Short answer

Four kinds of administrative work almost always pay off: first drafts of recurring texts, summaries from data you already have, sorting of what comes in, and making internal knowledge searchable.

Three almost never do: what happens rarely, where every case is unique, and what requires a judgement someone has to stand behind.

And the sum that decides which is which takes five minutes — see the next section.

Do the sum first

Before anything else: find out whether what you have in mind is worth the money. It takes five minutes and can sometimes save fifty thousand kronor.

Minutes per time × times per year ÷ 60 × hourly cost.

An example that is more common than you might think. Someone spends twenty minutes a week putting together a report. It feels like a lot, because it disrupts every Friday. But that comes to seventeen hours a year, and seventeen hours will not pay for a build costing forty thousand kronor — it takes years to earn back.

Another: four people each spend an hour a day writing quotes and customer replies from scratch. That is over two thousand hours a year. There, the build does not need to be especially good to pay off — it only needs to remove a third.

The difference between the two examples is not the technology. It is entirely volume, and that is the only variable that decides it. That is why I ask “how often?” before I ask “what?”.

Four that pay off

1. First drafts of recurring texts. Quotes, standard replies, contract drafts, adverts, minutes. The heavy part is rarely the writing — it is getting started. A draft that is seventy per cent right removes that threshold altogether, and the material is already in what you have written before. Quality control is also simple: someone reads it before it goes out.

2. Summaries from data you already have. The monthly report, the weekly follow-up, the material for the board. If the figures exist but someone spends half a day putting them together, that is half a day that can be won back every month. The condition is that the data is in reasonable shape — see the next section.

3. Sorting what comes in. Emails, forms, fault reports, cases. Getting the right thing to the right person with a summary at the top is often the biggest time saving in the whole business, particularly where several people read the same inbox and everyone starts from scratch. More on that part here.

4. Knowledge that sits in the heads of two people. The things “everyone knows” but nobody has written down: why something is done in a certain way, what you do when the customer asks that particular question, which supplier is used for what. Making it searchable is seldom anyone's favourite job, and it is one of the few things where the technology makes an impossible task manageable rather than just faster. It is also what protects you when someone leaves.

Three that don't

What happens rarely. The annual report, that application, the procurement every third year. Even if each occasion takes three days, the total is small, and the groundwork of describing how it should work costs more than it saves.

Where every case is unique. If the exceptions are the rule, the description of the exceptions becomes longer than the work you wanted to be rid of. A good test: can you describe how the task is done on one A4 page without writing “it depends” more than twice?

What requires a judgement someone has to stand behind. Assessments of staff, prices in negotiation, decisions with legal consequences. The technology can prepare the groundwork — which is often very useful — but the decision should have a person behind it, both because it turns out better and because someone has to be able to answer for it afterwards.

And a fourth, which is not really administration at all: when the data does not exist. The most common reason a project costs more than planned is not the technology but that the material sits in twelve spreadsheets with different naming standards. Then tidying it is the whole job, and that should be stated in the quote.

The most common mistake

Starting with what irritates most rather than what costs most time.

They almost never coincide. What irritates is often something that happens once a month, takes half an hour and feels pointless. What costs is something done every day, by several people, that nobody reflects on any more because it has always been that way.

That is why, when I take on an assignment, I don't ask “what would you like to automate?”. That question gives the irritation list. Instead I ask: what did you do the day before yesterday that you have done fifty times before? That question gives the time list, and the two lists rarely look alike.

Where you start on Monday

Without buying anything: ask three people to keep a simple log for one week. What they did, roughly how long it took, and whether they have done it before. No tools — a sheet of paper is enough.

After a week you have something almost no company has: an actual picture of where the time goes, rather than a feeling. Then do the sum with the formula at the top for the three largest items.

Often that is enough to decide whether you need help at all. Sometimes it shows the answer is a half-day training course and no build at all — and that is a perfectly honest outcome, which I often recommend.

The most important thing

If you only take one thing with you: frequency beats complexity.

A simple task done five hundred times a year is almost always worth more attention than a complicated one done five times — even if the complicated one is more fun to solve and the one you first thought of.

How a bounded build works, what it usually costs and how it is handed over is on the page about AI automation.

Best wishes, Mikael GlännströmWellbeing coach, author and father of two