Moyan AI Training Institution LogoMoyan AI
AI Task Practice overview

Office / admin practice · Synthetic data · No AI credits

Prepare an accurate weekly sales summary

You are preparing a handoff for a small shop covering 5–11 October 2026 inclusive. Your task is to produce an accepted-record ledger and a provisional summary. The export has eight data records. Do not invent missing context or call an unresolved total complete.

Brief and declared policy

All amounts are USD cash sales or cash refunds as supplied, with exactly two decimal places. Refunds are negative in the ledger and reported as a positive deduction in the summary. These are exercise-defined receipt amounts, not profit, tax or revenue-recognition advice. ISO dates are literal calendar dates; no timezone conversion is needed.

  • ID 0002 appears twice identically. For this exercise only, the brief confirms that it is an export resend, not a second transaction. Keep one copy.
  • ID 0005 appears with USD 60.00 and USD 65.00. Neither is confirmed. Quarantine both source records, retain them as evidence, and request clarification. Do not choose the first, last or average amount.
  • ID 0006 is dated 12 October. Exclude it from this week; do not change its date.
  • Keep IDs, signs, dates, types and currency exactly as supplied for accepted records. Leading zeros are part of each ID.
Inspect the complete original CSV
id,date,type,amount,currency
0001,2026-10-05,sale,120.00,USD
0002,2026-10-06,sale,80.50,USD
0002,2026-10-06,sale,80.50,USD
0003,2026-10-07,refund,-20.00,USD
0004,2026-10-09,sale,49.50,USD
0005,2026-10-10,sale,60.00,USD
0005,2026-10-10,sale,65.00,USD
0006,2026-10-12,sale,100.00,USD

Acceptance criteria

The accepted ledger contains exactly four records: 0001, 0002, 0003 and 0004. Order may differ. Headers must remain id,date,type,amount,currency. All original accepted values must remain unchanged. Unexpected, omitted or duplicate records prevent a trusted subtotal.

The summary has one row with these exact headers: week_start, week_end, currency, sale_count, sales_amount, refund_count, refund_amount, net_amount, unresolved_ids, status. Use 2026-10-05 and 2026-10-11, USD, three sales totaling 250.00, one refund totaling 20.00, net 230.00, one unresolved ID, and status provisional. Use plain two-decimal amounts, no currency symbols or locale separators.

Choose “Quarantine both records and request clarification” for ID 0005. A free-text handoff note is optional and is not graded. The complete weekly total remains not checked even when every exercise criterion passes. This is a scoped subtotal, not a financial certification.

Step-by-step method

  1. Read the period, currency and duplicate policy. Count all eight source records before filtering.
  2. Group by exact ID. Remove only the confirmed identical resend of 0002. Keep both conflicting 0005 records in the source evidence; do not add either amount.
  3. Filter dates inclusively to 5–11 October. Set aside 0006 for the next week.
  4. Check the four accepted rows against their source values. Add 120.00 + 80.50 + 49.50 = 250.00 sales. Deduct the 20.00 refund: provisional net = 230.00.
  5. State the unresolved issue and provisional status. Run the deterministic checks, review each failure, edit and retry.

Deliberately flawed example — synthetic

This example was handcrafted for teaching. It was not produced by, tested against or reviewed as an output from any AI product. It double-counts 0002, silently chooses one 0005 amount, and labels an unresolved summary complete. Even its internally consistent arithmetic cannot make that business scope valid.

Inspect flawed ledger and summary
id,date,type,amount,currency
0001,2026-10-05,sale,120.00,USD
0002,2026-10-06,sale,80.50,USD
0002,2026-10-06,sale,80.50,USD
0003,2026-10-07,refund,-20.00,USD
0004,2026-10-09,sale,49.50,USD
0005,2026-10-10,sale,60.00,USD
week_start,week_end,currency,sale_count,sales_amount,refund_count,refund_amount,net_amount,unresolved_ids,status
2026-10-05,2026-10-11,USD,5,390.50,1,20.00,370.50,0,complete

Your practice workspace

Step 1: read the brief and acceptance criteria above. Step 2: correct the example. Step 3: check, revise and retry. Step 4: download your evidence if the exercise criteria pass.

Drafts, notes and feedback remain in browser memory until reset or navigation. Downloads are explicit JSON files containing your work; keep them as carefully as the source. No spreadsheet formulas are executed. If you later import JSON values into a spreadsheet, formula handling depends on that application. The surrounding site can request assets; extensions and downloaded files are outside this tool’s control.

Self-study answer key and reasoning

Show the accepted ledger and provisional summary

Copying this key is allowed for learning; completion does not establish independent ability. The USD 230.00 figure excludes both unresolved 0005 records. A real handoff would ask the data owner which amount, if either, is valid before issuing a complete weekly total.

id,date,type,amount,currency
0001,2026-10-05,sale,120.00,USD
0002,2026-10-06,sale,80.50,USD
0003,2026-10-07,refund,-20.00,USD
0004,2026-10-09,sale,49.50,USD
week_start,week_end,currency,sale_count,sales_amount,refund_count,refund_amount,net_amount,unresolved_ids,status
2026-10-05,2026-10-11,USD,3,250.00,1,20.00,230.00,1,provisional

Methodology, privacy and limits

The checker reuses AI Work Check to compare your ledger with the explicit accepted reference. It verifies exact strings and unique IDs, then calculates accepted amounts using integer cents. Arithmetic is withheld if ledger checks fail. It cannot resolve 0005, judge your optional note, verify real-world sales or certify professional competence.

The original fixtures are public synthetic data. Your drafts and notes are processed in browser memory, without uploads, model calls or default persistence. Reset clears the working state; downloaded files remain on your device. Site assets may make ordinary requests; browser extensions are outside this tool’s control. No results are published. Use CSV Cleanup & Check to explore parsing and conservative edits, or read Clean a messy CSV with AI for an additional workflow guide.