Ask an owner why a job takes eleven days and you'll get an answer about capacity. Not enough people, too much work, everyone's flat out. It's a reasonable theory. It's also, in our experience, almost never what the numbers say.
The first instrument we run in an audit is a trace. We pick one real, completed piece of work — one matter, one client onboarding, one month-end close — and we follow it from the first inbound touch to the moment cash lands. Not a description of how it should go. One that actually happened, with a name and a date attached to it.
At every step we record six things: who touched it, what system they were in, what they did, what they had to go and look up somewhere else, how long the step took, and — this is the one that matters — how long it then sat before the next step.
Total handling time: four hours. Elapsed time: eleven days.
That shape shows up again and again. The work is fast. Then it waits. And an owner who has never seen those two numbers written on the same page has never really seen their own business.
Why the waiting is invisible
Handling time is visible because someone is doing something. It appears in timesheets, in project tools, in the felt experience of a busy afternoon. Waiting appears nowhere. Nobody logs the four days a file sat in a shared inbox because the person who could approve it was on site. Nobody bills the afternoon a proposal waited on one number from accounts.
So when a firm feels slow, everyone reaches for the explanation they can see. We need another person. We need a better CRM. We need to automate intake. Each of those is a proposed solution wearing the costume of a problem — and if you accept the framing, you build the thing that was asked for, it doesn't fix what was actually wrong, and everyone concludes the technology didn't work.
What actually causes the wait
In firms of ten to fifty people, the waits cluster in a small number of places:
- A single approver. Usually the owner, usually on something that stopped needing their judgment about two years ago.
- A handoff with no owner on the far side. Work is sent rather than assigned, so it belongs to nobody until someone notices it.
- A lookup. Someone needs one fact that lives in a different system, or in a colleague's head, and the job stops until they get it.
- A re-entry. The same information typed a second and third time, each transcription an opportunity to stall and a place for two versions to diverge.
None of those are technology problems. Three of the four are process problems, and the fourth is a data problem. Which is precisely why we put AI at the end of the sequence rather than the beginning — pointing a model at a workflow with a single-approver bottleneck gives you a faster path to the same queue.
The finding that usually wins
We score every finding the same way: hours reclaimed per month, multiplied by our confidence in that number, divided by the effort and the risk. Rank descending, apply the sequence rule, and the top two or three become the build.
The item that comes out on top is rarely the impressive one. In a worked example we ran recently, it was an invoice approval threshold — raise the number below which a partner doesn't need to sign, and roughly nine hours a month comes back. It takes a day. It breaks nothing. It's a policy change rather than a technology change, and it hands an owner back their own evenings.
That's the sort of thing a firm doesn't hire a consultant to discover, and is genuinely startled to receive. We lead the readout with it, every time — and we say plainly that it costs nothing and doesn't need us. A client who acts on a recommendation in week one and watches it work has already decided about the rest.
What to do with this before you call anyone
You can run a rough version of the trace yourself this week, and you don't need us to do it. Pick one job that closed last month. Open the email thread, the project record, the invoice. Write down the date of every step. Then put two numbers at the bottom of the page: how many hours of work it took, and how many days it took.
If those two numbers are close together, your constraint really is capacity and you should be hiring. If they're an order of magnitude apart — and they usually are — then the thing slowing your firm down isn't how hard anyone is working.
Caprenna is an operations engineering firm working with 10–50 person professional services firms. If you'd like the trace run properly, start a conversation.