A 20-lawyer commercial litigation firm came to us with a familiar complaint. Their lawyers were billing five to seven hours less per week than they were actually working. Not because the lawyers were lazy or dishonest — because they were exhausted and the timesheet was the last thing they did at the end of a sixteen-hour day, and they could not remember what they had done at 10:42 a.m. five days earlier.
The firm had tried better software. Three different platforms in five years. None of them solved it. The problem, as it turned out, was not the software. It was the timing.
The wrong question
Every timesheet system in existence is built around the same question: at the end of the day, what did you do? It is a terrible question. It depends on memory, and memory is the first thing that fails under sustained cognitive load.
The right question is the inverse. The system already knows what the lawyer did — it is in the email, the document edit history, the calendar, the call logs, the billing portal. The question is not what did you do. The question is: here is what we believe you did. Is this right?
What we built
An agent that runs throughout the day across each lawyer's communication and document activity. It infers time entries continuously: drafted reply to opposing counsel re: Wexler, 0.3 hours; revised motion in limine, Sanchez matter, 1.2 hours; client call with Patel, 0.8 hours. It assigns the matter codes automatically by reading the document and email metadata.
By the end of the day, the lawyer opens a single screen with the day's entries pre-populated. They confirm, edit, or reclassify. The action is reactive, not generative — and reactive work under fatigue is dramatically more accurate than generative work.
The numbers
Within sixty days, the firm's average billable capture rate moved from 71% to 89%. On a base of $4.2M in billings, that is about $850K of previously-lost revenue, recovered. The reclaimed time per lawyer averaged 11 hours per month — a combination of less time spent on the timesheet itself and more time captured because the entries were created in real time.
The annualized impact on the firm's economics was substantially larger than the cost of every system change they had made in the previous five years combined.
What surfaced underneath
The interesting finding was not the recovered hours. It was what the new system revealed about how the lawyers actually spent their time.
Two of the senior partners discovered they were doing roughly six hours a week of administrative work that had never appeared on a timesheet — vendor coordination, calendar negotiation, internal politics — because none of it was billable. Visible inefficiency, once seen, is fixable. The firm restructured an admin role around it within a month.
Another partner discovered she was spending fifteen minutes per day in unbilled context-switching between three matters. Once visible, the firm changed her case load to bunch related work together. Her billing went up, her fatigue went down, and the whole firm learned something about scheduling.
The general principle
Most of what people complain about as their busywork is not actually busywork. It is invisible work — work that happens, has economic consequences, but does not show up on any system. The job of an operating system is not just to do the work. It is to make the invisible visible. Once you can see it, you can decide what to do about it. Until you can, you are guessing.
