Cost cutting rarely fails because the numbers were wrong. It fails because the wrong things get cut: the tool that saved an hour a day goes, and the meeting that costs four people an hour a week stays.
Start from where the time goes, not from the invoice.
Find the work that repeats
Spend one week writing down every task you do more than once. Not a system, just a note at the end of each day.
By Friday the list sorts itself into three piles:
- 01Work only you can do, because it needs judgement or a relationship.
- 02Work anyone could do, because the steps never change.
- 03Work nobody should do, because it exists to feed a report nobody reads.
The second pile is where technology pays. The third pile is where the real money is, and the answer there is to stop, not to automate.
The two-and-five rule
If a task happens more than twice a week and takes more than five minutes, it belongs in a workflow rather than on a list. That is roughly forty minutes a month for one task. Ten of those is a working week a year.
Automate the boring layer first
The instinct is to automate the interesting work. The interesting work is exactly the work that needs a human, so it fights back.
Go after the plumbing instead:
- —Moving information from one place to another: a form to a spreadsheet, a payment to an invoice, a signup to a welcome email.
- —Chasing: reminders, follow-ups, anything that starts with "did you get a chance to".
- —Formatting: turning notes into a summary, a call into a list of actions, a table into a chart.
None of it is impressive. All of it is silent and daily, which is why it costs so much.
A process you have not written down cannot be automated, only guessed at. Write the steps first, in plain sentences, then find the tool.
Cut the attention tax
Meetings, status updates and switching between tasks are the expensive part of a small team, and none of it shows on a bill.
Replace the status meeting with a written update
One short document per week, written once, read whenever. It costs the writer ten minutes and gives everyone else their hour back.
Give every piece of work one owner
Two owners means two people waiting. One owner and a named deadline removes most of the coordination that meetings exist to fix.
Batch the interruptions
Answer messages in two or three blocks a day rather than continuously. Deep work does not survive a notification every four minutes, and no tool fixes that for you.
Buy leverage, not headcount
A subscription that removes an hour of daily work is cheap at almost any realistic price. A hire is a year of onboarding, management and salary. Reach for the subscription until the work genuinely needs a person.
Two tests before you pay for anything:
- —Name the task it removes. If you cannot name one, the tool is a hobby.
- —Decide how you will cancel it. Something that cannot be removed in an afternoon is not a tool, it is a dependency.
The tools directory is organised by the job you are trying to do, which is the only useful way to compare them.
Review the stack every quarter
Software spend grows quietly. Once a quarter, list every subscription next to the task it removes. Anything with no task next to it goes.
Do the same with the automations. A workflow built for a process that no longer exists is worse than nothing, because it keeps producing output that people still trust.
Start here this week
Pick the single task you complain about most and write down its steps. That is usually the one that pays first, and the writing down is most of the work.
If you want the practical version, the lessons cover the same ground with specific setups rather than principles.
