Why a feeling isn't enough
When you ask around how much time a certain task eats up, the answer that comes back is almost always an adjective, a lot, quite a bit, an enormous amount, and with an adjective you can't work anything out
The trouble is that time spent on repetitive work is hard to judge from the inside, whoever has done it for years has it so well in hand that it feels quick to them, whoever looks at it from outside sees it as slow because they only notice the moments when someone is stuck
To decide whether to automate something you need a number, even a rough one, as long as it comes from a method you can repeat in exactly the same way six months later and compare
First step, count the operations per month
The most solid measure to start from isn't time, it's the operations, that is how many times a month the thing happens, how many invoices come in, how many documents someone goes and downloads from the portal
The advantage is that you can count the operations per month without looking anyone in the eye, the number is already somewhere, in a folder, in a mailbox, in a list inside the business software you use, you just filter one month and read the total at the bottom
And it's a stable number, while the minutes per case change with the day, with the heat, with the colleague who drops by to ask something, the monthly volume stays where it is and holds up as a comparison even a long time later
Second step, watch ten cases in a row
Now you need the time for a single operation, and the way to get it is to watch ten cases one after the other, from opening the document to the moment the job is closed and the person moves on to the next one
Ten is too few for statistics, but it's enough to form an honest idea, because ten cases in a row will also include the one that goes wrong, the one that gets stuck, the one that forces someone to make a phone call, and those are exactly the ones that vanish from a rough guess
Take the time of the whole route, not of the single piece, if a case calls for opening two programs and an email, the time is what passes from the first click to the last, including the pauses where the person waits for a screen to load
Then don't just take the average, look at the longest case and the shortest one too, the distance between the two tells you how regular that work is, and so how realistic it is to automate it
Third step, ask the person who does it
The person who opens that screen every morning knows things no measurement will give you, they know which supplier sends the badly made PDFs, they know which day of the week the avalanche arrives, and above all they know the step nobody ever wrote down, the one you do because otherwise an error turns up a long time later
Ask openly, how much time does it take from you on a normal day, and how much on a bad one, you'll get two numbers instead of one, and the bad day is the one that counts most, because that's where the backlogs are made
Standing over someone with a stopwatch doesn't work
The temptation is to stand behind a person with your phone in your hand and time them, and it's the best way to get two things, a false number and an angry office
False because nobody works normally while being measured, someone who feels watched speeds up, skips the checks they usually do, and hands you back a time that exists on no other day of the year
And angry because a measurement with no explanation gets read for what it looks like, a check on the people, which is the worst way to open a conversation about automation
So say straight away what you're measuring and why, it's the process under examination, not the person, what you're looking for is the mechanical step to take out of the way, not who takes thirty seconds longer
The formula, and an example with made-up numbers
The sum is one multiplication, operations per month times minutes per operation, and you have the minutes per month, which divided by sixty gives you the hours
Say four hundred documents a month and six minutes a document
Four hundred times six makes two thousand four hundred minutes, that is forty hours a month, which in this example is a full week of one person, and in front of a number like that you stop asking yourself whether to look into it, you ask where to start
From there, if you multiply those hours by your company's hourly cost, you have the figure, and you can get it without a calculator on the work out the savings page
What to do with the number
The number you end up with isn't a promise of savings, it's the size of the problem, and it's good for two things, deciding whether that job deserves attention before ten others, and having a starting point to compare against later
Not all of those hours come off, inside there is the mechanical part and the part that takes a head, and only the first one gets automated
And if the number that comes out is a small one, you've gained something anyway, you've stopped thinking about that job and you can look elsewhere