I first recorded a stream-of-conscience document while walking into town. I use Letterly on my iPhone, with my Air Pods. I then share the resulting transcript as-is (no cleanup) via AirDrop to my Mac.
Testing, testing. Okay. So I’m taking some notes about AI policy. One possible approach is to start off with first discussing why an AI policy is a good idea. I say policy, but really it’s like AI guidelines. Policy sounds top-down, but bottom-up works better. You really want to have something that documents not only what you shouldn’t do, but what you should do and the whys on both so everybody understands. It’s like a legal contract. If you have to actually use the punishment clauses in a contract when things go bad, I feel like the contract failed. The contract should be very clear about everybody’s rights and responsibilities. And in a way, this is sort of an agreement between the organization and everybody using AI. Here’s how we expect you to be using it. Here’s what we expect you to not be doing and the reasons behind it so everybody understands. All right. So the goal is, the goal of good policy should be that…<snip>
I then gave the resulting document to Claude (running Sonnet 4.6), and asked it to organize the thoughts into a structured document, around no more than 10 topics. I knew I needed about 10-15 slides for the presentation, so that was about the right number of topics. I reviewed the result, copied it to a markdown file, did some light edits, and then started a new Cowork session in Claude Desktop, using Open 4.6. I attached that outline, plus a Powerpoint deck for a recent presentation as an example of a format that I liked.
Here is the outline for my AI Policy meetup presentation. I’m also going to attach a Powerpoint deck that contains the style I like, plus some slides that can be re-used or slightly tweaked.
Claude reviewed everything, and had a few questions:
Before I start building, I have a few questions to make sure I get this right:
- Audience & purpose — Is this guidelines deck for presenting to your own AI for Community team (staff, volunteers, board), or is it something you’d present to external organizations you’re training? The content reads like an internal document, but I want to confirm.
- Reusable slides — Looking at your LCCF deck, several slides map closely to your guidelines content. I’m thinking of reusing/tweaking these:
Should I include all of these? Any others you specifically want carried over, or any you’d rather leave out?
- Slide 15 (Hallucination & Sycophancy) → maps to “AI can get it wrong”
- Slide 17 (Privacy & Data Security with the 3-column Generally OK / Use Caution / Not OK) → maps to “Protect sensitive data”
- Slide 10 (AI-Assisted vs AI-Generated, the 3-step process) → maps to “Follow the writing process”
- Slide 16 (Bias & Equity) → could be relevant even though it’s not explicitly in your guidelines text
- Title slide branding — Should the title slide say “AI for Community” (your org) rather than “AI for Community Foundations” (the LCCF version)? And should the subtitle reference “AI Guidelines” specifically?
- Scope — Your guidelines have three sections (Non-Negotiables, Things to Watch Out For, How We Want You to Use AI). Should each get roughly equal coverage, or do you want to emphasize certain sections more?
I answered the questions, and then asked Claude to generate the Powerpoint deck. I downloaded the result, did some tweaks, and was ready for the meetup. You can see the resulting slides and a recording of the meetup here.