What do you do every week that a machine could do?
I automate the routine work that eats people's time: the sorting, the copying, the same report every week. It runs quietly in the background, and when something looks odd, a person gets the question.
- Typical
- Documents, integrations, reports, imports
- Control
- Unclear cases go to a person
- Traceability
- Everything that happens is written down
- Systems
- Built on what you already use
01 · Three steps
Start with the task someone quietly hates.
- Step 1
A review of the routines
We walk through a normal week and find the tasks that repeat.
- A list of what is worth automating, and what is not
- All you need to do is tell me how the job is done today
When you know the time disappears, but not where.
- Step 2
One flow
One task is automated from start to finish, on the systems you already use.
- Unclear cases land in a pile for a person
- Everything that happens can be looked up afterwards
When one task comes back every day or week.
- Step 3
More flows, kept running
Several tasks are connected, and I keep them going.
- You are told when something stops
- The flows keep up when the systems around them change
When the first flow is running and you want more.
02 · Fig. 1 · Before and now
A registry in, an app's worth of stories out.
Kulturjakten pulls heritage sites from the national registry. The import reads each one, writes the entry in the app's voice, checks facts and images and queues everything. An editor approves before anything is published.
03 · Where it usually starts
Four tasks that keep coming up.
Nearly everything I automate begins with a task someone does over and over.
Documents read and sorted
Invoices, forms, contracts and PDFs come in, and someone opens each one to see what it is. I build the step that reads them, pulls out what matters and files them where they belong. If a document is unclear, it lands in a small pile for a person.
Systems that do not talk to each other
The web shop has the order, accounting needs it, and someone types it across. I connect them, so the data moves itself and both sides agree.
Reports that put themselves together
The numbers come from the same places every week, and someone spends the morning copying them into the same template. I fetch them, build the report the way you already like it, and have it ready when the week starts.
Imports checked against the source
A spreadsheet with thousands of rows is going into a system, and nobody has time to check it all by hand. The import compares every row with the source, fixes only what is safe to fix, and asks a person about the rest.
I have automated a large part of my own work the same way, so I know where it pays off and where a person should stay involved.
04 · In depth · What separates a flow from a script
Built for the day something goes wrong.
Getting an automation to work once is easy. The job is making it work every time, also when the data is odd and the system on the other side is down.
Safe to run again
If the flow stops halfway, it can be started again without duplicate orders, duplicate emails or duplicate rows.
Checked against the source
Every row is compared with where it came from before it is saved. Only what is safe to fix is fixed automatically.
An error queue instead of guessing
What the flow does not understand is set aside with the reason, so a person can decide.
Retries when others are down
If the system on the other side does not answer, the flow waits and tries again instead of losing something.
Logs and alerts
Everything that happens is written down, so it is easy to see what was done and why. You are told when something stops.
Rules before AI
What can be decided by a fixed rule is decided by a fixed rule. AI is used where something has to be read and understood.
Common questions about automation.
What can be automated?
Most things that are done the same way every time: sorting email and documents, moving data between systems, reports, reminders and importing large files. A good sign is someone saying "I do this every week".
What is the difference between automation and an AI agent?
Automation does the same thing every time by fixed rules. An AI agent can read and assess, for example understand what an email is about. Often both are used: the automation moves the data, the agent reads what has to be understood.
What happens when something goes wrong?
It stops and asks a person instead of guessing. Everything that happens is written down, so it is easy to see what was done and why.
Do we have to change systems?
No. I build on what you already use, and connect the systems you have.
Tell me about the task.
Where the data comes from, where it should go, and who notices when something is wrong. I reply within a day, and together we find out whether it is worth automating.