Transparent about what AI can do,
and what it shouldn't
Responsible AI use means being specific and accountable for its output.
Policy last reviewed · July 2026Every workflow I build follows the same rule: AI handles the first pass, I read it, adjust it, and approve it before anything goes out. No deliverable I produce sends output directly from an AI to a funder, a donor, a board, or a client before I read and give it my personal approval. That's not an optional add-on. It's just how responsible work gets done, and I make sure it remains true for your team and organization too.
The reason is practical as much as it's ethical. When AI produces the draft and someone who understands the work reviews it, the output stays accurate, the organizational voice stays intact, and your team keeps the understanding of what was built and why. Cut the human out of that process and you get faster outputs that gradually drift from what makes your organization sound like itself. That's not what I'm here for.
I've watched this play out in workshops and client conversations. The organizations that do best with AI are the ones where at least one person genuinely understands the tool well enough to explain to a colleague what it produced and why. Building that understanding into every engagement from the start is how I work, and it's why I don't hand off a finished tool and disappear.
If all the work is handled by AI, there's no personality left. Speed isn't worth much if what gets delivered doesn't sound like you anymore.
Connor Jarvis · FounderWhatever is necessary to do the work — workflow documents, internal processes, communications templates, and any content we're building together. Nothing more. I don't access systems I don't need, and I don't hold onto files after an engagement closes.
Yes, and I tell you exactly which ones and why before we start. The AI tools I use process information to help me produce better work faster. You'll know what's being used, what data it's touching, and how to opt out of any part of that if you prefer a different approach.
I delete it. Not archived, not repurposed, not kept for portfolio or case study use without your explicit permission. If you want a written confirmation of what was shared and when it was deleted, I'll provide one.
Before any engagement begins, both parties sign a Service Agreement and a Data Processing Agreement. The Service Agreement covers scope, deliverables, and a confidentiality clause that protects everything shared during our work together. The Data Processing Agreement is specific to how AI tools are used on your data — what gets processed, how, and what happens to it after. These documents exist because accountability shouldn't be a verbal promise.
Not because AI can't produce output in these areas. It can. But there are parts of nonprofit work where the judgment, the relationship, and the community trust at stake aren't mine to run through a tool.
Messages to clients, community members, or the public during an active crisis require decisions that belong to your leadership. I won't draft those.
Anything that affects a client's access to services, resources, or support stays with a qualified person at your organization. No exceptions.
Documentation, follow-up notes, and communications in therapeutic or relational contexts shouldn't go through an AI workflow.
I can help draft the application. Who gets funded is not an AI call.
A personal note after a difficult conversation, or a follow-up after a loss, should come from a person.
This comes up more than you'd expect. An org arrives with a use case that involves sensitive client data, or a workflow that depends on relationships I wouldn't take near a tool, or a request that's more about looking like the org is using AI than actually getting something done with it. I'll say on the first call that it's not the right move. And I won't take the engagement.
If someone else is better suited to what you need, I'll say that too. The Honest Hour is supposed to give you a straight read, not close a deal.
I grew up handing out food and clothes to people who needed them. The organizations doing meaningful work, understaffed and running on grant cycles, are the ones I built this practice for. It's why I'm focused on nonprofits instead of tech companies or enterprise clients.
The human-in-the-loop principle isn't something I picked up from a framework. It comes from watching consultants hand organizations AI tools they don't understand, built during an engagement that ended when the invoice was paid. I want your team to understand what we built, ensure it will continue working, and have your team developing their own tools and workflows when I'm not around.