We delivered an AI college funding analyst for a Mount Pleasant father holding six award letters in six different formats — one that separates loans from gifts and renewable gifts from one-year gifts, projects all four years using each school's own tuition history, models the years both daughters are enrolled at once, and drafts the special-circumstances appeal to each school's stated process.

He has worked at the Wando Welch Terminal for twenty-two years, and his daughters are eighteen months apart, which means that for three years he will be paying two tuitions at the same time. The older one is a senior at a high school in Mount Pleasant. Nine applications, six acceptances, and by the first week of April, six financial aid award letters sitting on the counter. He had assumed the hard part was the applications.

No two letters were built the same way. One listed a fifty-five-hundred-dollar unsubsidized federal loan as a line item under the heading "Your Financial Aid Award," which is true in a technical sense and misleading in every other. One quoted a cost of attendance that left out travel, which for a school eleven hours away is not a rounding error. One offered a merit scholarship that renewed only at a 3.

5 grade point average, a condition that appeared in a footnote on the second page. One called its largest line a grant, and it was a grant, for the first year, after which it dropped by four thousand dollars. Three of the six did not state a four-year number anywhere. He built a spreadsheet.

He is a careful man and he got it wrong twice, once by about eight thousand dollars, because the thing being compared was not the same thing from letter to letter. There was also a problem underneath all of it: the aid formula looks at income from two years back, and two years back he had worked an unusual amount of overtime during a stretch when the terminal was short-handed. On paper he had a very good year. In the year his daughter would actually be enrolled, he did not.

What we built takes the letters however they arrive — PDF, portal screenshot, a photograph of a page — and normalizes all six into one comparable figure: total out of pocket across four years. Loans are pulled out and shown as debt with their interest rate and repayment, not counted as aid. Gifts are split into renewable and first-year-only. Every renewal condition is read out of wherever it is buried and stated plainly, and the model shows what each school costs if that condition is not met, because a 3.

5 is a real risk and not a footnote. Years two through four are projected using each school's own published tuition increases over the past decade rather than assuming the freshman number holds. And it models the overlap years, when the younger daughter enrolls and the calculation changes for both of them. Then it writes the appeal.

Every school has a process for reconsidering an award when circumstances have changed, and every school calls it something different and routes it somewhere different. It identifies the correct process and contact for each, drafts a letter that states the overtime anomaly in the school's own vocabulary, lists exactly which documents to attach, and tracks the window, which at two of the six was under three weeks from the letter's date. Normalized, the six offers did not rank the way they looked. The school that appeared cheapest on the page came out third over four years, and the spread between how it looked and what it was came to a little over nineteen thousand dollars.

Two appeals were filed. One produced an additional sixty-two hundred dollars a year for four years; the other produced two thousand dollars, once, which he took. She is going to the school that was second on the letter and first on the arithmetic. The younger one applies in the fall, and the file is already open.