The automations that pay for themselves first

The test, in one line

How often does it happen, how many minutes does it take, and what does a mistake cost. Multiply the first two, weigh the third, and automate downwards from the top of that list.

This sounds obvious and is almost never what people actually do. The instinct is to automate the task that irritates most, which is usually a quarterly job that takes an afternoon. Twenty times a day at ninety seconds beats four times a year at three hours, by a factor nobody guesses correctly in their head.

The third term matters because some cheap jobs are expensive when wrong. Sending the wrong price to a customer takes ten seconds and costs a week of goodwill. Weight those upwards even when the minutes are small.

The four that almost always pay

The reminder nobody has time to send, the report assembled by hand every Monday, the same information typed into two systems, and the status question customers ask by phone.

These four repeat across trades that otherwise have nothing in common. A dental practice, a garage and a training centre have the same four, wearing different words.

The arithmetic, worked through

Take reminders in a clinic with twelve appointments a day. If two per week do not turn up and half of those would have come with a reminder, that is one recovered slot a week, every week.

Put your own price on a slot and the annual figure appears immediately. Then compare it with what building the reminder costs: it is part of the first system, inside a starter pack of credits, and it does not recur.

Do this calculation before building rather than after. If it does not come out clearly positive on paper, it will not come out positive in practice, and that is a useful thing to learn in ten minutes.

Reminders before dashboards

A reminder changes what happens. A dashboard changes what you know. Both are worth having, and the one that acts pays back first almost every time.

Owners usually ask for the dashboard first, because the dashboard is what they personally lack. It is worth resisting for one round. A screen of numbers you look at on Monday improves decisions slowly; a message that goes out on Thursday evening changes an outcome this week.

Build the acting things first: reminders, alerts when something has been sitting too long, the message that goes out when a status changes. Then build the screen that tells you whether they worked.

What to leave alone at the start

Anything annual, anything mid-redesign, and anything that is one person's craft rather than a process.

• The annual job: automating it costs the same as automating a daily one and returns a fraction
• The process you are currently rethinking: you would be building the version you are about to abandon
• The judgement call: if the right answer varies with things nobody can write down, the software should collect the information and let the person decide
• The thing nobody has agreed on: automating a disagreement makes it permanent and gives it an interface

The honest part: automation exposes bad data

The first week of a reminder system is usually spent discovering how many phone numbers are wrong. The automation did not create that problem, it just made it visible for the first time.

This is genuinely unpleasant and genuinely valuable, in that order. A person sending reminders by hand silently works around a missing number; software either sends nothing or sends it into the void, and then you know.

Plan for it rather than being surprised by it. The first week is not the measurement week. Fix what the first week shows, and start counting from the second.

How to tell whether it worked

Pick one number before you build, write it down, and look at it a month later. No-shows per week, hours to produce the report, days from job finished to invoice sent.

One number, chosen in advance, is worth more than any amount of dashboard afterwards. The failure mode without it is that everybody agrees the system is helpful and nobody can say whether it is worth its cost.

Bulgaria is a useful place to be deliberate about this. Eurostat has AI use among Bulgarian companies at 8.6% and basic digital intensity among small firms at 38.3% against an EU average of 71.4%, which means most of your competitors are still doing these four jobs by hand. That is an advantage while it lasts, and it lasts longer if you can prove which parts of it are working.

QUESTIONS