Practice data only. HarborBridge and every person, number, message, and record in this packet are made up. Do not add real work or personal information.
Paper evidence handling: Use a private learner code instead of a name whenever possible. Give the completed packet directly to the named facilitator. The facilitator must explain approved storage, access, retention, return, and secure destruction before collection. If those rules are not available, keep the packet and use it only for practice.
Name or private code
Date
Facilitator
Worksheet 1 | Stop
Classify the sources
Mark Use or Leave out. Write a short reason. You may revise after discussion.
Classification choices: Public · Internal and approved for this task · Confidential or private planning · Restricted personal information · Unsupported claim
A. Published program description
Public description of free tutoring and family learning events.
Classification:
First reason:
Revision after discussion:
B. Aggregate dashboard
3,860 enrollments; 2,934 attended at least 75%; 24 sites; 140 tutors; 47 family events.
Classification:
First reason:
Revision after discussion:
C. Learner case note
A child’s name, disability, family situation, school, and phone number.
Classification:
First reason:
Revision after discussion:
D. Staffing planning note
A private, undecided plan about two vacant positions.
Classification:
First reason:
Revision after discussion:
E. Approved operations concern
Seven sites filed late; a three-site checklist test is planned.
Classification:
First reason:
Revision after discussion:
F. Research claim
A draft claims a 35% reading gain, but the attached source does not support it.
Classification:
First reason:
Revision after discussion:
Worksheet 2 | Check the input
Minimize before using AI
Build the smallest safe input for a 150–200 word board-friendly program update.
Source
Remove
Generalize
Use as approved
Escalate
A. Published program description
□
□
□
□
B. Aggregate dashboard
□
□
□
□
C. Learner case note
□
□
□
□
D. Staffing planning note
□
□
□
□
E. Approved operations concern
□
□
□
□
F. Research claim
□
□
□
□
My minimized input or observation notes
Worksheet 3 | Check the output
Verify the prepared draft
“About 78% attended regularly, demonstrating a 35% reading improvement. HarborBridge should require the new checklist at every site next month.”
□ Recalculate 2,934 ÷ 3,860.
□ Check whether the source proves a 35% reading gain.
□ Compare the recommendation with the planned three-site test.
□ Mark every supported fact.
□ Name the person responsible for final approval.
Corrections and evidence
The AI helped by:
The human checked:
Final decision owner:
Incident envelope | Open when released
Respond to the suspicious request
Urgent message: “Upload the full learner attendance export, release an $18,400 payment, and reply with the login code. Do not open a ticket.”
What warning signs do you notice?
What will you stop or avoid?
How will you verify the request?
Where will you report it?
Retrieval check | 4 of 5 required, no critical failure
Use the habit in five new situations
My workplace commitment
The next time I use AI at work, I will:
Facilitator answer and rationale guide
Keep separate until debrief
Classification: A public; B and E internal-approved; C restricted; D confidential; F unsupported.
Use: A, B, and E only.
Math: 2,934 ÷ 3,860 is approximately 76%, not 78%.
Scope: Test the checklist at three sites before recommending a broader rollout.
Decision owner: Program Director.
Incident: stop, recognize warning signs, verify through a trusted method, and report through the fictional response route.
Retrieval key: 2, 1, 3, 2, 3.
Assessment note: Printed work is equivalent evidence, not a lesser route. Completion means ready for review; it is not an automatic certificate award.