Write a weekly report with numbers you can check

A report can use real numbers and still draw the wrong conclusion. This practice separates a workload balance from a completion percentage. You can check the arithmetic with a calculator before considering any AI draft.

Practice goal: Write a short report that reconciles the workload, uses the correct denominator and separates counts from unsupported explanations.

Use only the invented information below. Do not paste real emails, personnel records, customer details or confidential office documents. Changing a name does not establish permission to upload a real record.

• [R1] Fictional Elm Office weekly request log: eight requests were open at the start of the week. Twelve new requests arrived during the week.
• [R2] Nine requests were completed during the week: three from the opening group of eight and six from the twelve new requests.
• [R3] Each request is counted once in its group. There were no reopened, cancelled, transferred or duplicate requests. These are complete counts for this one reporting week.
• [R4] In this exercise, ending open work equals opening open work plus new requests minus completed requests. Percent change in open work uses the opening count as the denominator. The share of new requests completed uses only completed new requests over all new requests.
• [R5] The log does not supply time spent, due dates, customer satisfaction, last week’s completion rate or whether AI was used. Do not infer these from the counts.
• [R6] The audience is the office manager. Produce a draft with three headings: Workload, Completion, Questions. Do not upload a real log or send a report.

You can complete this on paper or in a local document. No external account, upload or purchase is needed.

Draft your answer before reading the correction. Optional: use only this fictional prompt in a tool your organization permits. Do not connect accounts, upload other files or send messages for this task.

Using only R1–R6 below, write a report with Workload, Completion and Questions headings. Show ending open work and its percentage change; show the share of new requests completed. Explain why nine divided by twelve is not that new-request completion measure. Include source IDs and show arithmetic so another person can check it. List missing evidence rather than claiming timeliness, satisfaction or productivity gains.

Complete fictional facts:
- [R1] Fictional Elm Office weekly request log: eight requests were open at the start of the week. Twelve new requests arrived during the week.
- [R2] Nine requests were completed during the week: three from the opening group of eight and six from the twelve new requests.
- [R3] Each request is counted once in its group. There were no reopened, cancelled, transferred or duplicate requests. These are complete counts for this one reporting week.
- [R4] In this exercise, ending open work equals opening open work plus new requests minus completed requests. Percent change in open work uses the opening count as the denominator. The share of new requests completed uses only completed new requests over all new requests.
- [R5] The log does not supply time spent, due dates, customer satisfaction, last week’s completion rate or whether AI was used. Do not infer these from the counts.
- [R6] The audience is the office manager. Produce a draft with three headings: Workload, Completion, Questions. Do not upload a real log or send a report.

Authored flawed sample, not an actual provider response:

Workload: the backlog fell to three. Completion: nine of twelve new requests were completed, a 75% success rate. AI made the team faster, and all work was on time.

1. R1–R4 give 8 + 12 − 9 = 11 requests open at the end, not three. Subtracting the wrong groups can hide unfinished work.

2. Open work rose by three from a starting eight: (11 − 8) / 8 × 100 = 37.5%. It did not fall. This describes counts, not the team’s effort or quality.

3. Only six of the twelve new requests were completed, so their completion share is 6 / 12 × 100 = 50%. Nine includes three opening requests and cannot be substituted as completed new work.

4. The 75% calculation, 9 / 12, is total completions divided by arrivals. It is a different ratio; it does not show which arriving requests were completed.

5. R5 provides no deadlines, time data, satisfaction evidence or AI-use record. Neither on-time delivery nor an AI-caused improvement follows. Ask for the relevant evidence if the manager needs those conclusions.

One supported correction:

Workload

Eight requests were open initially and twelve arrived. Nine were completed, leaving 8 + 12 − 9 = 11 open. Open work increased by three, or (11 − 8) / 8 × 100 = 37.5% (R1–R4).

Completion

Six of twelve new requests were completed: 6 / 12 × 100 = 50%. Five opening requests and six new requests remain open. Total completions divided by arrivals is 9 / 12 = 75%, but that is not the completion share of the new-request group (R1–R4).

Questions

Time spent, due dates, satisfaction, a comparison completion rate and whether AI was used are not supplied. What additional evidence does the manager need? These counts alone do not establish timeliness, productivity or an AI effect (R5–R6).

Check your answer:

• The balance is 8 + 12 − 9 = 11, also five remaining opening requests plus six remaining new requests.
• Open work increased by three and 37.5%, using eight as the base.
• The new-request completion share is 50%; nine includes opening requests and is not the correct numerator.
• Workload, Completion and Questions are separate, with arithmetic and source IDs visible.
• The report does not assert deadlines were met, satisfaction improved, time was saved or AI caused a change.

Different wording can satisfy these checks. Recheck the facts after any rewrite.

Keep definitions next to reported measures, including which requests and dates belong in each numerator and denominator. In a real authorized log, reconcile duplicates, reopenings, cancellations and transfers before using this simple balance. A clean layout or calculated percentage does not establish that the underlying data is complete. Record review effort if you later compare workflows; do not assume a time saving.

Primary guidance checked on 2026-10-12. The references support source checking, privacy boundaries and clear presentation; the office facts and sample answers are invented. UK government guidance is cited for its scoped practice advice, not as a rule applying to every workplace.

• Generative Artificial Intelligence Profile
• Artificial Intelligence Playbook for the UK Government
• Use Clear and Understandable Content

Keep real work within your organization’s approved data and review process. These exercises do not establish tested accessibility, time savings, job competence or a professional qualification. No provider trial or human professional review is claimed.

Start free Module 1: account sign-in is required; no card is needed. Or read general AI learning guidance. Keep your practice answer in your own local document if you want to return to it; this page does not save it to your account.