Most of what is written about AI is strategy. This is seven pieces of work. They repeat every week in most offices. For each one: what the machine does and what the person still does.
Documents come first because they are the clearest example. But the same method covers the rest of what repeats: the calls you do not get to, the inbox, quotes, typing between systems, follow-up, invoicing and reminders, reports.
1. The documents
The same fields out of the same document type. Lease, deed, voucher, case file. The machine reads the document and pulls the fields out. Every line carries page and paragraph, so you check instead of hunt. If a field is not there, it says so. It does not guess. A person approves. Always.
2. The calls you do not get to
The machine answers, handles the fixed questions and books into the calendar. Afterwards a summary is waiting for you. Complaints and negotiation stay with a person. That is where tone decides the outcome.
3. The inbox
Mail is sorted by case and urgency. What is urgent sits at the top. Standard replies are drafted and waiting. Anything that needs a position stays with you. The machine has read the content. It has no opinion on it.
4. Quotes
The draft is built from your own prices and your earlier quotes for similar work. You correct the figures and send. The pricing itself stays with you. The gain is that the quote goes out the same day.
5. Typing between systems
The same values, typed from one system into another. Customer number, amount, date, case number. The machine reads them in one place and writes them in the other. A person checks that they match. It is the dullest item on the list. It is also the one that is measured least often.
6. Follow-up, invoicing and reminders
The invoice is built from the work that was done. The reminder is prepared on time on your template. The follow-up on a quote is ready the day it is due. You approve before anything goes out. When to chase a client stays with you. That is a relationship, not a rule.
7. Reports and status
The numbers are fetched from the systems they already sit in. They are assembled and waiting at the hour you usually sit with them. The machine fetches and assembles. The person reads and draws the conclusion.
Where to start
Not with all seven. That is exactly what fails. In 2025 MIT reviewed more than 300 AI initiatives in companies. 95% had no measurable effect on the bottom line. Most had started too much at once.
Pick the task with the highest number when you multiply times a week by minutes. Not the most annoying one. The most repeated one. Build it to the end. Measure it for a month. Then take the next.
How much of a task the machine can carry depends on the task. In our calculator we count on 45, 60 or 75% of the hours, and we show the cautious figure first.
One workflow in production is worth more than five that are close.
That ranking is what we do with you in the free audit. Tell us what repeats. Then we find the one that saves the most.
The next step
You tell us what repeats. We map it for free and write it down within a week: hours, kroner, where to start, and one fixed price — or an honest no. It takes at most 3 hours of your time.
Tell us what repeats.
We map the work and put hours, kroner and one place to start in writing. Free, because nobody can price an automation before seeing the work.
