Extract & Verify
Extraction is fast but not trustworthy until you check it. Here you'll pull fields from a sample record block into a table, then verify each against the source — and catch the value the AI got wrong.
Work through it
The source records
BIO-101 / Sec 02 / Dr. Reyes / Enrolled: 23 / Online HIST-210 / Sec 01 / Dr. Okafor / Enrolled: 41 / In-person MATH-150 / Sec 04 / Dr. Lindqvist / Enrolled: 18 / Hybrid
Write your extraction instruction
Before you look at the AI’s output, write the instruction you’d give it to pull Course, Section, Instructor, Enrolled, and Modality. Saved on this device.
The AI-extracted table
The AI returned:
| Course | Section | Instructor | Enrolled | Modality |
|---|---|---|---|---|
| BIO-101 | 02 | Dr. Reyes | 32 | Online |
| HIST-210 | 01 | Dr. Okafor | 41 | In-person |
| MATH-150 | 04 | Dr. Lindqvist | 18 | Remote |
Your verification
For each row, note what you checked and whether it holds. Saved on this device.
| Row | What you checked | Holds? (yes/no) |
|---|---|---|
| BIO-101 | ||
| HIST-210 | ||
| MATH-150 |
Saved automatically on this device.
Reveal model answer
A strong extraction instruction (copy and adapt):
Extract these fields into a table, one row per section: Course, Section, Instructor, Enrolled, Modality. Keep values exactly as written — do not reformat or infer. If a field isn’t present, write "not found". Add a column noting the source record each row came from.
Then verify the output against the source:
| Row | Finding |
|---|---|
| BIO-101 | WRONG — Enrolled should be 23, not 32 (digits transposed). |
| HIST-210 | OK — all fields match. |
| MATH-150 | WRONG — Modality should be "Hybrid," not "Remote" (changed term). |
Takeaway: two of three rows had errors that look perfectly clean in the table. Only the check against the source caught them — extraction is a draft until verified.