“Put these tasks in date order” leaves two decisions unanswered: what happens when dates match, and where do undated tasks go? A neatly formatted AI answer can hide both guesses. Agree on the rules before handing over the rows.
Try a small list with a deliberate tie
Use these fictional records. Dates follow year-month-day, and each task ID is unique:
- T02 — 2026-10-12 — Nora
- T01 — 2026-10-12 — Luis
- T03 — no date — Mei
- T04 — 2026-10-14 — Omar
Specify: earliest date first, missing dates last, then task ID in ascending order when dates match. Keep every field exactly as supplied. The expected order is T01, T02, T04, T03. Nora and Luis still belong to their original tasks; only the positions of the complete records change.
Without the ID rule, either order of the two October 12 tasks could satisfy “date order”. That is an underspecified request, not proof of an error. Do not let the model invent a deadline for Mei to make the list look complete.
Check the records as well as the order
Keep an untouched copy with an original row number. After sorting, compare the number of rows and the occurrences of every ID. Then check that each ID still has the same date and person. A correct sequence of dates does not reveal a name attached to the wrong task.
For repeatable work, Excel can sort on several columns using additional sort levels. Select the whole dataset, not just the date column, and confirm which row contains headers. AI can help describe the intended rule; the small example above gives you something concrete to verify before applying it to a larger list.