You got out of the dugout. You started with the work. You picked one use and turned it into a practice, with an owner, boundaries, and a written guideline.
Now we need to talk about what happens when that practice crosses the organizational boundary.
I heard a funder say something at a conference last year that I haven’t been able to shake.
“We thought it was written by AI, so we didn’t review it.”
Said matter-of-factly. As a process update.
Sit with that for a second.
Somewhere, an organization’s application, their case for the work they do, maybe their best shot at that funding cycle, didn’t get read because someone guessed at its authorship and let that guess stand in for judgment.
We should be evaluating what’s in front of us, not trying to determine who or what wrote it.
The checkbox was already the wrong question
Somewhere right now, a grant application still asks:
Did you use AI in preparing this application? Yes / No.
It felt like a reasonable question in 2024. I’m increasingly convinced it tells us very little.
AI use can mean translation. Brainstorming. A first draft. Editing. Summarizing forty pages of community feedback. Analyzing data. Drafting an entire application from an organization’s existing materials.
Those uses don’t carry the same implications for judgment, ownership, or risk.
And we’re quickly adding another category. AI systems can increasingly do things on our behalf, not just produce something for us to review. An agent might identify funding opportunities, decide which ones meet an organization’s criteria, draft an application, or eventually manage parts of the correspondence with a funder.
One checkbox collapses all of that into a single category.
What are we supposed to learn from the answer?
One funder made a different choice
I’ve been grappling with hard AI moments alongside United Way Suncoast as they navigate exactly this question.
Their community investment guidance tells applicants plainly that using tools like ChatGPT, Claude, or Copilot for grant drafting is neither endorsed nor discouraged, much as they wouldn’t discourage using spell-check.
Applicants don’t have to disclose AI use, and they aren’t penalized for it.
Instead, the guidance names what can weaken an application: generic responses, repetition of the prompt without substance, factual inaccuracies or invented examples, and formulaic writing that doesn’t reflect the organization’s actual work.
The focus is on whether the application is specific, accurate, substantive, and grounded in the organization’s experience.
Does the work hold up?
That feels considerably more useful to me than trying to determine whether AI touched it.
It also matters from an equity perspective. Smaller organizations, organizations writing in a second language, and organizations without professional grant-writing capacity may have perfectly legitimate reasons to use AI to strengthen an application.
If “sounds AI-written” becomes a reason to read something less carefully, we risk judging whether an applicant’s writing conforms to our expectations of authenticity rather than evaluating the substance of their work.
Funders are already defining disclosure differently
United Way Suncoast isn’t alone in wrestling with this, and I find the range of approaches emerging across philanthropy encouraging.
The Internet Society Foundation doesn’t prohibit applicants from using AI. Its public guidance focuses instead on the role AI should play, supporting an applicant’s work without replacing their own ideas and voice.
Wellcome asks applicants to disclose substantive generative AI assistance. Importantly, it also tells applicants why: the information is collected for monitoring and isn’t used to assess their application.
The Wenner-Gren Foundation takes another approach. It encourages applicants and grantees to disclose generative AI use in a confidential section and grounds that guidance in transparency, learning, and safety.
So we’re already seeing a range:
No required disclosure, with clear expectations for quality and accountability.
Guidance focused on preserving the applicant’s ideas and voice.
Disclosure when AI assistance becomes substantive, separated from assessment.
Encouraged disclosure as part of transparency and learning.
I don’t think we need to pick a winner.
These organizations are doing something more important: they’re defining what matters in their context and being clear with applicants about it.
That takes us back to the same work I’ve been writing about throughout this series.
Judgment.
Ownership.
Boundaries.
What are we actually trying to understand?
When a funder asks whether someone used AI, we’re usually trying to understand something deeper.
Where did judgment sit?
Did AI help polish the writing, or did it interpret the underlying information? Did someone with knowledge of the organization review what it produced? Did the final application still represent what the applicant knows, believes, and intends to say?
Who owns the result?
Someone still has to stand behind the accuracy and substance of an application. AI involvement doesn’t change that. If anything, it makes ownership worth stating more clearly.
What were the boundaries?
What information went into the system? What was AI allowed to do with it? Did sensitive grantee, financial, community, or personal information enter a tool that wasn’t appropriate for it? Was AI producing material for someone to consider, or was it making a consequential choice?
These questions tell me much more than whether AI was used.
The useful question isn’t whether AI was involved. It’s where judgment sat, who owned the result, and what boundaries governed the work.
And those aren’t particularly exotic AI questions.
They’re governance questions.
The same questions belong on our side of the table
This is where disclosure gets more interesting for philanthropy.
If we ask applicants to explain how AI shaped the work they submit, we should be equally specific about how it shapes our own work.
If AI records or transcribes a grantee conversation, does the grantee have a meaningful opportunity to consent? Do they know how the recording, transcript, or summary will be stored and used?
If AI supports due diligence or media monitoring, what sources does it rely on? Does the applicant have an opportunity to correct inaccurate or misleading information before it affects the funder’s judgment?
If AI helps draft a follow-up question, decline letter, or other applicant communication, who remains responsible for its accuracy, tone, and impact?
The relevant question is not simply whether AI was used. It is how that use affects the people on the other side of the process, and whether the funder has taken responsibility for the judgment, boundaries, and consequences involved.
As the role becomes more consequential, the governance questions become more consequential too.
Scoring applications, comparing applicants, recommending funding decisions, contacting applicants, or triggering payments each involve different levels of authority.
By the time an AI system can take an action without someone approving that specific action first, we’ve moved into a different governance conversation.
I’m going to come back to that one soon.
For now, the point is simpler.
Applicants aren’t the only people who need to be able to explain where judgment, ownership, and boundaries sit. Funders do too.
This is reporting all over again
If this sounds familiar, it should.
I’ve made a similar argument about grant reporting for years. We collect information because we want to learn something from it; drive impact.
The same discipline belongs in AI disclosure.
Before adding an AI question to an application, ask:
What are we trying to learn from the answer?
What would concern us?
What would we do differently?
Who will actually use this information?
If disclosure doesn’t affect assessment, risk, accountability, or learning, I’m not sure why we’re collecting it.
And if it does matter, a yes/no checkbox probably isn’t giving us enough information to make the judgment we’re asking it to support.
Third base: sophistication, not surveillance
The range of funder approaches emerging right now tells me something useful.
We don’t need a universal AI disclosure sentence.
We need organizations capable of defining their own practice, explaining it to the people affected by it, and revisiting it as the work changes.
That’s third base.
We started by finding where AI was already showing up. Then we made choices about individual practices and wrote those choices down. Now we’re bringing those decisions into the relationship with applicants, grantees, partners, and communities.
That requires more than transparency for transparency’s sake.
It requires clarity about judgment, ownership, and boundaries, on both sides of the relationship.
Home isn’t full AI adoption
It’s worth saying plainly because it’s easy to assume the endpoint of all this is more AI.
Home is knowing what we’re trying to accomplish, keeping meaningful human agency where the work requires it, and using technology in ways that support the relationship between funder and grantee rather than quietly redesigning it out from under both of us.
A checkbox can’t tell us whether we’re doing that.
Better questions can.
And sometimes, reading the application is still a pretty good place to start.
Ready to talk?
💬 If your applications and reports still ask “did you use AI” as a yes-or-no box, it’s worth rethinking what you’re actually trying to learn.
I work with COOs, executive leaders, boards, and grants management teams to rebuild the questions your organization asks — of applicants, of grantees, and of itself — so disclosure actually tells you something.
I take on a limited number of new engagements each quarter. Q4 is filling.
➡️ Let’s talk.
About the Author
Rachel Kimber is managing director of Full Circle Impact Solutions, where she helps foundations and nonprofits build the infrastructure to use AI effectively, ethically, and in service of mission. She works with boards, leadership teams, and grants management staff to align technology adoption with institutional values.
She is faculty for PEAK Grantmaking’s Practical AI for Grants Management Learning Series and NTEN and TAG’s certification programs. Rachel co-created the Stuck to Strategic Framework and speaks internationally on AI, governance, and organizational strategy.