Work Mentor / Field notes

How to check an AI-generated weekly sales report

A practical five-point checklist: paid orders, refunds, channel revenue, weekly growth and unsupported causal claims. Includes a free fictional-data practice.

An AI can write a fluent paragraph from inconsistent numbers. A useful review checks the data and the claim separately. Start with this small, repeatable rubric.

1. Fix the reporting boundary

Choose the current and previous week before adding numbers. Count only paid orders inside the current period. A cancelled order can still carry an amount in a spreadsheet; that does not make it revenue.

2. Reconcile gross sales and refunds

For this exercise, net revenue equals the sum of paid gross amounts minus the refunds on those orders. Keep cents until the end. Real businesses may use different accounting policies; document those before applying a template.

3. Rank channels using the same measure

Group current-week net revenue by channel. Do not compare one channel’s order count with another channel’s revenue. If channels tie for the top, either tied channel is accepted in this exercise.

4. Use the previous period as the denominator

Weekly growth = (current net revenue − previous net revenue) ÷ previous net revenue × 100%. A negative result is a decline. If the previous value is zero, percentage growth is undefined; our practice datasets keep it above zero.

5. Separate observation from explanation

“Revenue rose” is an observation. “Our ads caused the increase” is a causal claim. Two weekly totals alone cannot establish the cause. Ask for campaign exposure, timing, alternative explanations and a suitable comparison before making that claim.

Use AI without hiding the evidence

Give an assistant the fictional CSV and ask it to show which rows it included, its net-revenue formula and the evidence for each claim. Verify the numbers yourself. Then attempt a new case with the same tool choice and without the mentor’s hints. A correct answer on a copied example does not establish transfer.

What the free practice includes

Fictional order datasets, five deterministic checks, targeted hints, limited AI explanations, new cases, local history and backup export. Optional AGI cloud membership costs 9 USDT for 30 days plus the matching decimal and network fees shown at checkout; it adds cloud storage, not extra AI quota.

Method updated 1 October 2026. Fictional practice, not a credential or a promise of job performance.