Here's everything you need to know before you begin. It should only take a few minutes to read.
About This Assessment
At Klook, we use AI tools every day to work smarter. This assessment is a chance to show how you think and collaborate with AI — not just whether you know how to use it.
What you'll do
Two tasks using a real customer review dataset
Tool required
Claude Cowork (free trial available)
Time limit
None — quality over speed
What we look for
How you prompt, iterate, and reason with AI
What is Claude Cowork?
An AI workspace by Anthropic. Upload files, ask questions, and work through problems together — like having a sharp colleague available at any time.
Analyse data fastUpload a file and ask Claude to find patterns, label rows, or summarise
Structure your thinkingDescribe a problem and let Claude help break it into clear steps
Draft and refineGenerate a first draft, then improve it by giving feedback
Work back and forthIt's built for multi-step sessions, not one-shot answers
If you haven't used Claude Cowork before — read these first
💡 Free Claude Pro Trial — You need Claude Pro to access Cowork. It comes with a 30-day free trial (credit card required, no charge during trial). Cancel anytime.
🛠️
Using a different AI tool? That's totally fine.
You're free to use whichever AI tool you're most comfortable with. Just make sure it lets you export your Skill as a shareable file — this is how we verify that your workflow is reproducible and can be handed off to others, not just a one-time result.
You can complete this assessment using Coze (扣子) or Qwen (通义千问) instead.
Task 01 Same process — use any AI tool to label the CSV with Sentiment and Category columns. No change needed.
Task 02 Same — generate your merchant optimisation report using whichever AI you prefer, export as PDF.
Task 03 Instead of a Claude Skill file, submit a plain .txt file containing the complete system prompt / instructions you configured in Coze or Qwen. Your file should include: the full instructions, the expected input format (CSV with Sentiment + Category columns), and the output structure (sections, charts, recommendations).
Recommended tools: Coze (coze.cn · ByteDance) · Qwen (tongyi.aliyun.com · Alibaba). Your submission will be evaluated using the same scoring criteria.
Your Assessment Tasks
Read the background below, download your dataset, and produce three deliverables. No time limit.
Background
Every day, Klook receives millions of user reviews across thousands of activities worldwide. It's impossible to read them manually — so our teams use AI to analyse reviews at scale, across every merchant and product on the platform.
In this assessment, you'll use AI to do exactly that. You'll work with a set of fully anonymised reviews and use the reviewer's region and feedback content to identify what a merchant is doing well, where they're falling short, and how they can improve. You'll then package your analysis workflow into a reusable AI Skill — so it can be applied to any merchant in the future.
Your Input — Download First
300 anonymised customer reviews for the Klook Mt. Fuji One-Day Tour from Tokyo. This is your sole input for all three deliverables.
Use AI to add two new columns to every row in the CSV:
CategoryThe main topic of each review (e.g. guide, transport, scenery, value). Design your own scheme — there's no predefined list.
SentimentThe overall tone. Use any scale you like (e.g. Positive / Neutral / Negative).
✓ All 300 rows labelled
Submit aslabelled_dataset.csv or labelled_dataset.xlsx
02
Merchant Optimization Report
Build an AI Skill that takes your labelled dataset and generates a report for the tour operator — covering what's working, where to improve, and how to strengthen their competitive position across different reviewer regions.
✓ At least 2 charts✓ Actionable recommendations
Submit asmerchant_report.pdf
03
Exported Skill File
Export your AI Skill and upload it to us. This shows us how you designed and instructed the AI — and lets us verify that the workflow is reusable for other merchants and products.
✓ Same input always produces the full report
Submit asskill_file/ (folder or .json or .md)
Submit a plain .txt file with your complete AI system prompt.
Your .txt file should include:
The full instructions / system prompt you configured in Coze or Qwen
The expected input format (e.g. "CSV with Sentiment and Category columns")
The output structure (sections, chart descriptions, regional analysis, recommendations)
This will be scored using the same rubric as a Claude Skill file — what matters is the quality and completeness of your prompt design.
Submit Your Work
Upload your three deliverables and answer the reflection questions.
Submission received!
Thanks for completing the assessment. We'll review your work and be in touch via email.
Analyse data fast
Upload a file and ask Claude to find patterns, label rows, or summarise
Structure your thinking
Describe a problem and let Claude help break it into clear steps
Draft and refine
Generate a first draft, then improve it by giving feedback
Work back and forth
It's built for multi-step sessions, not one-shot answers
