What is worth automating
A good candidate meets three conditions at once: it repeats, it has clear rules, and a person does it today.
- Issuing invoices from orders — including sending them and chasing due dates
- Moving data between the e-shop, warehouse and accounting
- Payment reminders on a fixed schedule
- Reports someone assembles by hand on the first Monday of the month
- Creating records — a new customer in one place means a record everywhere it belongs
- Checks — an alert when something crosses a limit or a value is missing
What is not worth automating
- Activities that happen a few times a year. Setting it up costs more than doing it by hand.
- Decisions that need judgement. An automation can prepare the material; leave the decision to a person.
- Processes that are still changing. Settle them first, then automate.
- Anything nobody would check afterwards. A silent error is worse than noisy manual work.
Working out whether it pays back
You do not need a spreadsheet for this. Three numbers will do.
| What to work out | How | Example |
|---|---|---|
| Time saved | Hours per week × 50 weeks | 3 h weekly = 150 h a year |
| Value of that time | × hourly cost of the person | 150 h × 400 CZK = 60,000 CZK |
| Cost of mistakes | Annual cost of fixes and complaints | e.g. 20,000 CZK |
When the total beats the price of the solution, it pays back within a year. For most small automations, which start at 15,000 CZK, it lands considerably sooner.
Do not leave out what is hard to measure: that nobody waits for the result, that a stand-in can handle it, and that three different versions of one spreadsheet stop circulating.
Off-the-shelf tool or custom build?
| Off-the-shelf tool | Custom | |
|---|---|---|
| When | Standard procedure, mainstream systems | Your own rules, older systems |
| Cost | Monthly fee, grows with volume | One-off build plus running it |
| Speed | Days | Days to weeks |
| Limits | What the tool cannot do, you cannot do | Limited by budget, not by the tool |
For a straightforward chain between mainstream systems, an off-the-shelf tool is usually the sensible answer. Custom work starts paying off where the rules are your own, where an older system has no API, or where the monthly fee grows faster than the volume it handles.