What we use AI for

At Liminal, AI is a research assistant and a thinking partner. It speeds up the parts that do not require judgment.

• Background research on your field and your funders
• Reading long inputs: interview transcripts, survey responses, old messaging
• Pressure-testing an argument before we bring it to you
• Drafting, rewriting, and proofreading our own work
• Analyzing data and modeling scenarios

What we do not use AI for

• Strategy. The recommendation you get comes from people who have worked inside nonprofits.
• Anything that reaches you unreviewed. A person at Liminal reads and approves every deliverable.
• Images of people. We do not generate faces or scenes that could be mistaken for your staff or the people in your programs.
• Quotes, statistics, or stories. Nothing gets invented. Every number we give you came from a source we can show you.
• Community voice. When your message needs to reflect what people said, it comes from what people said.

Who is accountable

Every deliverable has a named author at Liminal who is responsible for its accuracy and its argument. No part of your project runs on autopilot.

How we protect your information

• We use business accounts configured so client material is not used to train models
• We ask before putting confidential material into an AI tool, including donor data and unpublished financials
• If you tell us to keep something out of these tools, it stays out
• If your organization has its own AI policy, ours defers to yours

How we tell you what AI contributed

We use the AI Contribution Scale created by Blair Enns of Win Without Pitching. It describes how much AI shaped a piece of writing. Help with spelling and grammar is allowed at every level. The highest applicable level governs the whole piece.

• AI-0 Human. No AI used.
• AI-1 AI Researched. Written by a person, with AI contributing to the research.
• AI-2 AI Assisted. The ideas and arguments are ours. AI worked as an editor and thought partner.
• AI-3 Human and AI collaboration. We directed the ideas. AI drafted passages that survived into the finished piece.
• AI-4 AI Drafted. We supplied the direction and source material. AI produced most of the first draft, which we edited and verified.
• AI-5 AI Generated. AI produced the piece from a prompt, with light human editing.

Most Liminal writing lands at AI-2 or AI-3. Research documents and early drafts of long content sometimes start at AI-4. We do not send client-facing work out at AI-5. Ask where any deliverable fits within this range, and we will tell you.

The AI Contribution Scale by Blair Enns / Win Without Pitching is licensed under a Creative Commons Attribution 4.0 International License.

Bias, and how people get described

AI models learn from what has already been published, and that material often misrepresents the communities our clients serve. Ask a model to describe a neighborhood or a movement, and it reaches for the most common framing, often the exact framing your organization exists to change.

So AI never gets the last word on how people are described. Language about community comes from the community, through interviews and listening sessions, and your own staff.

Does AI change what we charge?

No. We price on the value of the work, not the hours behind it.

What it changes is how much we can do inside the same engagement: more research before we write, more drafts to react to. If it ever takes real work out of the scope, we will say so and adjust.

If you would rather we not use AI

Let us know, and we will scope your project without it and be straightforward about what those changes mean for the timeline and cost. Some organizations have funder or board requirements here, and we work within them.