How to find manual work in business without guessing
The short answer: run a five day time log where every person touching admin or customer work writes down every task, the tool they used, and how many minutes it took, logged as it happens rather than recalled at the end of the week. That is the whole method. It works because it replaces memory with a record, and memory is where every attempt to find manual work in a business goes wrong.
Most founders try to shortcut this by asking around. "What eats your day?" gets an honest answer, but it is an answer built from whatever someone did an hour ago or whatever annoyed them most recently, not from an accurate total. A five day log fixes that by making the record automatic instead of recalled. Five working days is enough to catch the Monday report, the Wednesday order backlog, and the Friday status chase without turning into a month-long project nobody finishes.
Build the log before day one
Set up a shared sheet with five columns before the week starts: date, person, task, tool, minutes. Add a sixth for frequency once the week is over, that comes later. Keep it to those columns. A log with more fields turns into a form people dread filling in, and a dreaded form gets abandoned by Wednesday.
Everyone who touches admin or customer-facing work logs, not just the operations team. That includes the account manager copying a signed quote into the CRM, the founder answering "where's my order" from the shared mailbox, and the salesperson rekeying a lead from a form into a spreadsheet. If a shared mailbox is involved, note it directly in the tool column, mailbox triage is one of the most undercounted categories because it looks like communication, not process.
Brief the sheet in under two minutes: one line per task, log it the moment you finish it, don't batch it at the end of the day. Batching is where the five-minute tasks get rounded off and disappear.
Log minutes, not hours
Round to the nearest hour and the sheet lies to you. A task that takes four minutes and happens forty times a week is two hours and forty minutes of monthly cost that never shows up if the smallest unit on the sheet is sixty minutes. Minutes are the only unit small enough to catch it.
The conversion that matters happens after the week, not during it: minutes per instance times frequency per week times roughly 4.3 weeks per month gives monthly hours. A ten-minute task done fifteen times a week is 10.75 hours a month, more than a full working day, and it will not look like much on any single Tuesday. This is the same arithmetic behind our cost calculator, and it is worth running on your own numbers rather than trusting a rule of thumb, since a sales admin's ten minutes and a founder's ten minutes cost the business two different amounts.
Why the interview question gets the wrong answer
"What takes you the longest" surfaces the task someone remembers, and memory is biased toward two things: what happened recently and what irritated them. A frustrating hour spent chasing a signature yesterday will get named before a routine that quietly consumes six hours a month, because the routine stopped registering as work a long time ago.
This is habituation, not dishonesty. The person doing a task the same way for a year no longer experiences it as a task, the same way you stop hearing a fridge hum. A time log has no memory bias, it just totals what happened. Run both and the two lists rarely match, which is exactly why a log beats a conversation when the goal is to find manual work in business rather than confirm what people already suspect.
The manual work nobody logs because it doesn't feel like work
Some tasks are invisible because they are habitual, not because they are hidden. Rekeying a customer's details from an order form into the CRM, answering a status question in Slack that already has an answer sitting in another tab, rebuilding the same revenue summary every Monday from three different tools: none of these feel like a task, they feel like part of the job.
This is also where CRM data goes stale. Nobody logs "I didn't have time to update the deal stage," so the CRM quietly drifts out of date, and nobody notices until a report from it is wrong. Watch for these four during the audit week specifically: manual data transfer between two systems, answering a repeat question that already has a recorded answer, a report rebuilt from scratch on a schedule, and a status update chased down instead of looked up. If your team is asking each other where a job stands, that is exactly the kind of gap covered under job status and single source of truth.
Running the five days
Day one, send a two-line kickoff message: what the sheet is, why minutes not hours, and that logging takes seconds if done in the moment. Link the sheet, don't attach it.
Day three, send one nudge. People forget by Wednesday, not because they don't care, but because the habit of logging hasn't set yet. A short "still logging as you go?" message catches most of the drop-off.
Treat interruptions as loggable events, not noise. A lead reply that pulls someone off a call, an order status question that comes in over email mid-task, a customer chasing a return: these count. If your team is slow to answer new leads, this week will show it directly in the gap between when a lead came in and when someone logged a reply to it, which is the actual measure behind speed to lead, not a guess about it.
By day five, stop. A longer window invites people to start editing their entries for how they think it looks rather than what happened.
Reading the sheet once the week is over
Group every row by task type, not by person, and sum the minutes. Multiply each total by its weekly frequency, then by 4.3, to get a monthly hour figure per task. Sort that list high to low. The top of the list is rarely a surprise once you see it in writing, but it is almost never the task anyone named out loud during the week.
Three patterns are worth flagging specifically. High frequency with low judgment, meaning the same decision gets made the same way every time, is the clearest automation candidate. Data moving from one system to another with no analysis in between, the kind of manual rekeying covered under order and data entry, is the second. A report rebuilt on a fixed schedule from the same sources, the kind covered under weekly reporting, is the third. An ecommerce business will usually find its biggest total in order handling; a services business usually finds it in status chasing or CRM upkeep. Either way, the sheet tells you which one applies to you rather than which one applies in general.
Turn the totals into a decision
Rank the tasks by monthly hours recovered, then multiply each by what an hour of that role actually costs, loaded, not just salary divided by hours. That gives a monthly figure per task, and it is the only number worth acting on from the whole exercise.
Not everything on the list should be fixed. A task that happens twice a month, or one that requires a judgment call specific to that customer, costs less to leave alone than to build around. The honest call there is to skip it, even if the total minutes look tempting on the sheet. Automate the routine, keep the judgment calls with a person.
Once you have a ranked list with real hours and a real cost attached to each line, you have exactly what a Roadmap engagement starts from, which is why we score every idea the same way before pricing a build.
If running this audit turns up more hours than you have time to fix on your own, that ranked list is the starting point for a Revenue Roadmap: two weeks, a fixed price on each item, and the fee credited against whichever one you build first.
Common questions
Do I need special software to run the audit?
No, a shared spreadsheet with five columns is enough; the discipline of logging matters more than the tool.
Should managers log their own time too?
Yes, chasing a status update or approving an order is manual work whether an admin or a manager does it.
What if the week I run it isn't typical?
Pick a normal week; a spike week overstates one task and hides the steady, repeatable ones you actually want to find.
How is this different from just asking the team what to automate?
The log records what people actually do minute by minute, an interview records what they remember or resent, and those two lists rarely match.
Start with the Revenue Roadmap
Two weeks, EUR 1,997, money back if it does not find at least 20 hours a month worth automating. The fee comes off your first build.
