Institutional knowledge doesn't leave with a reviewer. It was never captured.

A reviewer at a mid-size firm gets a text from a client at nine at night, second week of March: a Roth conversion from eighteen months ago, and does the basis carry the way the client remembers. The reviewer works it out, texts back an answer, moves on to the next return in the queue. Nothing about that exchange goes anywhere but the reviewer's phone. Eight months later, during extension season, a different reviewer opens the same client's file and has to work the same question out from scratch — not because the first answer was wrong, but because the file has no record that the question was ever asked.

The note dies with the file

A review comment left in a workpaper, or a diagnostic cleared and signed off in the prep software, exists to get this year's return out the door. Nothing about how either one is built treats it as an asset the firm still owns once the return is filed. The engagement closes, the file archives, and next season's version of the same client starts from the numbers that carried forward — not the reasoning that produced them. The knowledge wasn't deleted. It was never given anywhere permanent to be.

Turnover eased. The leak didn't, because it was never the emergency

Turnover across public accounting firms fell to 11.8% in 2025 — the lowest figure Inside Public Accounting's Trends data has recorded in more than two decades of tracking it, down from a 15.9% peak during 2022's "Great Resignation." That reads like the crisis is over. It isn't the crisis that was ever doing the damage: over the same two decades, the rate has never dropped below 11.8%, which means roughly one reviewer in nine has left every single year, good years included, for as long as the data goes back. A rate that low and that steady isn't a spike a firm can wait out. It's the firm's ordinary operating condition, and it has been quietly emptying out case-specific judgment at that pace the entire time.

Two people work the return. Only one record survives either of them touching it

Tax prep software is built around a handoff: a preparer enters the return, a reviewer checks it, clears the diagnostics, and signs off. What persists afterward is a record of that handoff happening — a timestamp, a cleared flag, an approved entry — not a record of the judgment behind it. The system can tell you a position was reviewed. It can't tell you why the reviewer accepted this year's treatment, or what made them override the diagnostic that flagged it in the first place. The tools were built to move a return through a pipeline, and they do that well. Preserving the reasoning that moved with it was never the job they were designed for.

Busy season doesn't leave a slot for writing it down

Seventy-eight percent of tax and audit professionals surveyed for the 2025/26 busy season worked more than fifty hours a week, and forty-three percent logged fifty-one to sixty. Under a schedule like that, documentation isn't skipped because reviewers don't value it. It's the first task cut, because a fifty-five-to-seventy-hour week has no built-in slot for writing up a call that already got made out loud, on the phone or in the margin of a PDF, when the alternative is starting the next return in the stack ten minutes later.

Where Pernee fits

None of these four causes is a staffing problem waiting on a hiring market to fix it. They're all the same structural gap in different clothes: a judgment gets made, and nothing in the systems around it is built to keep it. Pernee attaches a reviewer's reasoning to the return line it explains, at the moment the call is made, so it survives the file closing, the reviewer leaving, and the next season's version of the same client — without asking anyone to find extra time during the season to write it down twice.