The half hour that disappears every day
It is rarely the big projects that eat up time. It is the report that someone assembles from three files every Monday. The folder where invoices are renamed and filed by hand. The list that gets filtered all over again before every stocktake. Taken on its own, none of it looks like much.
Add it up over a year and you quickly reach several working weeks. On top of that comes a risk that is hard to put a figure on: routines like these usually exist in one person’s head and nowhere else. If that person is away, nobody knows exactly what has to happen in which order.
In practice, automation does not mean robots and artificial intelligence. It means a properly built script that carries out the same steps reliably every time. It runs at night, keeps a log of what it did and reports back when something is wrong.
Before, by hand
After, automated
What can typically be automated
What we automate with
For most tasks Python is our first choice, because it reads well and there is a proven library for almost every format. On Windows servers we also work with PowerShell, on Linux with Bash and systemd. For routines that touch several systems we wrap everything in Docker so that it behaves the same way everywhere.
Scripting languages
Data formats
Scheduling
Operation
How we go about it
Watch the routine
We take the time to have the task shown to us as it really runs, not as it is described on paper.
Collect the special cases
What about cancellations, late additions or that one customer number with a different structure. This is exactly where automation otherwise fails.
Build and test the script
First in test mode with real data, without anything being written or sent.
Run both side by side
For a while the manual work and the script run in parallel. That way you can see in black and white that the result is correct.
Go live and monitor
After that the automation runs on its own, with a log and a notification if anything goes wrong.
What you get out of it
Time for more useful work
Your people look after customers and specialist tasks instead of copying and pasting.
Consistent quality
A script handles the five hundredth run just as carefully as the first.
Independent of individuals
The procedure is written down and keeps running when the person in charge is on holiday.
Quick to implement
Many automations are finished within a few days and pay for themselves within a few months.
Easy to trace
Every run is logged. You can check at any time what happened.
Ready to extend
Once the groundwork is in place, further steps can be added with little effort.
Frequently asked questions about automation
A rough calculation helps: if a task costs one hour a week, that adds up to around five working days a year. Many automations take less effort than that. For tasks that only come up twice a year we usually advise against it, unless they are particularly prone to errors.
Scripts are not set in stone. If a file format or a process changes, we adapt it. That is why we document from the start what the script does and what it depends on. An adjustment is then usually a matter of hours, not days.
We always test without write access first and then let the automation run alongside the manual work for a while. Only once the results match do we switch over. We also back up the affected data beforehand.
Not necessarily. Small automations run on an existing server or on a machine that stays switched on anyway. If there is no suitable place for it, we provide a small environment, either on your premises or in our data centre.
No. Automation follows clear rules that you define in advance and that are always applied in the same way. For tasks with fixed procedures that is exactly right. Where rules are not enough, for example because free text has to be interpreted, an AI component can be a sensible addition.
Yes, because that is exactly what we build the monitoring for. You get a message when a run fails, and also when a run did not start at all. Silent failure is the most common flaw in home-grown automations, and we deliberately avoid it.
Which task do you do the same way every week?
Describe one of these routines to us. It is often possible to say in the first conversation whether automating it will pay off.