Four Kids, Four Supply Lists, One Spreadsheet
Four school supply lists, four authors, four conventions. Consolidating them is a normalization problem, not a copy-paste problem. What made an AI handle it well: ask the clarifying question first, surface the ambiguity, and write down the assumptions.
Every August our school posts the supply lists as four separate PDFs. That’s not a complaint about the school. It’s just how it works when you have a kid in Pre-K4, one in 2nd, one in 4th, and one in 6th. Four grades, four documents, four different people who wrote them, four different ideas about what a supply list should look like.
The lists themselves are short. The work is in the reconciliation. Somebody has to sit down and figure out that four separate documents asking for glue sticks add up to twenty-four glue sticks, in three different sizes, one of which has to be purple.
This year I handed the whole thing to Claude. Here’s what actually happened, because the interesting part wasn’t “AI made a list.” It was watching where the judgment calls landed.
The setup
I pasted four URLs and said: consolidate these.
Claude pulled all four PDFs and asked me two questions before doing anything. The second one was about output format. The first one was the one that mattered:
Which grade is your middle-schooler in? The 6th to 8th grade list has grade-specific add-ons.
That’s the right question, and it’s the one I would have gotten wrong at Target. The 6th to 8th grade list is a shared document with conditional branches buried in it: Algebra 1 students ONLY need a TI-84 Plus CE. 7th grade ELA students ONLY: 1 spiral notebook, 1 plastic pocket folder. My kid is in 6th. That’s a $120 graphing calculator I don’t have to buy, sitting in a line item that reads like a requirement until you notice the qualifier.
I’ve spent enough time around requirements documents to recognize the failure mode. A shared spec with per-role exceptions is exactly where people skim, and skimming is how you end up with the wrong thing in the cart.
Where the real work was
The naive version of this task is concatenation: glue four lists together, sort alphabetically, done. That gets you a longer list, not a better one. The actual work is normalization, and it’s genuinely fiddly:
Same item, different specs. Glue sticks appear on all four lists. The 6th-grade list says “Glue Sticks, 2.” The 4th says “Glue Stick (large), 4.” The 2nd says “Glue Sticks (small), 6.” The Pre-K list says “Purple Elmers Glue Sticks Pack of 12.” These belong on one row for shopping purposes and four separate specs for buying purposes. Claude collapsed the row and kept every spec in a notes column, which is the correct call. You want one line on the list and the details when you’re standing in the aisle.
Same word, different units. The 4th-grade list says “Red pens - pack 2.” The 2nd-grade list says “2 Red Pens.” One of those is two packages and one is two pens. Claude summed them to 7 and then flagged the mismatch explicitly rather than pretending the total was clean. I’d rather have an honest number with a caveat than a tidy number that’s wrong.
Items that don’t consolidate at all. Crayons look like one item across three lists and are actually three products: Crayola 24-count, Crayola regular 16-count with “no Twistables” called out, and Crayola large 8-count with “(not jumbo)” in the original text. Merging those would produce a list that’s shorter and useless. They stayed as three rows.
Conditional logic based on data I didn’t supply. The Pre-K list assigns Ziploc bag sizes by last name: A to J snack bags, K to Q gallon, R to Z two-gallon. Claude resolved it for our household and wrote the reasoning into the notes column rather than just asserting the answer. That’s the behavior I want. Show the inference so I can check it.
What came back
A workbook: a master tab with 48 normalized line items across six categories, a quantity column per child, a formula-driven total, and a “Got it?” column for checking things off. Then a tab per kid for the ones that need to be packed and labeled separately. Then a sources-and-assumptions tab listing all four original PDFs and every judgment call made along the way.
That last tab is the part I’d push anyone to insist on. Six assumptions, written out, including “the 6th-grade list requests 1 to 3 boxes of tissues; 1 box is counted in the totals” and “the 2nd-grade list does not mention a book bag, so none is counted.” Those are the seams. If something’s wrong, it’s wrong there, and I can check six sentences in thirty seconds instead of re-reading four PDFs.
The transferable part
Strip out the crayons and this is a shape I recognize from work: several source documents, overlapping but not identical, written by different authors with different conventions, that need to become one actionable artifact without losing the exceptions.
Three things made it work, and none of them are specific to school supplies.
The clarifying question came before the work, not after. One question about which grade my middle-schooler was in changed the output materially. A tool that guesses and moves on produces something that looks finished and isn’t.
The ambiguity got surfaced instead of smoothed. Every place the sources disagreed, whether units, sizes, or brands, ended up visible in the output rather than silently resolved. The temptation with a consolidation task is to make it look clean. Clean is the wrong goal. Checkable is the goal.
And the assumptions got written down. Not in the chat, where they scroll away, but in the artifact, where they’ll still be there in three weeks when I’m wondering why I only bought one box of tissues.
The total, for the record: 48 line items, 24 glue sticks, 48 pencils, 10 rolls of paper towels, and eight tennis balls that have to be cut open and jammed onto chair legs. There are apparently YouTube videos for that part. Claude didn’t offer to help with it and I didn’t ask.
Have a project in mind?
Let's talk through the path from idea to production.