From our 2026 webinar for Nonprofit Builder, presented by AlignIQ co-founders Dan Garmat and Caroline Stauss.
Before asking how AI can help, name what it could put at risk. Then turn each value into a simple filter question.
Key takeaways
- Constraints free up experimentation. When staff know the boundaries, they can try things without asking permission every time.
- Start with risk: name what AI could put at risk for the people you serve before listing what it could do for you.
- Privacy and consent, as a filter: would a reasonable participant be surprised their information touched an AI tool?
- Equity, as a filter: could this output systematically disadvantage a group we serve?
Try this with your team
Take your organization’s top two values. For each, write one yes-or-no question someone should ask before using AI or sending an AI-assisted output.
Going further
See how your current policy holds up with the free AI Policy Health Check, or build your team’s rules with us through Ethical AI Guardrails.
More from this series
Transcript
Lightly edited from YouTube’s automatic captions for readability. Wording and order are the speakers’ own.
Policies are constraints, but they’re constraints that can free up creativity and problem solving when you’re experimenting with new technology, and free up growing fluency.
So before we talk about how AI might help, we need to name what AI could put at risk. Examples we have are privacy and consent. That came up a bit before, about using stakeholders’ names in generative AI programs. How does that relate to what your stakeholders might expect?
We put equity too, because if equity is a key value for you, AI can put it at risk with biased information. You can see how this starts connecting to AI usage itself.
One first step is to write a simple yes-or-no question you can use as a filter before using AI, or before sending an AI-generated output. There are a couple of examples on the screen. If privacy and consent is a value, a yes-or-no question would be: would a reasonable participant be surprised their information touched an AI tool? Or for equity: could this output systematically disadvantage a group we serve?