The reference shelf
These pages go deeper than this one, and each owns its topic. Start here, then follow the link.
Governance & data tiers
The four data tiers, what each one allows, and the Governance Acknowledgment that turns on your seat.
Open governance →Plain-English glossary
Every term the program uses, defined in a line. Open it the moment a word stops you.
Open glossary →Key terminology
The fuller reference for AI terms, with examples drawn from DC CAP work.
Open terminology →The 4D framework
Delegation, Description, Discernment, Diligence. The four habits the program builds.
Open the 4Ds →Desktop App tour
The nine-stop, hands-on walk through the Claude Desktop App.
Take the tour →How Claude works
The mechanism in plain language, spread across the optional “Under the hood” sections of Weeks 0–2: prediction, the two sources of knowing, the window.
Start the mini-course →Practice floor: the data tiers
Hands-on drills on what enters Claude and what stays out. Ten minutes, no stakes.
Run the drills →Practice floor: the regression check
Keep a shared Skill honest as it changes: three held-out inputs, rerun on every edit. Ten minutes.
Learn the habit →Practice floor: pick your model
Which Claude model fits the task in front of you, as a two-question decision.
Open the picker →How we measure success
The measures behind the pilot and how the program tracks progress over the eight weeks.
Open the measures →Chat, Cowork, and Code
Three ways to work with Claude. Most teams live in Chat and Cowork. Code belongs to the Innovation Hub.
claude.ai
Chat
Your everyday workspace for quick drafts, one-off research, brainstorming, and personal analysis. Conversations stay private by default.
Get in: sign in at claude.ai with your DC CAP credentials.
Desktop App
Cowork
Your production studio. Claude makes real files here: Word docs, slide decks, spreadsheets, and PDFs. It runs DC CAP's Skills and links to tools like Monday.com and Google Drive.
Get in: toggle Cowork in the Desktop App, then point it at a working folder.
Terminal
Code
The Innovation Hub's build tool for software, data pipelines, and multi-agent workflows. Most staff never open it. It sits here so you know what the org can build.
Used by: the Innovation Hub, for building Skills and workflows the rest of us use.
Projects, Cowork, and SharePoint
Three places hold your work, each with one job. Your team builds shared context in Projects, Claude makes the files in Cowork, and the finished version lives in SharePoint.
Place 1
Projects: where your team thinks together
A shared workspace that carries DC CAP's context into every conversation: policies, templates, reference docs, past work. Your teammates open the same Project and build on each other's work. Put anything the team reuses here.
Place 2
Cowork: where files get made
Point Claude at a folder on your computer and it produces real deliverables: board decks, grant drafts, student communications, branded documents. Cowork also runs DC CAP's Skills.
Place 3
SharePoint: where final work lives
When a deliverable is ready for review or sharing, move it to your team's SharePoint folder. Version history, co-authoring, and access controls live here.
Your Cowork working folder is a scratchpad. Make a folder like Claude_Working on your Desktop and point Cowork there. Move finished files out to their real home, and keep Cowork away from Documents, Downloads, or anything you can't afford to lose.
SharePoint is the archive. Save finished deliverables to your team's SharePoint, in an "AI Drafts" subfolder. Version history and team review live there.
A Claude conversation is a working log. It records how you reached a draft. For the latest version of anything, go to SharePoint.
Five rules for every session
These five hold across Chat, Cowork, and Code, from your first session. They come from DC CAP's AI Governance Policy and the AI Fluency framework.
Classify before you upload
Every piece of data has a classification, so check it first. Anything with student names, grades, financial details, or other personal information is Tier 1 Restricted and stays out of Claude. When in doubt, ask.
Review before you send
Treat every output as a first draft. Check it for accuracy, tone, policy, and audience fit. Reviewing the draft is the actual work.
Disclose when it matters
When Claude helped meaningfully on something headed to funders, partners, or the public, note its role. Internal work products don't need a disclosure unless your team lead asks for one.
Push past the first draft
Claude's first answer is rarely its best. Push back, add constraints, and ask it to try again. Fluency grows in the back-and-forth.
Own what you publish
The work ships under your name. You answer for the accuracy, framing, tone, and policy of everything that leaves your desk, however it was made.
Eight ways to use Claude this semester
Every one of these clears the data rules you signed, because they all run on the same habit: the scholar never enters the chat; the situation does. Describe what is happening, keep names and records in your systems, and personalize the draft yourself.
Outreach from a situation
“A first-gen sophomore has missed our last two check-ins and seems at risk of not re-enrolling.” Describe the situation and ask for a warm, low-pressure draft to personalize. Your read of the scholar made the email work; the name and the record only added risk.
Financial aid, translated
Paste public policy language, like DCTAG rules, verification steps, or a satisfactory-academic-progress policy, and ask for a family-friendly explanation with one clear next step. Public text in, plain language out.
Partner requirements, gathered
Ask Claude to help you assemble what a partner campus asks of applicants or enrolled students, then verify against the campus’s own page before it reaches a family. Numbers, dates, and requirements always get the check.
Meeting notes into next steps
Paste your own session notes, with no identifying details, and ask for the follow-up list and a plan for the week. Your notes stay internal, and the judgment stays yours.
Workshop and family-session materials
Agendas, handouts, and slide outlines for family sessions, built from public sources. Anything carrying the DC CAP name follows the brand rules.
Patterns, not people
Ask about themes in an aggregate you are cleared to use. Watch the small-group trap from Week 1: outcomes for six scholars at one campus point back to real people even with no names attached. When a group is small enough to re-identify, describe the pattern instead of sharing the rows.
Scholarship search, curated
Give a generic profile, like a major and a class year with no individual data, and ask Claude to organize external scholarship options into a table you verify. Claude gathers; you confirm every deadline.
Your own plate
Recommendation-letter scaffolds from your talking points, a skimmable summary of a long email thread, prep notes for a partner check-in. The hours these return are the point of the whole program.
Not a coach? The same eight patterns fit your desk
Swap the nouns and the gate holds. Development: a grant-reporting Project (guidelines, past reports, style samples — never donor names or gift amounts; describe the relationship, keep the record in your systems), a monthly funder-update Skill, a skimmable summary of a long proposal thread. Finance and operations: policy language translated for staff, meeting notes into action items, a recurring report run against a checked file. The habit is identical: your situation goes in the window; the record stays in your systems.
Before your first session
New here? Week 0 walks you in, and the pilot hub lists your first-week steps: baseline assessment, Desktop App tour, the required Governance Acknowledgment, and two Anthropic courses.
Take the baseline assessment
The AI Fluency self-check. Do it before you dive in so your starting point stays honest.
Walk the Desktop App tour
Nine stops through the Claude Desktop App, plus its first-run checklist.
Read and sign the Governance Acknowledgment
The one required step. Reading the data rules and checking the box turns on your Claude seat.
Finish two Anthropic courses
Claude 101 and AI Fluency for Nonprofits. Both are free and earn certificates.
Voices, tools, and ideas
Curated places to stay sharp between sessions: practical voices, frontier updates, the official Claude docs, and how other mission-driven teams put AI to work.
Voices in AI for work
Practical thinking for knowledge workers and leaders.
On the frontier
Stay current on the edge of AI development.
Claude docs & reference
Official Anthropic documentation for getting more from Claude.