A repetitive task is any step your team does the same way most weeks, with the same inputs, and a result you could write down before they start. Filing a signed letter. Chasing an overdue invoice. Copying a new client's details from a form into the practice system. The work is not hard. It is just the same.
Most of these are not worth a project. A few of them are eating the week. The difference is not how boring they feel. It is whether they pass a short test you can run this afternoon.
What it actually costs
People talk about this as inefficiency. That word is too clean. The bill is messier.
The first cost is context switching. Someone is in a matter, a set of accounts, or a campaign. Then they stop to move a file, update a status, or send the same reminder they sent last Tuesday. Getting back to the real work takes longer than the step itself.
The second is errors that show up downstream. A name copied wrong. A date left on last month's template. An invoice that went out without the latest hours. The person who did the step may never see the damage. The client does, or the partner does, two weeks later.
The third is the quiet fact that the person who knows the steps is the only backup. If they are on leave, or they leave, the work pauses until someone reconstructs it from memory and a messy inbox. You do not have a process. You have a person. In a firm this size that person is often the founder.
The kill-list test
Walk your own week. For each candidate, ask five questions. Yes to all five and it is worth killing, or at least writing down. A no on any of them and you leave it alone for now.
1. Same trigger every time. An email arrives. An invoice is overdue. A form is submitted. You can point at the event that starts the work. If the start is "someone remembers," it is not a candidate yet.
2. Same steps, in the same order, most of the time. If people argue about the order, or skip half of it depending on mood, you do not have a settled process. You have a habit that has not settled.
3. The exception rate is low enough that a person can still catch the weird ones. Some files will be odd. That is fine if a human can sit at the end and catch them. If half the cases are special, you are looking at casework with a similar-looking start.
4. Someone can name the done state in one sentence. "The signed PDF is in the matter folder and the status is filed." If you cannot say what done looks like, do not kill the task yet.
5. It happens often enough that the time adds up. Weekly is enough. Once a quarter, with extra decisions each time, is usually not. Count the last four weeks.
Run this on the work that already annoys you: the same status update, the same chase, the same copy-paste between two tools that do not talk. Filing signed letters usually passes. A partner reading every cover email usually does not. Retyping invoices from one system into another usually passes. The write-off conversation does not.
What looks repetitive and is not
Pricing a matter is not this kind of work. Hiring is not. A difficult client call is not. These can have a checklist around them. The core is judgement, and judgement changes with the facts in front of you.
Agencies see this in scoping. Law firms see it in advice. Accountants see it in anything that starts "it depends." The surrounding admin might pass the five questions. The professional call will not. Keep the two apart on the list.
If two experienced people would reasonably disagree on the next step, it is not a candidate. Disagreement is the job.
People problems in a process costume
Sometimes a step exists because nobody trusts the last step. A partner re-reads every letter. A bookkeeper re-keys figures that already live in the system. A founder asks for a screenshot because the status field has been wrong before.
That is not a process gap. It is a trust gap. Connecting the tools will not fix the reason the check appeared. You will just move the mistrust one screen over.
If the step is there because someone got burned, deal with the burn. Write down who is accountable. Make the first step reliable. Then look again. The extra check may go, or it may stay on purpose as a human catch. Either is better than dressing a people problem up as a workflow.
What to do this afternoon
You do not need a project. You need a list.
Take one hour. Name the tasks that already run on rails. Put each through the five questions. Keep the ones that pass. For each keeper, write the steps down, including the trigger and the done state.
Then stop. The next decision is not which tool. It is which of three answers fits.
A rule. If this, then that, with no one in the middle.
A connection between tools, so the data moves without a person copying it.
A person staying on it, because the volume is low, the exceptions are ugly, or you are not ready to trust a rule yet. Writing the steps down still helps the next time someone is away.
Most firms we sit with in Cape Town do not have an IT department. The founder is the backup. The kill list is how you stop that being the whole week.
Pick one repetitive task that passed. Write the steps. Decide which of the three answers it gets. That is the whole job for now.
