The instinct is to start with whatever is most annoying. It feels right and it is usually wrong, because annoyance and cost are different measurements and only one of them repays automation.
Frequency beats intensity
The quarterly report that ruins a day is memorable. Four days a year.
Copying details from an enquiry email into a spreadsheet takes ninety seconds and nobody mentions it. Fifteen times a day, that is well over a hundred hours a year, and not one of those hours ever felt like a problem.
Automation repays frequency. The small invisible task is nearly always the better candidate, and it is nearly always the last one anyone suggests.
The four that qualify
Most genuinely good candidates fall into one of these.
Moving information between systems
Someone reads something in one place and types it into another. Enquiry to spreadsheet, order to accounts, form to calendar.
This is the strongest category there is. It is pure mechanical transfer, it is high frequency, and it is where typing errors come from. Automating it removes the time and the mistakes together.
The message that always follows an event
A confirmation when a booking is made. A reminder two days before. A notification when something is ready.
These are usually done inconsistently by people who are busy, which means the value is not only time saved but a customer experience that stops depending on how hectic Tuesday was.
Chasing
Unpaid invoices, unreturned forms, unanswered approvals.
Chasing is unpleasant, so it gets deferred, so it happens late or not at all. This is the category with the most direct financial return, because a reminder sent on day thirty rather than day fifty is money arriving three weeks earlier for no additional effort.
Filing and naming
Saving attachments to the right folder with a consistent name.
Trivial individually and the reason nobody can find anything two years later. The payoff is not the seconds saved, it is that search starts working.
What to leave alone for now
- Anything needing judgement. If the instruction contains “it depends”, it needs a person.
- Anything infrequent. Twice a year is not worth building and maintaining.
- Anything nobody agrees on. Three people doing it three ways is a process problem first.
- Anything expensive to get wrong. Payments leaving the business, contractual commitments. Automate the preparation, keep a person on the decision.
Finding yours
Ask the team a specific question rather than a general one. Not “what should we automate”, which produces the annoying things. Ask: what do you do more than five times a day that a stranger could be taught in two minutes?
That phrasing surfaces the mechanical high-frequency work, which is exactly what qualifies. It also surfaces things nobody has mentioned because they never seemed worth mentioning.
Then count honestly. Times per week, minutes each, multiplied out over a year. That number decides it, not how much anyone dislikes the task. If the total is large, it is also worth asking whether the honest answer is a person rather than an automation.
Start with one
Pick the single best candidate and do only that. Not a programme, not a review of everything.
One working automation teaches you more about what is realistic here than any amount of planning, and it gives the team something to react to. It also keeps the failure small if it turns out the process was less settled than everyone believed, which happens often.
Then build in the check that most first automations skip: how you would know if it stopped. That is a bigger question than it sounds.
The honest summary
Start with the small frequent thing nobody complains about, not the big annoying thing everybody does. Count the real hours before deciding.
If the count comes out low, the honest answer is to leave it alone. Not every irritating task is worth removing.
If you have a few candidates and cannot tell which is worth doing first, describe what happens and how often. Ranking them is usually straightforward once the frequency is written down.
Once you have picked one
A worked example of the first one most businesses build, then the part nobody plans for: what happens when an automation you have come to rely on stops, and how you find out.
Common questions
Which task should we automate first?
The small frequent one nobody complains about, not the big annoying one. Automation repays frequency, so ninety seconds fifteen times a day beats a dreadful task that happens quarterly.
How do we find the good candidates?
Ask the team what they do more than five times a day that a stranger could be taught in two minutes. That phrasing surfaces mechanical high-frequency work, which is exactly what qualifies.
What should we leave alone?
Anything needing judgement, anything infrequent, anything nobody agrees on, and anything expensive to get wrong. For that last one, automate the preparation and keep a person on the decision.