Tool · Where do you start!?I checked the maths behind this on · 26 Aug 2026
Not which tool you buy, but which work you put into it. Describe a task, give six answers, and it lands on a map next to the rest of your week.
too long? here it is in six lines
At nearly every client I see it happen in the same order. A subscription gets signed, a training session follows, and then everybody starts looking for work to put into it. Six months later usage has collapsed and somebody concludes that it does not deliver much after all.
It goes wrong because the first question gets skipped. Not which tool you buy, but which work you actually want to put into it. Because that is not all of your work. There is work you could hand over tomorrow, there is work where you have to keep taking the last step yourself, and there is work you should stay away from because you cannot check whether it is right.
This card helps you tell those apart. You describe a task, give six answers, and it lands on a map. Do that for five or six tasks and you get something more useful than a single recommendation. You see your own work divided up, and usually you also see that your attention is in the wrong place.
All six questions eventually come down to two things.
Is it worth it. How much time does this task cost you per month, all told. Not per go, but per month, because that is the number that matters. A ten minute job you do every day is worth more than a half day job you do once a quarter.
Can it go. That is what the other five questions are about. Can you tell whether the result is any good. What does it cost when it goes wrong. Does someone expect this to come from you personally. Is the knowledge you need written down anywhere. And how sensitive is the content.
The second group decides whether something may go, the first whether it pays. You need both. Work that could easily go but barely costs you any time is a nice experiment, not a project.
Of those five, one towers over the rest: can you tell whether the result is any good.
That sounds like a detail but it is the heart of the matter. If you can see at a glance that a piece of work is right, then things are allowed to go wrong now and then. You pull it out and you move on. If you cannot see it, you are not handing over work, you are taking a gamble.
And then the combination I would really watch out for. You cannot check whether it is right, and a mistake costs money or reputation. That is the most dangerous position there is, because it goes well for a while and then once it does not. You find out when it is already out the door. In nearly every AI project I have watched fail, this exact combination was hiding somewhere.
There is one more question in here that you will not find in any other comparison: is the knowledge you need for this written down anywhere.
Ask that about a task you have been doing for ten years. In a great many cases the answer is no. You know which client you address how, which exception you make for which product, and when you deviate from the standard. That is written down nowhere. That sits in your head.
As long as that is true, you cannot hand that task over. Not to a machine, and not to a new colleague either. That is not an AI problem, that is a documentation problem that happens to be coming to light now.
The good part is that the fix has value regardless. If you write down how you do something, you are no longer the only one who can do it. That is a gain before anything has been switched on at all. Which is why in that case the card does not say it cannot be done, it says something has to be cleared up first.
The map has four boxes. To the right means more time saved, upwards means more suitable to hand over. Every task is a numbered dot.
Top right is where you start. Plenty of time and nothing in the way.
Bottom right is worth it but something is blocking it. Often something you can fix, like writing down knowledge or agreeing where the data is allowed to go.
Top left can go but returns little. Do it if it takes no work, but do not build anything for it.
Bottom left you leave alone.
Below the map you get, per task, what the outcome is and what the first step would be. There are four outcomes: hand it over, do it together, repair it first, or keep doing it yourself. That last one is a real answer and not a failure.
At the top sits the number I consider the most important one, and it is a range. The bottom end is the work that can go entirely. The top end adds the work where you keep the last step yourself, because there you get the first draft back, and that is the bulk of the job.
Why not a single number: how much of that second group you actually get back depends on how long you take over that last bit. That differs per task and per person, and I do not know it. An invented factor of a half or two thirds would sit on your screen just as firmly as the hours you filled in yourself, and that is exactly the kind of precision I want no part of. Two bounds that both come out of your own answers are less precise and more true. It is also the number you tell your manager or your client: at least this, at most that.
Take five or six tasks. Not the most interesting ones, but the tasks your time actually goes to. If someone asks you where your week goes, those are the ones.
Then do it again with your team, and have them fill it in separately first. The tasks you disagree about are exactly the conversations you want to be having. Usually it turns out that what one person can check in two minutes costs someone else an afternoon. That difference says more about how you work than the score does.
And start with one task from the top right box. Not with three. You get more done with one job that runs smoothly than with four you half finish.
The map only calculates with what you fill in yourself and nothing goes to a server. Nothing is saved either, so when you close the page your list is gone. If you want to keep it, use the button to copy the overview.
And treat the outcome as something to talk about, not as advice. Whether something may actually be handed over depends in the end on the agreements that apply where you work. Not on a score on a website.
Everything happens in your browser. Your tasks go nowhere and nothing is saved. Close the page and your list is gone.
What there is to gain
8 to 16 hours per month
The bottom end is work that can go entirely. The top end adds the work where you keep the last step yourself: you get the first draft back, not the approval. How many of those hours you actually get back depends on how long you take over that last bit, and I do not know that. Hence a range and not a single number. Start with the first box. Not with the most interesting problem.
Your work on one map
little time savedmore time saved →
Task by task
No blockers worth the name. You can see whether it is good, a mistake is manageable and the knowledge is written down.
This is where you start. Set this up first, measure for a month what it returns, and use those numbers to get the next task approved.
8 hours per month · suitability 92 out of 100 · example
A mistake is expensive here, but you can check for it.
Put a fixed check in its place. Not reading it over and assuming it is fine, but one thing you look at specifically before it goes out.
8 hours per month · suitability 31 out of 100 · example
You cannot check whether it is right and a mistake is expensive. That is the most dangerous combination there is.
It goes well for a while and then once it does not, and you find out when it is already out the door. Find a way to check the result first. If you cannot, this work belongs with a person.
2.5 hours per month · suitability 0 out of 100 · example
The outcome is something to talk about, not advice. Whether something may actually be handed over depends in the end on the agreements that apply where you work, not on a score on a website.