The SMART.AI Framework Suite
Three frameworks for nonprofits putting AI to work without spending down the trust they took years to build.
These are the July 2026 review drafts. They're complete and readable, and they aren't final. I'm collecting comments from a small group of practitioners through August 15, then publishing the finished versions in September. If you got here from a footer or a printed page, you're in the right place.
Start here
The Suite Overview runs three pages. It explains what each framework does, how the three fit together, and which one to reach for first. Read it first if you only have ten minutes.
SMART.AI: the adoption and governance spine
Six steps, discovery through stopping rules. Reach for this one when a client says they want to start using AI.
Step one isn't a policy draft. It's declaring amnesty, before you ask staff what AI they're already using. Nobody tells the truth to a process that punishes honesty, and every shadow-AI problem I've seen started with an organization that skipped that step.
Thirteen pages. Plan on 30 to 45 minutes.
SAFE.AI: the power and justice lens
Four moves for tools that land on the people your client serves. Reach for this one when a client is about to buy something that touches their community.
It gives you the questions to ask in the vendor meeting, before the contract gets signed instead of after.
Nine pages.
SCOPE.AI: the context and knowledge framework
Five moves for the question of why the output is wrong and how to fix the input. Reach for this one when a client says the AI keeps getting their stuff wrong.
Most bad AI output isn't a model problem, it's a context problem. The tool doesn't know the organization. This is also the framework where I walked back my own earlier advice on footnotes, because I got it wrong the first time. That reversal is in the document.
Ten pages.
What I'm asking reviewers for
Read SMART.AI closely and skim the other two. Then tell me three things: where a framework would fall apart in a real engagement, where the language gets in the way, and what's missing that should be there.
Comment inside the documents if you can, since that shows me exactly which sentence you mean. Email works too if that's easier. Either way, I need your notes by Saturday, August 15, so I have time to work them in before the September release.
I'm not paying reviewers and I'm not asking anyone for a testimonial. Reviewers who want credit get named in the Acknowledgments of the published version, and you're welcome to decline that. These drafts aren't public yet, so please keep the links to your own reading and hold off on sharing them until September.
Three sharp comments beat a full markup. Disagreement is the most useful thing you can send me.
How you can use these
The Suite is licensed under Creative Commons Attribution-ShareAlike 4.0 (CC BY-SA 4.0).
You can use these frameworks in your own work, including paid client engagements. Running the steps with a client counts as use, not adaptation, so it carries no obligation beyond crediting the source. You can also adapt them, translate them, shorten them, or build on them. If you share the adapted version with anyone, that version goes out under the same license and keeps the credit.
To credit the work, copy this line: Adapted from the SMART.AI Practice Suite by Gayle Roberts, CFRM. gayleroberts.com/smart Licensed CC BY-SA 4.0.
One request that isn't a license term. Please give an adapted version its own name rather than calling it SMART.AI, SAFE.AI, or SCOPE.AI. Credit this as the source so readers can tell your work from mine.