Tools
Every tool used in the labs is free and was tested in 2026, and any restriction on a “free” tool is stated. You need a laptop and a Google account. Nothing has to be purchased or installed beyond what a lab requires.
AI assistants
Use whichever you prefer; labs sometimes assign one deliberately so the class can compare.
- ChatGPT, Claude, Google Gemini
- University of Michigan tools, free with your uniqname. The university runs its own generative AI services. U-M GPT provides the U-M community with no-cost access to several chat models and image generation, and U-M Maizey builds a custom assistant over your own documents. Google Gemini and NotebookLM are also covered by U-M’s enterprise agreement when you sign in with your U-M account, which clears them for certain kinds of sensitive data that a personal account is not. For anything that touches client, partner, or otherwise sensitive information, use a U-M-covered service, sign in with your U-M account, and check the ITS guidance for which data each tool is approved to handle. Do not paste a partner’s private data into a personal chatbot account.
- Disclosure rule: every use in graded work is disclosed on the class form, and every factual claim is verified before it enters your work. This applies to the U-M tools exactly as it does to the public ones.
- Claude Console Playground is Anthropic’s page for testing requests to the Claude API, with a separate system-prompt field, model and temperature settings, token counts, and a code view of each request. You join through the class organization with Playground-only access, so there is no API key and no charge to you. Outside the class organization, API use is billed per token (current prices). It is not a U-M-covered service, so only fictional or public text goes into it.
- Week 5’s prompt-writing exercise; Anthropic’s prompt engineering overview and OpenAI’s prompt engineering guide.
AI mechanics (Week 4)
- Transformer Explainer runs a real GPT-2 in your browser, and its temperature slider changes what “the same model” says.

- 3Blue1Brown, LLMs explained briefly (video, ~8 min).
AI policy and adoption (Weeks 5–6)
- Mental Health AI Policy Project, 50-state tracker: every enacted law and active bill on AI in mental health, filterable by state, with bill text, key provisions, and enforcement authority. As of July 2026 it tracks 44 bills, 22 of them enacted laws. Check your own state before making an adopt/defer/refuse call on any AI tool that touches mental health care.
Story production (Weeks 8–11)
There are two lanes. Cloud editors upload your footage to someone else’s servers, so if your piece contains an identifiable vulnerable person, edit in the offline lane. For everything else (your own narration, staff on camera with consent, consented public footage, stock), the easy lane is fine and much faster.
Easy lane (cloud, text-based editing):
- Clipchamp, free with a Microsoft account (your U-M account works), no watermark up to 1080p, auto-captions included free. It runs in the browser with nothing to install.
- Vrew edits video like a text document: it transcribes your footage and you cut by deleting sentences. It suits interview-driven pieces. Its free tier has monthly limits on speech-to-text and AI features that change often, so check the current terms before committing a project to it.
- Whichever you use, the accessibility rule applies: auto-captions are a draft, and you hand-correct them before submission.
Offline lane (required for sensitive footage):
- Video: DaVinci Resolve, free, no watermark. The free version does not auto-generate captions, which is a Studio-only feature. Caption externally and import the .srt.
- Captions: Subtitle Edit with Whisper runs offline on the lab machines, so a vulnerable subject’s story is never uploaded to a third-party server. YouTube Studio’s auto-captions are the no-install fallback; export them and correct them by hand.
- Audio: Audacity, free and sufficient.
Either lane:
- Narration: ElevenLabs turns a script into speech, with its newest model covering more than 70 languages (model list). The free plan gives 10,000 credits a month, roughly ten minutes of speech, for non-commercial use with attribution. A video the partner organization will publish needs a paid plan, which starts at $6 a month. AI narration is labeled inside the piece under the Digital Story rules, and cloning anyone’s voice other than your own needs that person’s permission. It is not a U-M-covered service, so no client or partner information goes into it.
- Design: Canva free tier for titles and thumbnails; an organization you work with can apply for Canva for Nonprofits if it is a registered nonprofit.
- CapCut: its free tier now caps auto-captions and puts .srt export behind the paid plan, so the tools above are the safer route.
Accessibility (Week 11)

- W3C WAI: Captions/Subtitles, the captioning standard used in the course.
- Section 508 synchronized media checklist.
- NVDA, the free screen reader. Try your own story with it.
- WebAIM contrast checker for anything with text on color.
Partner work (Weeks 9–13)
- Technology readiness self-check, our ten-minute browser tool for the Week 9 capacity assessment, quoting the NCVO Digital Maturity Matrix, the Data Orchard Data Maturity Framework, and CIS Implementation Group 1.
- NTEN Tech Accelerate assessment plus their Equity Guide for Nonprofit Technology (2025), the instrument behind the Week 9 capacity assessment.
- TechSoup Digital Assessment Tool as the alternative instrument.
- Community Tool Box, ch. 3, assessing community needs and resources.
- Our Data Bodies, Digital Defense Playbook for the Week 7 lab and for partner conversations about data.
Humanitarian mapping (optional)
- Missing Maps and the HOT Tasking Manager, if you want to try humanitarian mapping on your own time; discussed in Week 12’s commons cases.
Tool churn
Free tiers and links change often, and features move behind paywalls. This page was accurate in July 2026 and is re-verified each offering. If you find a change before the instructor does, post it on the Design Wall.