Ask a business owner how much time his team loses to retyping and you get a shrug. "It is not that bad." And it is not, per time. The problem is that it is per time.

At a home furnishings retailer with five branches we started counting. Not estimating, counting.

What we found

Every order was written down by hand almost three times. First on a paper carbon-copy form in the shop. Then as a quote in a spreadsheet on the iPad. And again after the measuring up, plus retyped into the order system. The salesperson emailed the file to the back office, where the customer details were copied over one more time.

The owner estimated fifteen to forty-five minutes per order, depending on the number of lines.

That figure on its own says nothing. What matters is what sits around it: fifteen to thirty-five deliveries a week, five to fifteen measuring-up visits, and in a busy week fifty appointments that all get scheduled by one person.

That is how you get to four hours in a quiet week, twelve in an average one, twenty-six in a busy one. Over a year: 550 to 600 hours.

How to work it out yourself

You need three things, and none of them is in a system.

One. Which action exactly. Not "the back office", but "copying the customer address from the email into the order system". If you cannot write it down in one sentence, you have not pinned it down yet.

Two. How long that action takes. Sit next to someone and time it. Do not ask — people systematically underestimate their own routine work, because they do it on autopilot.

Three. How often. This is where it usually goes wrong. Do not take the average. Take a quiet week, an average one and a busy one, because that spread is exactly where it hurts. In a quiet week retyping does not stand out. In a busy week it is the reason an order sits there.

What the figure is worth

At a cautious internal hourly rate, this company comes out at around €22,000 a year. Of that, an estimated half to two thirds can be automated away.

Note that word: estimated. That is a calculation up front, not a saving measured afterwards. Anyone who papers over that difference is selling hot air, and you only notice when the customer starts counting for himself.

But even as a calculation it does its job. Because the question was never whether fifteen minutes of retyping is bad. The question is whether you want to spend six hundred hours a year on something no customer pays a cent more for.

Nobody feels fifteen minutes. Everybody feels six hundred hours.

And then?

In this case we did not throw a tool at it. First we mapped the whole process, twelve steps from the conversation in the shop to the invoice. Then we drew up 119 requirements, prioritised, eight of them knock-outs. Only then came a recommendation.

What that recommendation was is in another piece — it was not what you expect from an AI agency.