An AI chat assistant can turn terse daily notes into a readable weekly report, but totals such as hours and completed-task counts should be checked separately. This guide uses a small synthetic work log, an AI drafting prompt, and a Python reconciliation script.
Who this is forFor people who keep short daily work logs and want AI to draft weekly reports without losing control of hours, counts, and factual details.
AI is useful for reorganizing terse work notes into readable prose, but it should not be treated as the source of truth for arithmetic. A weekly report can sound coherent while quietly changing a total, counting one activity twice, or presenting a planned task as completed. A safer workflow separates narrative drafting from numerical reconciliation.
The following work log is synthetic. Each row records one activity, the hours spent, and whether the activity was completed during the week. Keeping these fields structured makes later reconciliation much easier.
| date | task | hours | status |
|---|---|---|---|
| 2026-09-14 | Clean customer CSV | 2.0 | completed |
| 2026-09-14 | Draft import script | 1.5 | completed |
| 2026-09-15 | Test import script | 2.5 | completed |
| 2026-09-16 | Investigate duplicate IDs | 1.0 | completed |
| 2026-09-17 | Update validation rules | 2.0 | completed |
| 2026-09-18 | Prepare user notes | 1.5 | in progress |
By hand, the hours are 2.0 + 1.5 + 2.5 + 1.0 + 2.0 + 1.5 = 10.5 hours. Five of the six activities have status completed, so the completed-task count is 5. The final activity is still in progress and must not be counted as completed.
Paste the log into an AI chat assistant and explicitly limit the task to rewriting and organizing the supplied data. Also ask it to state its totals so they can be checked.
Prompt to the AI: Using only the work log below, write a concise weekly report with three parts: work completed, work still in progress, and weekly totals. Do not invent outcomes or tasks. State the total logged hours and the number of completed activities. Keep the original task meanings.
A typical AI answer might say: This week, customer data was cleaned, an import script was drafted and tested, duplicate IDs were investigated, and validation rules were updated. User notes remain in progress. Total logged time was 10.5 hours, with 5 completed activities.
Save the same synthetic example as work_log.csv with the columns date, task, hours, and status. The reconciliation script should read this source file directly rather than trying to recover numbers from the AI-written prose.
| Metric | Expected value from the log |
|---|---|
| Rows | 6 |
| Total hours | 10.5 |
| Completed activities | 5 |
| In-progress activities | 1 |
This separation is important: the CSV is the evidence, while the weekly report is a presentation of that evidence.
The following Python 3.12 script reads work_log.csv, validates the required fields, totals the hours, counts statuses, and writes a reconciliation report to outputs/weekly_reconciliation.txt. It does not overwrite an existing outputs directory.
from pathlib import Path
import csv
input_path = Path("work_log.csv")
output_dir = Path("outputs")
if output_dir.exists():
raise SystemExit("outputs already exists; remove or rename it before running again")
if not input_path.exists():
raise SystemExit("work_log.csv was not found")
required_fields = {"date", "task", "hours", "status"}
rows = []
with input_path.open("r", encoding="utf-8", newline="") as f:
reader = csv.DictReader(f)
if reader.fieldnames is None:
raise SystemExit("CSV has no header")
missing = required_fields - set(reader.fieldnames)
if missing:
raise SystemExit(f"Missing columns: {sorted(missing)}")
for line_number, row in enumerate(reader, start=2):
try:
hours = float(row["hours"])
except ValueError:
raise SystemExit(f"Invalid hours value on line {line_number}")
if hours < 0:
raise SystemExit(f"Negative hours on line {line_number}")
status = row["status"].strip().lower()
if status not in {"completed", "in progress"}:
raise SystemExit(f"Unexpected status on line {line_number}: {status}")
rows.append({
"date": row["date"].strip(),
"task": row["task"].strip(),
"hours": hours,
"status": status,
})
total_hours = sum(row["hours"] for row in rows)
completed_count = sum(row["status"] == "completed" for row in rows)
in_progress_count = sum(row["status"] == "in progress" for row in rows)
output_dir.mkdir()
report_path = output_dir / "weekly_reconciliation.txt"
report = (
f"Rows: {len(rows)}\n"
f"Total hours: {total_hours:.1f}\n"
f"Completed activities: {completed_count}\n"
f"In-progress activities: {in_progress_count}\n"
)
report_path.write_text(report, encoding="utf-8")
print(report, end="")
print(f"Saved: {report_path}")For this example, the independently expected values are 10.5 total hours and 5 completed activities. Compare those values with every numerical statement in the AI draft.
| Check | Log result | AI draft | Decision |
|---|---|---|---|
| Total hours | 10.5 | 10.5 | Matches |
| Completed activities | 5 | 5 | Matches |
| In-progress activities | 1 | User notes remain in progress | Consistent |
This example counts completed activities by row. A real organization may instead count tickets, deliverables, projects, or milestones, so the counting rule must be defined before reconciliation. The script also checks arithmetic and simple statuses only; it cannot determine whether a work-log entry itself is accurate or whether the reported work quality was sufficient.
2026-09-21 · hand-checked example · Python 3.12
The explanations and examples were written for this site. See the official sources below for the related behavior and concepts.