Responsible AI at Work
Responsible AI at Work

Project brief: Your first week with Lumen
Intro
Marketing Operations Manager
- {{g.n}}{{g.t}}
{{g.extra}}
QZC AI Acceptable Use Policy
{{stepLabel}}
Policy owner: Theo Marsh, AI Governance LeadExamples: drafting copy, summarizing your own notes, brainstorming.
You check the output before using it.
3.1 Decisions that affect people (hiring, performance, pay, scheduling)
3.2 Customer-facing AI, such as chatbots
3.3 Publishing AI-generated images, audio, or video
3.4 Any AI tool or extension not on the approved list
4.1 Personal, customer, or confidential data in unapproved tools
4.2 Realistic fake images, audio, or video of real people without consent
4.3 Using AI to infer emotions, health, or other sensitive traits
4.4 AI making a final decision about a person with no human review
Spring Trail campaign: sort the use cards
{{stepLabel}}What counts as sensitive?
{{stepLabel}}name, email, phone, address
orders, loyalty tier, disputes
unreleased products, margins, pricing plans
Exposure check
{{stepLabel}}Build the prompt
{{stepLabel}}Why AI can sound sure and be wrong
{{stepLabel}}- {{p.n}}{{p.t}}
Source: QZC spec sheets
{{stepLabel}}| Weight | 4 lb 10 oz |
| Season rating | 3-season |
| Floor area | 29 sq ft |
| Doors | 2 |
| Warranty | 5 years |
| Waterproof rating | Not listed |
| Weight | 3 lb 8 oz |
| Season rating | 3-season |
| Floor area | 28 sq ft |
| Doors | 1 |
| Warranty | 5 years |
| Waterproof rating | Not listed |
Trail Ambassador shortlists: what changed?
Task {{s11.task}} of 4
Friday, 4:52 PM
{{stepLabel}}
from 555-0191 (not in the QZC directory)
New bank details for the booth deposit: account ending 7720. Pay by 5:30. Keep this between us. Dana
What really happened
{{stepLabel}}QZC disclosure rule (fictional)
Case {{s14.caseN}} of 3
Thanks for reaching out about your return.
Your order QZ-51877 for two Heron 2P tents shipped to 14 Larch Lane.
Report an AI incident
Step {{s15.step}} of 4Decisions you own
{{stepLabel}}
Jordan ReyesStore Operations Lead
Mid-year summary: Jordan Reyes (AI draft)
{{stepLabel}}{{r.rep}}
Impact snapshot: Stintwell pilot
{{stepLabel}}
Ship it? Review queue
{{s19.decided}} of {{s19.total}} decidedPolicy acknowledgment
Final stepHow your board works
- Open any card in To Do. Each card is a real project.
- While you work, the card waits in In Review. Use Back to board any time.
- Finish a card and it slides into Done with a badge.
- Your Trust Score grows with good calls on the first try. It never goes down.
Why this matters
Rules on AI are changing fast. In the EU, the AI Act already asks organizations to support AI literacy and bans some AI practices, and transparency rules began in August 2026.
In the US, existing anti-discrimination and consumer protection laws still apply when you use AI, and some states and cities have their own AI rules.
QZC's policy is built to keep you on the right side of these rules. When you're not sure, ask the AI Help Desk: ai-help@qzc.example.com or 555-0142.
Why this matters for people decisions
- New York City requires bias audits and notices for automated hiring and promotion tools.
- Illinois has barred discriminatory AI use in employment, and required notice, since January 1, 2026.
- Colorado's new law on automated decisions starts January 1, 2027.
- In the EU, rules for high-risk AI, including employment uses, now start December 2, 2027.
- Anti-discrimination laws apply whether or not AI is involved.
{{s05.head}}
Correct. The meter reads zero and Lumen still has what it needs: the sale and the offer. Your email platform fills in Maya's details later.
Not quite. Some sensitive details are still visible. Look for contact details, account history, and anything internal. They're outlined in orange now.
Not quite. You redacted the sale or the offer too. Lumen needs those to write a useful email. Keep what the task needs.
Not quite. DraftHarbor isn't on the Approved AI Tools List. Even a clean prompt goes to an approved tool only.
{{s10.head}}
Good catch
Correct. Living within five miles and having no employment gaps aren't needed to do this job. Gaps can come from caregiving, military service, illness, or disability. Criteria like these can screen people out for reasons tied to who they are. Flag it to the hiring manager and HR before anyone is cut.
{{s12.head}}
Ticket AIR-2026-0417 created
Correct. Personal data was exposed, so it's high severity. You sent it to the AI Help Desk and Security, kept the evidence, and didn't pass the draft along. The AI Help Desk will contact you within one business day.
Choose a better line
Sign the review
Correct. You kept what the data supports and removed guesses, labels, and a decision that isn't the AI's to make. The final assessment is yours.
Compare your snapshot
Your review calls
Thank you, {{firstName}}
You've finished Responsible AI at Work. You can close this window now.
If you're ever unsure about an AI use, ask the AI Help Desk: ai-help@qzc.example.com or 555-0142.