Sync my progressJoin a session to keep progress across devices and share it with your instructor
Learning objectives
Frame a real work task using Trigger / Input / Task / Decision / Output / Reviewer
Rewrite a vague instruction into clear inputs, outputs, and constraints
Build a Context + Prompt Pack with Task, Context, Constraints, Output Format, and Review Criteria
Compare AI output under minimal context versus improved context
Complete one full workplace exercise with instructor guidance
Frame the Real Work Task
30 min
Frame the Real Work Task
Objective: frame a real work task using Trigger / Input / Task / Decision / Output / Reviewer.
This course does not start with "what do I want AI to do?" That question jumps straight to the tool and skips the judgment that matters most: which specific task in my work is eating time, causing errors, or needing judgment? Find that first, and you will know whether and how AI can actually help.
The task frame: six fields
Field
Meaning
Trigger
What event starts this task?
Input
What data or documents does this task actually use?
Task
What specifically needs to be done?
Decision
Where does a human need to make a judgment call?
Output
What is finally produced, and for whom?
Reviewer
Who confirms the result before it ships?
Jumping straight to the tool
"I want AI to help clean up the meeting notes": no mention of which meeting, what to extract, who it is for, or who confirms it.
Framing the task first
Trigger
The cross-department meeting ends
Input
The meeting notes
Task
Extract action items
Decision
Which items count as formal action items?
Output
An action tracker
Reviewer
The meeting organizer
Info
"Using AI" is not a task definition. A task definition answers: who needs what, under what circumstances, and who is accountable for the result.
Warning
Data safety reminder: every exercise in this course uses only the simulated scenario files the course provides. Unless authorized, never paste an organization's confidential data, customer data, or personal data into an AI tool.
Check yourself
A colleague says "can you use AI to knock this report out." Which approach best matches this course's task-framing method?
Key takeaway: Good AI collaboration starts with a clear task frame, not with "what should I tell AI to do."
Inputs, Outputs and Constraints
30 min
Define Inputs, Outputs and Constraints
Objective: rewrite a vague instruction into clear inputs, outputs, and constraints.
Once the task is framed, the next step is to spell out the input, output, and constraints: where the data comes from, what the output should look like, what must not be done, how to flag missing data, and who the final reviewer is.
Input: the source and scope of the data (e.g. "use only this meeting note," not "use whatever you know")
Output: the expected format (e.g. "a table with owner, due date, status columns")
Constraints: explicitly forbidden moves (e.g. "never guess a due date the source does not mention")
Non-goals: what this task does not include (e.g. "no commentary on meeting quality")
Review owner: who confirms the result before it ships
A vague instruction
"Help me clean up the meeting notes."
A clear task
Input
Use only this meeting's original notes
Output
Extract clear action items with owner, due date, and status columns
Constraint
Use only owners and due dates the source text supports; mark unknown when data is missing, never guess
Info
Human, AI, Human: a human defines input, output, and constraints; AI executes within those constraints; a human confirms the constraints were actually followed before it ships.
Check yourself
The meeting notes never mention a due date for one action item. Which constraint best prevents AI from guessing?
Key takeaway: Spell out input, output, and constraints clearly, and AI has a real chance of getting it right; a vague instruction only produces a vague result.
Context + Prompt Pack
35 min
Build a Context + Prompt Pack
Objective: build a Context + Prompt Pack with Task, Context, Constraints, Output Format, and Review Criteria.
A Context + Prompt Pack combines the task frame and the input/output constraints into one reusable work package. It is not a one-off chat message: it is a work spec that can be handed off and checked.
The five fields of a Context + Prompt Pack
Field
Content
Task
What specifically needs to be done (from Lesson 1's task frame)
Context
The background information and data this task needs
Constraints
What must not be done, and how to flag missing data
Output Format
The expected output structure
Review Criteria
What standard the reviewer will use to confirm the result
Info
AI 101L only teaches the basic judgment behind "work-grade context." The full context taxonomy (RAFS; Include / Retrieve / Remember / Exclude) is covered in depth in AI 102.
List what this task "definitely needs" first, then what is "helpful but optional"
Sensitive or unauthorized data never goes into context, no matter how helpful it looks
Review Criteria should let the reviewer quickly confirm whether the output meets the task requirements
Check yourself
Which field of a Context + Prompt Pack answers "how will the reviewer confirm this output is acceptable?"
Key takeaway: A Context + Prompt Pack combines the task, data, constraints, format, and review standard into one reusable, hand-off-ready work spec.
Execute and Compare
35 min
Execute and Compare
Objective: compare AI output under minimal context versus improved context, and identify where the difference actually comes from.
Run the same task twice: once with only the bare minimum instruction (minimal context), and once with a full Context + Prompt Pack (improved context). Compare the two outputs for completeness, factual support, usefulness, hidden assumptions, and how much verification work each one still needs.
Minimal Context
A short instruction missing constraints
The output may read smoothly but carry unverified assumptions
The reviewer needs much more time to check it line by line
Improved Context
A clear Task, Constraints, Output Format, and Review Criteria
The output better matches the real need, with gaps flagged as unknown
Review is more focused and verification time is shorter
Warning
The difference does not necessarily come from "a longer prompt": it comes from "the right context." Separate the two, so you can tell whether the real investment should go into preparing context rather than writing longer instructions.
Check yourself
After comparing minimal-context and improved-context outputs, the second one is clearly better. Which question best helps you understand why?
Key takeaway: A good output usually comes from good context, not a longer prompt; when comparing, ask why, not just whether the result looks good.
Guided Workplace Lab
50 min
Guided Workplace Lab
Objective: with instructor guidance, pick one scenario and complete a full "frame the task -> build a Context + Prompt Pack -> execute -> verify" exercise.
Info
Choose one of the scenarios below and complete the full workflow. All materials are simulated scenarios and contain no real organization or personal data.
Three scenario options
Scenario
Flow
A: Meeting / Staff Work
Meeting notes -> summary -> action items -> human validation
1Pick a scenario and a task Choose one scenario and frame the task with Trigger / Input / Task / Decision / Output / Reviewer.
2Build a Context + Prompt Pack Write out Task, Context, Constraints, Output Format, and Review Criteria.
3Run it once with minimal context Give only the bare-minimum instruction and note what goes wrong with the output.
4Run it again with improved context Provide the full Context + Prompt Pack and compare the two outputs.
5Verify by hand Check each item against the source and flag anything that needs a fix or confirmation.
6Write down three observations Note how the difference in context affected output quality in this exercise.
Done
Human, AI, Human: a human picks the scenario and designs the context, AI executes per instruction, and a human verifies the output and decides whether it is usable. This loop grows into a formal workflow in Module 2.
Key takeaway: Frame the task, prepare the context, execute, verify: walk through this loop once and you have a full method you can bring back to your real work.
