You know your retention rate. You could recite your loss ratio and your book growth without opening a file, and you would have a fair guess at what your top three carriers did last quarter.
Now tell me how many hours your agency spent last week re-typing information that already existed somewhere else.
That number does not exist. It is almost certainly bigger than several of the ones you just recited, and it is the only one in the building nobody has ever written down.
Every business has a version of this. The work that quietly takes the most hours is the work nobody thought to measure, because it never shows up as a project with a deadline. It shows up fifteen minutes at a time, all day, and then the day is gone.
Everyone can name it. Almost nobody can rank it.
Walk through your office and ask what wastes the most time. You will get an answer in about four seconds, and it will be a good answer. Ask that same person to rank their top three by hours and watch the answer soften.
That gap is the problem worth solving. What people report is the loudest annoyance, and loud is not the same as expensive. The carrier portal that logs you out mid-quote is memorable because it is infuriating. The eleven minutes spent finding last year's renewal summary is forgettable because it happens quietly, four times a day, every day, and never once feels like a crisis.
This is how Admin Drag survives in a well-run agency. It is not hiding. It is just never counted.
What a time audit is, and what it costs you
A time audit is a one-week record of where your working hours actually went, sorted into a small set of categories you fix before you start, so that at the end of the week you can rank them instead of guessing.
That is the whole method. It is not a timesheet, nobody is billing anything, and precision to the minute would defeat the purpose. The goal for the audit week is an entirely ordinary week, described honestly. Working harder while you measure only guarantees you measure the wrong week.
The categories have to be set before you begin or the exercise collapses. Left to invent labels as they go, most people end up with thirty categories by Thursday, each with a single entry, and no signal at the bottom of the page. A small fixed set is what makes the totals mean something.
The cost is ninety seconds a night and ten minutes on Sunday.
The three categories
Three categories cover the overwhelming majority of the drag in an agency, and they are the ones the audit measures.
Re-keying. Typing information that already exists somewhere else. The quote request into the portal. The portal into the agency management system. The management system into the spreadsheet the carrier wants in their own format. None of this creates information. It moves it, by hand, one keystroke at a time.
Chasing. Work that only exists because someone else has not done their part yet. The unsigned application. The underwriter who has not come back. The client who has been meaning to send the loss runs since the second week of the month. Chasing is the follow-up, the reminder, and the overhead of keeping a list of everything you are waiting on in your head.
Hunting. Looking for something that already exists and belongs to you. The file, the email thread, the current version, the number you had last Tuesday. Hunting almost never appears in anyone's estimate of their own week, because each instance is small enough to feel like nothing.
Then one more line, and it matters more than it looks:
Three lines a night, plus one for anything that fits none of them.
The fourth line is Other, and you should be generous with it. If something in your week genuinely does not belong in the first three, it goes there, and you write down what it was. That line is the honest catch, and it is often where the surprise lives.
How to run it
- Before you log anything, write down your guess. Rank the three categories in the order you believe they will finish, fold the paper, and put it in a drawer. This takes thirty seconds and it is the most important step in the exercise. You will want it on Sunday.
- Make one page. Four lines: Re-keying, Chasing, Hunting, Other. Paper is fine. A note on your phone is fine.
- Ninety seconds at the end of each day. Put a number of minutes next to each line and round to the nearest fifteen.
- Run it for a full week, including the bad day. The day everything goes sideways is the day the log earns its keep.
- Sunday, give it ten minutes. Total each column, rank them, and then open the drawer.
No app, no tool, no subscription. If this needs a purchase order it will not survive Wednesday. The printable version, with the categories already on it, is Appendix A of the AEP prompt book, and the digital edition is free.
Have your producers and your service staff run the same week you do. Four logs from four seats will tell you something one log cannot.
Where this goes wrong
Four failure modes, in the order I see them.
Logging during the day. Stopping to record an interruption is another interruption, and by Wednesday it feels like surveillance. The log happens once, at the end, from memory. Memory at nine o'clock is accurate enough for a ranking, which is all you need.
Chasing precision. Nobody is auditing your audit. Fifteen-minute increments and an honest shrug will produce the same ranking that stopwatch discipline would, and one of those two approaches survives to Friday.
Skipping the bad day. The instinct is to restart the week when Tuesday falls apart. Tuesday falling apart is the data. Agencies do not get a clean week, and a log that only contains clean days describes an agency that does not exist.
Exempting yourself. If you hand this to the team and do not run it, you will read four logs, recognize none of your own week in them, and quietly conclude that your time is the well-managed part. Run it. Principals are usually the worst offenders on Hunting, and nobody is in a position to tell them.
Reading the totals
The ranking matters more than the raw numbers, and the comparison against your folded guess matters most of all. If the order surprised you, that surprise is the finding, and it is worth more than the hours.
When Re-keying finishes first, your systems are not talking to each other. This is structural, not a discipline problem, and no amount of asking people to be more efficient will move it. The work exists because the same information has to live in four places and only a human is connecting them. It is the most mechanical of the three columns, which also makes it the most straightforward to hand off.
When Chasing finishes first, your bottleneck is mostly outside your walls. Carriers, underwriters, clients, and the mail. The instinct is to follow up more, and that instinct is wrong, because more follow-up is more of the exact work you just measured. A Chasing-heavy week points at a system that tracks what you are waiting on and surfaces it at the right moment, so the follow-up stops depending on somebody remembering.
When Hunting finishes first, your information exists and nobody can find it. This is the column people most often misread. The usual conclusion is that the staff needs training on the system, so the agency buys training, and the number does not move. Training was never going to move it. Your people knew exactly what they were looking for. They could not find it, and the reason sits in how the agency files things, not in what anyone knows.
When Other finishes first, stop and read what you wrote in it. Your categories are not wrong. Your week simply contains something significant that nobody in the agency has ever named out loud, and now you have seven days of evidence describing it. That is the most valuable page in the log.
What to do with what you find
You pick the column that finished first, and you fix one thing inside it.
Mike Welker manages benefits at The Insurance Center in Utah, and his first column was never in doubt. Group benefits gets none of the automatic carrier downloads that property and casualty enjoys, so every carrier's production and cancellation reports had to be collected by hand, analyzed, keyed into Applied Epic, and then keyed again into Agency Zoom. He had been trying to standardize those carrier spreadsheets for something like a year.
At a summit in May, he built his first AI assistant. Sixty days later, after four or five rounds of getting it wrong and adjusting, he had a working pipeline: mask the protected health information, collect every carrier report, standardize them, restore the PHI locally, and distill the whole thing into a dashboard his team uses all day.
Claude has created an output that gives me more than what I thought I could ever do on my own.
Mike Welker, Benefits Manager, The Insurance Center (Utah)
The part worth borrowing is not the pipeline. It is that he aimed at one column and stayed on it through five attempts. He was not experimenting with AI. He was fixing a specific, measured, named piece of his week.
A prompt is a fine place to start, and the time it gives back is real. If Re-keying finished first for you, a well-built prompt will reformat a census in seconds and hand you most of an afternoon.
A prompt is also rung one of a ladder, and the rung above it is a Digital Crew.
A Digital Crew is a governed team of AI agents with named roles, defined jobs, a human manager, and decision rights that stay with people. The crew takes the busywork. Your best people get back to the work only they can do.
The difference is who does the work. A prompt helps your producer get through the census faster, and she is still the one doing it. A crew has a seat whose job is the census, a person who owns that seat, and a decision it never makes on its own. Designing that seat is a different piece of work than measuring your week, and it is the work I do with agencies directly.
Measuring is yours, and it starts tonight. Get the AEP prompt book, print the Drag Log out of Appendix A, and write down your guess before you log a single minute. Sunday you will know something about your agency that you did not know on Monday.
Related reading: What Admin Drag looks like in a Medicare agency walks through the same week from the producer's side. Our survey of insurance agents on AI adoption covers what agencies are doing about it. When your team is ready to build the thing that takes the biggest column, Claude training for teams is where that happens.
