Workshop 4: AI for Documents & Visuals (Adobe)
Workshop 4: AI for Documents & Visuals
Adobe Creative Cloud Pro — Acrobat AI Assistant for PDFs and Firefly for communication graphics
Upcoming — Fall 2026 | Date TBD | 1 hour | Bring a laptop
This session is planned for the fall. The date will be posted on the home page once scheduled. The agenda below is a draft.
What You’ll Leave With
By the end of this hour, you will be able to:
- Use Acrobat AI Assistant to summarize, query, and pull from long PDFs
- Use Firefly / Adobe Express to make a clean communication graphic without design software experience
- Know what Drexel’s Moderate Risk Data classification does and doesn’t allow for Adobe tools
- Decide when an AI-generated visual is appropriate for published public health materials
- Spot representation bias in generated images and disclose AI-made visuals responsibly
Run of Show
| Time | Segment |
|---|---|
| 0:00–0:08 | Overview: Adobe at Drexel, the two AI tools, data and “commercially safe” generation |
| 0:08–0:23 | Ethics & Governance: representation, accuracy, and disclosure in AI media |
| 0:23–0:42 | Activity 1: Acrobat AI Assistant on a real PDF |
| 0:42–0:57 | Activity 2: A communication graphic in Firefly / Express |
| 0:57–1:00 | Wrap-up and where visuals fit your work |
Overview (8 min)
What you have. Drexel provides Adobe tools with AI features. Two AI capabilities matter most for this audience:
- Acrobat AI Assistant — works inside a PDF to summarize it, answer questions about it, and generate text (a summary, an email, key points) with links back to the source passages.
- Firefly / Adobe Express — generative image tools (text-to-image, generative fill and expand) and templates for flyers, infographics, social posts, and slide visuals. Drexel’s tool page notes monthly generative-credit limits.
Data rule for these tools. Drexel lists Acrobat AI Assistant, PDF Spaces, Creative Cloud, and Express as usable with Moderate Risk Data. Confirm the current classification on Drexel’s AI Tools page before putting real documents into Acrobat AI Assistant.
“Commercially safe” generation. Adobe trains Firefly on licensed and public-domain content and positions its output as commercially safe to use. That matters when a graphic ends up in a published report, grant, or public campaign. Still review any generated image for accuracy and appropriateness—generative tools can produce misleading or off-tone visuals.
Reach for Acrobat AI Assistant when the work is trapped in a PDF. Reach for Firefly / Express when you need a visual and don’t have a designer. Both are about getting a usable draft fast, then applying your judgment.
Ethics & Governance: Representation, Accuracy, and Disclosure in AI Media (15 min)
The thread continues. Earlier workshops dealt with AI text and team use. Making images and visuals adds a distinct set of risks, because pictures carry meaning—and bias—in ways that land fast and stick.
Representation bias is the big one. Image models learn from internet-scale pictures and reproduce the stereotypes in them. Ask for “a doctor,” “an epidemiologist,” “a health department director” and you’ll often get the same narrow demographic; ask for “a low-income neighborhood” or “a person struggling with addiction” and you may get stigmatizing, stereotyped scenes. In public health this is not a small problem: biased visuals can misrepresent and stigmatize the very communities we serve, and signal who “belongs” in a profession. Generate a few of these in the activity and you’ll see it directly.
Accuracy and misinformation. A generated image can depict an event that never happened or a “chart” with invented numbers; Acrobat AI Assistant can misread a table or miss a caveat in a PDF. Never present AI-generated imagery as documentary evidence, and verify any figure or claim AI pulls from a document against the source.
Dignity and consent. Avoid generating images that resemble identifiable real people or specific community members, and take extra care depicting vulnerable groups. The communities in your materials should be represented with accuracy and respect—ideally informed by the people themselves, the “nothing about us without us” principle.
Copyright and “commercially safe.” Adobe trains Firefly on licensed and public-domain content and positions output as commercially safe—a real advantage for published reports and campaigns. That is not a blank check: still review each image, don’t prompt to mimic a living artist’s style or real logos and brands, and confirm usage rights for anything consequential.
Disclosure. Label AI-generated visuals according to Drexel and your unit’s norms, especially for public-facing materials. Audiences increasingly expect to know when an image was AI-made.
Before using a generated image of people, ask: Would the community depicted recognize themselves with dignity, or as a stereotype? Generate the same prompt two or three times—if the demographics never vary, the model is showing you its bias, not reality.
Activity 1: Acrobat AI Assistant on a Real PDF
Activity 1: Acrobat AI Assistant on a Real PDF
Goal: Turn a long document into something usable in minutes.
Instructions
Open a long PDF in Acrobat — a public report, an article, a grant guideline, or a document appropriate for Adobe’s Moderate Risk Data classification. No high-risk regulated data unless Drexel explicitly approves that use case.
Open AI Assistant and try each:
- Summarize: “Give me a one-paragraph summary and then 5 key takeaways.”
- Ask: Ask a specific question the document should answer (“What is the eligibility criteria?” / “What does it say about data sharing?”). Notice it cites the passage.
- Extract: “List every deadline and requirement mentioned, as a table.”
- Repurpose: “Draft a 4-sentence email to my team summarizing what they need to know from this.”
Verify one answer against the source passage it cites. Confirm it didn’t miss or misstate anything.
Reflection
- How long would that summary or extraction have taken by hand?
- Where did it cite cleanly, and where would you double-check before trusting it?
Activity 2: A Communication Graphic in Firefly / Express
Activity 2: A Communication Graphic in Firefly / Express
Goal: Produce a clean, shareable graphic for a real communication need.
Instructions
Pick a real need: a flyer for an event, a social post, a simple infographic, or a title visual for a slide or report.
In Adobe Express, start from a template that fits the format (social, flyer, presentation).
Use generative tools where they help:
- Text-to-image for a background or illustration (“a diverse group of community members at an outdoor health fair, warm daylight, flat illustration style”).
- Generative fill / expand to adjust an existing image to the frame.
- Express’s text and layout tools to add your headline and key message.
Keep it readable: one clear message, legible text, accessible color contrast. Generative AI fills space; the communication discipline is still yours.
Reflection
- Is this good enough to use, or a starting point for a designer?
- Did any generated image misrepresent the community or topic? (A real risk worth naming.)
Generated images can include inaccurate or stereotyped depictions. For anything public-facing—especially representing communities you serve—review carefully and get a second set of eyes. Credit/disclose AI-generated visuals per Drexel and your unit’s norms.
Wrap-Up (3 min)
- Acrobat AI Assistant turns long PDFs into summaries, answers, and drafts—with sources to verify.
- Firefly / Express makes communication graphics fast, using your monthly generative credits.
- Drexel lists these tools as usable with Moderate Risk Data; confirm before using documents that contain confidential, regulated, or restricted information.
- The tools produce drafts; accuracy, tone, and representation are your call.