Collaborating on Projects

Moderate · Communication & Dissemination track · ~35 min hands-on + readings and quiz

← Back to the Communication & Dissemination track

Give your team one shared AI context — files, instructions, and reusable tools — so no one re-explains the basics.

What you’ll be able to do

  • Set up a shared ChatGPT Edu Project (files, instructions, chats) for a team
  • Build a custom GPT your team reuses
  • Apply data, consent, and access rules to shared AI use

Overview

A shared Project in ChatGPT Edu gives a team one place for related chats, files, and instructions, so everyone works from the same context and a colleague can pick up where you left off. Custom GPTs turn a good prompt into a reusable team tool, and a shared prompt library spreads what works. This is the self-paced companion to Workshop 3 — Collaboration on Projects; do that workshop for the live, hands-on version.

Shared use raises new questions: whose data is this, who agreed to it, and who has access. Keep sharing inside Drexel (“Only those invited”), keep PHI and sensitive or identifiable data out of shared files, and match the tool to Drexel’s data classification.

Practice activities

Activity 1 · Moderate — Build and share a Project

Time ~18 min · Tools ChatGPT Edu

Goal. Create a shared team workspace a colleague can step into.

Setup. A real team effort (a brief, a report, an onboarding guide). Use non-sensitive or approved files only.

Steps.

  1. Create a Project and add project instructions — the standing context everyone should get automatically:

    This project supports the DSPH [team] working on [goal]. Always write in plain, non-technical language. For summaries, use bullets and a one-line bottom line. Flag anything that needs a human decision. Never include identifying or sensitive data.

  2. Upload one or two approved reference files (a template, a style guide, a public report). Start a chat inside the project: “Using the uploaded template, draft a project update for this month.”

  3. Open share settings and invite a colleague with “Only those invited.” Choose Can chat or Can edit, and have them continue your chat.

Expected result. A shared Project your partner can pick up with the same context.

Check your work. Does the partner get your files and instructions without re-pasting anything?

Common pitfalls. Use “Only those invited,” never “Anyone with a link,” for Drexel work. No PHI or sensitive data in shared files. Keep sharing inside Drexel.

Stretch (optional). Add a second reference file and watch how the project’s answers change.

Activity 2 · Moderate — Build a reusable custom GPT

Time ~15 min · Tools ChatGPT Edu

Goal. Turn a repeated team task into a shared tool.

Setup. A task your team does over and over — a standard recap, a plain-language rewriter, a grant-deadline checker.

Steps.

  1. Create a custom GPT with clear instructions:

    You are a [role] for the DSPH [team]. When given [input], always produce [output] in this format: [structure]. Use a [tone] tone. If key information is missing, ask for it first. Never include sensitive or identifying data.

  2. Optionally add an approved reference file (a style guide or template). Test with two different inputs; refine where it drifts.

  3. Note how to share it inside your Drexel workspace — and don’t share outside Drexel.

Expected result. A working custom GPT your team can reuse.

Check your work. Does it hold the format and tone across two very different inputs?

Common pitfalls. Don’t add PHI, sensitive PII, or confidential material to GPT files. Share inside Drexel only.

Stretch (optional). Start a shared prompt library with 2–3 entries (title, prompt with placeholders, “use when”).

Activity 3 · Moderate — Give your team a shared folder-based context with Claude Cowork

Time ~20 min · Tools Claude Cowork (Claude desktop app) · a shared folder (OneDrive or Google Drive synced to your computer)

Goal. Build the same “step in without re-explaining” team context as Activity 1 — but as a shared folder that Cowork reads and writes directly, instead of files uploaded into a chat.

Setup. Data classification first, and this one is stricter than Activities 1–2: Claude and Claude Cowork are consumer/PI-purchased tools, not Drexel-supported, and are approved for Low Risk Data only unless separately approved (see the Tools page). ChatGPT Edu is the Drexel-supported tool cleared for higher-risk data — so for this activity, use only fabricated, publicly shareable content. Create a test folder in a synced OneDrive or Google Drive location (e.g., Cowork-Test) containing two small files you write yourself: team-brief.md (a one-paragraph fictional project brief, e.g., a made-up community walking-program pilot) and style-guide.md (3–4 bullets: plain language, bullets plus a one-line bottom line, flag decisions for a human). Add an empty templates subfolder.

Steps.

  1. Open the Claude desktop app, start a Cowork session, and connect it to your Cowork-Test folder. Ask it to orient itself first:

    Look through this folder. Summarize what the team brief says and list the style rules you’ll follow from the style guide.

  2. Ask for a draft that requires both files together — the brief for content, the style guide for voice:

    Using the team brief and following the style guide, draft a one-page project update for this month. Save it as a new file called project-update-july.md in the templates subfolder.

  3. Open the new file in the folder itself (Finder/Explorer or OneDrive/Drive) — not in the chat. Then edit style-guide.md yourself (add a rule like “always end with ‘Questions? Contact the project lead.’”), and ask Cowork to revise the update. Notice it picks up your change from the folder.

Expected result. A shared folder where the context lives in the files: any teammate who connects Cowork to the same synced folder gets the identical brief, style guide, and outputs automatically — no re-uploading, no invitation to a chat workspace — and Cowork’s drafts land as real files the whole team can open in Word, Drive, or anything else.

Check your work. Did the draft clearly use both files (brief content, style-guide voice)? Does project-update-july.md exist in the folder outside of any chat? Could a colleague on the shared folder open Cowork tomorrow and pick up exactly where you left off?

Common pitfalls. The big one: do not put real Drexel work in this folder unless it is Low Risk Data — for anything beyond that (student records, research data, anything sensitive), your team should default to ChatGPT Edu Projects, the Drexel-supported route. Also: Cowork can modify and create files in the connected folder, which is the point — but it means everyone on the shared folder sees its edits, so use a dedicated test/project folder, not your whole Drive. And remember the folder’s own sharing settings (who’s on the OneDrive/Drive share) now control who gets the AI context — check them the way you checked “Only those invited” in Activity 1.

Stretch (optional). Drop a real (non-sensitive, public) template into templates and ask Cowork to fill it out from the brief — then compare the experience with Activity 1: same goal, but folder-native with direct file read/write versus files uploaded into a Project’s chat. Jot one sentence on which model fits which of your team’s tasks.

Check your readiness

Answer these, then check — your score suggests whether to dive in or skim the readings first.

Useful resources