Do I need artificial intelligence to automate the office

Almost never, and whoever puts it everywhere is probably selling you the word of the moment

An old office computer switched on, showing a text screen

The short answer is no, almost never

To automate the repetitive work of an office, artificial intelligence is almost always unnecessary, because most of that work is made of rules somebody in the company can already explain out loud, and a rule you can say out loud can be written down

Take the supplier who sends the invoice every month on the same layout, the number is always in the same place, so is the taxable amount, so is the total, and the person who posts it isn't interpreting anything, they're repeating steps they learned years ago and by now do without thinking

When the steps are these, the work isn't asking anybody to guess, it's asking to be written down once and then run always the same way, and a program has been doing that for thirty years with no need for any new word

Where artificial intelligence really is needed

Artificial intelligence is needed where the data arrives dirty and has to be interpreted, meaning where there is no hard rule that covers the cases, and this happens more often than it looks

One example is the skewed scan of a handwritten document, with a stamp over the numbers, where looking in the right place isn't enough because there is no right place, or the email dashed off by a customer who orders the goods in words, with no codes, no table, and a different turn of phrase every time

In situations like that the part that interprets makes sense, and Muffin Suite uses it when the case calls for it, but keeps it under control, meaning what it comes up with goes through the checks like any other data, the totals have to add up, the VAT number has to exist, the dates have to make sense, and if something doesn't add up the system stops instead of writing something at random into your business software

On accounting data you want a rule, because with a rule you know in advance what it will do, while with an interpretation you find out afterwards

Why they offer it to you anyway, and how you decide

If somebody is putting artificial intelligence into a job that consists of copying three fields from a tidy file into a screen, that isn't a technical choice, it's the word of the moment that sells today, and you're paying for it twice, once in the monthly fee and once in unpredictability

The way to notice is simple, ask what happens when the document arrives different from usual, if the answer is that the system works it out on its own, push, because the good answer is that the system stops and tells you, and on this you can also read the guide on the checks and stops in an automation

The right question isn't whether to use artificial intelligence, it's whether the data coming in is tidy or not, if it's tidy you write the rule and sleep soundly, if it's dirty you interpret it and put more checks around it

Almost every process in a small office sits in the first group, and to tell which group yours is in you only have to look at where the documents come from, and how much they resemble each other, on the underlying distinction there is more in the guide on automation and artificial intelligence

What people usually ask us

No, almost never, because most repetitive office work is made of rules somebody in the company can already explain out loud, and a rule you can say out loud gets written into the system and run always the same way, with no need for anything that interprets
It is needed where the data arrives dirty and has to be interpreted, for instance a skewed scan with a stamp over the numbers, or an email where the customer orders in words with no codes, there a hard rule does not cover the cases and interpreting is the only way
Yes, where the case calls for it, and always under control, meaning what the artificial intelligence reads goes through the checks like any other data, the totals have to add up and the dates have to make sense, and if something does not add up the system stops
Because with a rule you know in advance what it will do, while with an interpretation you find out afterwards, and on an accounting entry being predictable is worth more than being clever, so the part that interprets at most proposes a reading, it does not have the last word
Ask what happens when the document arrives different from usual, if they tell you the system works it out on its own, push, because the good answer is that the system stops and tells you, instead of guessing and writing something into your business software
Look at where the documents come from and how much they resemble each other, if they are tidy and always the same then writing the rule is enough, if they arrive photographed, handwritten or different every time then interpreting is needed, with more checks around it

Show us what you still do by hand

Describe the process or send a short screen recording and we'll tell you what can be automated

WA