonprofit organizations lives here. I recently consulted with an organization about their fundraising and asked to review their theory of change. What they gave me was clearly AI-generated and didn’t reflect what they actually do. When I asked about the process behind the document, they told me they had been working on it for nine months. My honest thought was that if they had just used their own minds and brainstormed as a group, they would have been done eight months ago.

Two organizations can adopt the same AI tools and get wildly different results. The one with a clear voice and a leadership team that knows what the organization stands for will use AI to move faster without losing coherence. The one without those things will produce more content, faster, and none of it will land.

This also relates to how “success” is defined. When a measure becomes a target, the economist Charles Goodhart warned, it ceases to be a good measure. This is known as Goodhart’s Law. So if your nonprofit celebrates emails sent and proposals submitted, AI will inflate those numbers. What it won’t inflate is trust, donor retention, or the feeling that someone at your organization actually cares.

The Window Is Still Open

Ten years from now, we will look back at this period the way we now look at the early days of social media: a chaotic window where everyone was experimenting without rules, and the organizations that established thoughtful norms early gained a lasting advantage.

The nonprofit sector has a real opportunity to do this right. Its relationship-first values provide an ideal framework for deciding where AI helps and where it doesn’t. Most nonprofits, approximately 85 percent, are already using generative AI tools. The question is whether they will use them in ways that reinforce what makes their work distinctive, or in ways that quietly erode it.

Organizations should start with an honest conversation about where AI adds value and where it subtracts meaning. To start, identify one category of communication in your nonprofit where AI should not be used. Not because AI can’t perform the task but because human effort is the point. This could be feedback on colleagues’ work, donor thank you calls, or your next theory of change. Pick one and hold it. That is the beginning of a norm.

And yes, the technology will keep accelerating but the norms are still ours to set.

Your boss probably does use AI to write some of her emails. Whether that should make you mad depends on whether the tool wrote what she already knew, or whether it did the knowing for her.

Look for an expanded version of this piece in the author’s Substack, Nonprofit Futures.