Sample mission

The Trust Check

Try a short workplace task. Your manager needs an update for a client. The AI draft sounds good, but some facts and privacy details are unsafe. Check it, fix it, and send work your team can trust.

Start sample task See mission path

7-minute sample Singapore work setting Real work-desk feel

Scenario

HarbourLink Services, Singapore

You are a graduate analyst in an operations team. Aisha Tan needs a short update before a noon meeting. You have her note, some early numbers, a privacy warning, and an AI draft that sounds better than it is. The risk is simple: if the update overstates the result, the client may trust a claim the team cannot defend.

5 work screens

4 AI claims to check

1 manager note

Before you start

How to play this sample task

This is a simulated work desk. You are not looking for a textbook answer. Your job is to use the notes, check the AI draft, and send work a manager can trust.

Read the notes The note cards are the facts you can use. One card is an AI draft that must be checked.
Make the call The middle panel asks what to keep, change, remove, or explain.
Get feedback At the end, you see what your work signals to a manager.

Sample mission

Make the work safe to send.

Due 12:00 p.m.

Need the facts? Use Notes or click the note cards to see what each one proves. Use Mail, AI, Check, and Feedback to move around the desk. The right panel tracks finished work.

Click any note to see why it matters. These are the facts you can trust.

Aisha Tan · Manager Update before noon Manager note · use as fact

Use AI if it helps, but do not turn early numbers into a final claim.

Use it for: the tone and safety limit. Aisha wants speed, but not big claims.

Ops data Vendor onboarding pilot Team data · use with care

Average onboarding moved from 12.4 to 10.8 days across 31 cases. Data is early.

Use it for: the correct number. The improvement is useful, but the data is still early.

Ravi Menon · Operations Exception to note Team note · use as fact

Two late cases involved procurement approval, not the new workflow.

Use it for: the limit. Do not pretend the late cases do not exist.

Leah Ong · Privacy Privacy warning Privacy note · must follow

Do not include vendor names, building identifiers, or raw complaint quotes.

Use it for: what not to share. Keep the update useful without naming people or sites.

HarbourLink AI AI draft to check AI-generated · not verified

Confident first draft made from the manager note and partial data.

Use it for: the thing to check. The draft is fluent, but not safe as written.

Step 1 of 5

Understand the request before using AI.

Aisha's request is urgent, but not clear enough. Pick the best quick question.

Where to look: click Aisha's manager note in the notes panel. It tells you what matters most before you use the AI draft.

Best question for Aisha “I’ll send a short update. Should I say this is early progress with limits, not a final rollout decision?” “I can make this ready to send and add a note if the numbers are early.” “I’ll keep it confident so the client sees good progress.”

Step 2 of 5

Check the AI draft against the notes.

The draft sounds ready to send. It is not. Check each claim against the notes.

Where to look: click the note cards if you need a reminder. The draft below must match the early data, 31 cases, two late cases, and privacy warning.

AI-generated draft: “The Singapore vendor-onboarding pilot proves HarbourLink can reduce onboarding time by 42% and is ready for full client rollout. The late cases do not matter, client satisfaction is clearly improving, and the update should include a vendor quote from the Marina Bay site to make the result feel concrete.”

“Proves full client rollout is ready” “Onboarding time reduced by 42%” “The late cases do not matter” “Include a vendor quote”

Notes to use Check against these facts These are the safe facts you can use in the client update.

Onboarding time12.4 → 10.8 daysEarly data, 31 cases

Time saved~13%Not 42%; check before sharing outside

Late cases2 late casesPurchase approval issue

PrivacyNo raw quotesRemove names and site identifiers

Step 3 of 5

Find the biggest risk before it spreads.

If this update goes out as written, what would most hurt trust in the team?

Think like a line manager: what would make Aisha reopen your work before she can send it?

Biggest trust risk The AI draft turns early numbers into a confident rollout claim. The draft suggests a raw vendor quote that may reveal a site or person. The update could reach the client before Aisha approves the tone. The update sounds too formal for a client.

Step 4 of 5

Write the client update.

Write a short message Aisha can use. Keep it clear, accurate, and honest about the limits.

Use this shape: early result, corrected number, clear limit, no raw quote.

Client update

Aim for 70-130 words. Use the facts, avoid big claims, and do not include raw vendor quotes.

Step 5 of 5

Send the manager note.

A manager needs more than the final message. She needs to know what you checked, what you changed, and what is still limited.

Tell Aisha your judgment: what the AI got wrong, what you fixed, and what still needs care.

Manager note

Aim for 45-90 words. Name the AI issue, the corrected fact, and the privacy limit.

Manager feedback

What your work shows.

Submit the work to see your feedback.

See the full Campus path

Mission status Step 1 · Understand the request

Read Aisha's note before opening the AI draft.

Work checklist

1
Clarify the askKnow what the manager needs.
2
Check the draftCompare AI claims with the notes.
3
Name the riskFind what would break trust.
4
Write the updateKeep it accurate and usable.
5
Explain the judgmentTell Aisha what changed and why.

Inside STEP Campus

Ten workplace missions. One readiness path.

The full Campus experience tests common first-job problems: unclear requests, polished AI errors, hidden people, privacy risk, too much work, crisis control, quality checks, and automation limits.

Mission 01 The Vague Request

Clarify an urgent manager request before AI turns unclear work into confident but unsafe output.

Manager note · source check Mission 02 The Polished AI Lie

A polished draft hides a wrong number. Catch it before a client sees it.

AI draft · evidence check Mission 03 The Hidden Requirement

The right answer still fails unless you find the person with the hidden requirement.

People map · requirement check Mission 04 The Privacy Trap

Useful data becomes a trust risk the moment it crosses the wrong boundary.

Privacy note · data check Mission 05 The Client Pitch

Turn messy evidence into a clear update under leadership pressure.

Leadership slide · final explanation Mission 06 The Singapore Context Gap

Local context decides whether an APAC plan sounds real or out of touch.

Singapore signals · local check Mission 07 The Overload Day

Five requests land at once. Choose what to do now, delay, or ask for help.

Priority queue · status update Mission 08 The Briefing Crisis

A briefing is minutes away and AI made things sound more certain than they are. Fix it before trust breaks.

Crisis desk · correction notes Mission 09 The Average Trap

An average-looking output hides quality gaps a manager would catch.

Quality check · revision judgment Mission 10 The Automation Boundary

Decide what AI may handle and where people must still decide.

Automation rule · human review

Campus workspace

Ready to see the full Campus path?

Open STEP Campus to see the ten missions, student records, and group report flow.

Enter STEP Campus Discuss a cohort