Practice Exercises

In Progress

This page is currently under active development. Content may be incomplete, evolving, or placeholder-only. Please check back later for finalized documentation and fully structured examples.

These exercises help you test dynamic threshold evaluation using memory-stored policy logic in the Threshold Memory version of the Expense Policy Decision Engine.


📁 What You’ll Need

  • Expense Policy Book (Threshold Memory Version).md (this may already be preloaded — check your workspace before uploading)

  • Any of the following receipt files:

File Name
Description
Expected Outcome

Receipt_1_Only.pdf

Blacklisted vendor: ACME Consulting

❌ Rejected

Receipt_2_Only.pdf

Equipment > $3000 default threshold

❌ Rejected

Receipt_3_Only.pdf

Blacklisted + over threshold

❌ Rejected

Receipt_4_Only.pdf

Split invoices, suspicious behavior

⚠️ Escalate

Receipt_1.pdf

Compliant meal receipt with guest names

✅ Approved

Receipt_2.pdf

Hotel < $200/night, pre-approved travel

✅ Approved

Receipt_3.pdf

Alcohol + per-person limit exceeded

❌ Rejected

This folder contains all the required files for testing the Threshold Memory template, including the .yml workflow file, the preconfigured policy book (.md), and a set of labeled receipt PDFs. Students can use these to simulate various approval scenarios and observe dynamic threshold evaluation in action.

🚀 Instructions

1. Load or Upload the Policy File

Check if Expense Policy Book (Threshold Memory Version).md is already loaded in your workspace. If not, upload it when prompted in the Start Block.

2. Upload a Receipt File

Choose one of the provided PDFs to upload as your test input.

3. Use This Prompt in Chat:

Can you review this receipt and let me know if it complies with policy?

4. Observe the Output

  • Check that the assistant uses policy_memory to reference thresholds.

  • Confirm whether it uses category-specific limits or global fallback.

  • Look for mention of blacklist flags, metadata requirements, or compliance notes.

This screenshot showcases the GLIK assistant evaluating uploaded receipts using memory-based policy thresholds. It dynamically references structured rules from the policy book, compares amounts against category-specific or global limits, and clearly explains each approval or rejection decision. Ideal for demonstrating how compliance logic is enforced through memory-scoped orchestration.

✅ Success Criteria

  • Assistant clearly identifies the matching policy rule

  • Output includes a comparison (e.g., $3,400 > $1,500 limit)

  • Final status is shown: ✅ Approved, ❌ Rejected, or ⚠️ Escalate

  • Reasoning is transparent and uses memory-based logic


Try different combinations of receipts to explore various logic paths in the workflow.

Last updated

Was this helpful?