All free tips and guides

Step-by-step guide · Recurring work

Match two lists—and keep the exceptions

A small matching task is a good place to test an AI workflow. The goal is to distinguish clear matches from the items a person still needs to look at.

Steps

1. Make a small practice copy of the two lists. Remove private information. Keep a stable reference ID for each row.
2. Define what counts as a match: for example, location, task type, and due month. Tell the AI which fields must agree.
3. Require three outcomes: one clear match, several possible matches, or no match. Ask it to explain the evidence.
4. Check the rows yourself and reconcile the total count. This is a review table; it should not edit the original schedule or records.

Prompt to copy

Replace the text in brackets with your own details, then paste it into the AI tool you use.

Help me compare two lists from my own work and find matches, gaps, and exceptions. Start by asking which two lists I want to compare and what a match should mean. Invite me to attach the files or share a small sample, and verify that you can read the relevant rows and columns.

Propose a matching rule from the actual fields, then ask me to confirm it. Keep each original row ID; if there are no stable IDs, label the rows clearly in the review output. Do not edit the source files.

For every row in the first list, return its ID, status (clear match, ambiguous, or missing), candidate ID(s) from the second list, the fields supporting the result, and anything I need to check. Don’t choose between equally plausible matches or invent missing dates or values.

Give the total count for each status and reconcile their sum with the first list. Flag records used more than once and unmatched records in the second list. Show a few matches with their source rows so we can check the rule before expanding it.

What to check

• Did every schedule row appear exactly once?
• Do the status counts add up to the input row count?
• Are ambiguous and missing cases still visible?
• Do clear matches actually meet every part of your rule?
• Is any record matched to more than one row, and is that expected?

Then try this

Try a second small sample with a known answer. If the matching rule is stable, a spreadsheet lookup or conventional automation may be the more dependable long-term solution.

Updated September 2026. Examples are fictional. This is a working method, not a guarantee of what any tool will produce.