Software takes over the work you repeat every week.
Capturing documents, retyping data, building the same report every Monday: those are hours nobody notices, because they are spread thinly across the month. We work out what a routine really costs first, then automate the one that pays back twice: once in the effort you stop spending, and once in the volume the same team can take on afterwards.
- The numbers come first
- The old way keeps running
- Fixed price after the consultation
Four signs that somebody in your business is doing a machine’s job.
Someone retypes data from a PDF
A delivery note or a form arrives as a PDF, and then the same thing is typed into the system again, field by field. The mistake does not happen the first time round, it happens on the twentieth one in the afternoon.
Supplier invoices entered one by one
The same pile every month: open it, type in the amount and the account, file it away. Every invoice looks different, and the handling is identical every time.
The Monday morning report
Figures copied out of three systems into a spreadsheet, formulas dragged down, the whole thing sent round as a PDF. Whoever takes it on is busy until lunchtime.
Notices sent out by hand
Appointment confirmation, reminder, status update: one email each, each of them typed separately. With ten customers that works; with two hundred it does not.
Six routines that look much the same in almost every business.
Capturing documents and invoices
The invoice comes in as a PDF or a photograph, the software reads out supplier, amount, date and number and puts it up for approval. Anything that stays unclear goes on a small pile for a person.
Moving data between systems
The order from the shop appears in the accounts, the address from the form appears in the CRM. Entered once, the same everywhere, with nobody catching up in the evening.
Reports that build themselves
The same report, the same sources, in the inbox first thing every Monday. As a page that always shows the current state, if you would rather not hunt for an attachment.
Notices and reminders
Appointment confirmation, a reminder two days ahead, a note about an unpaid invoice. Triggered by a rule, not by somebody happening to remember.
Approvals with a route and a deadline
Who checks, who signs off and what happens when nobody has reacted after three days. The route is settled in advance and can be followed afterwards.
Visual checks by camera
Counting, sorting, spotting the odd one out: a camera checks what somebody inspects by eye today, and flags the cases a person needs to look at.
The sum is simple: how often a month, times how many minutes
Automation rarely fails on the tools and usually fails on the wrong first process. So it does not start with technology, it starts with a multiplication anyone in the business can check. And it has two sides: what the routine stops costing, and what now fits into the same week.
How often, times how long
Count how many times a month a task comes round, and time the minutes honestly, once. Twelve minutes, eighty times a month, is sixteen hours. You can hold a conversation about that number; you cannot hold one about a feeling.
The first process is the dullest one
Not the most visible one, and not the one people complain about loudest, but the one that repeats itself in exactly the same shape most often. Dull here means predictable, and therefore cheap to build and easy to check.
What sits on the other side
Building costs once, running costs monthly, and the routine wants looking after as soon as something changes. If the sum does not add up over a year, we say so before we start.
Most of what saves time needs no AI at all
Rules are cheaper, faster and checkable: they do exactly the same thing every time. A model only comes in where something has to be read or sorted that looks different on every occasion.
What rules take care of
When A happens, do B. An order arrives, so write the document and adjust the stock level. Chains like that run unchanged for years, cost almost nothing to run and give the same result every time.
Where AI genuinely helps
When every supplier invoice is laid out differently, when enquiries written in plain language have to be sorted, when a picture has to be judged. That is where a rule becomes endless and a model belongs.
Why the mixture wins
In most routines AI handles a single step, the reading or the sorting, and the rest stays rules. That is cheaper to run and easier to inspect on the day something does not add up.
What runs on its own and what needs a signature
Not every step belongs to the machine. We settle the line with you beforehand, and it falls where a mistake costs money or trust.
What runs without asking
Steps that can be undone and hurt nobody: moving data, filing documents, sending a report, triggering a reminder. Asking first adds nothing here except waiting.
What needs sign-off
Payments, contracts, anything with a deadline and anything that goes outside the house. The software prepares it, a person confirms with one click. That takes seconds instead of minutes and is still a decision.
What the machine did
Every step is logged: what was triggered, with which values and with what result. When something goes wrong you can see where, instead of guessing.
Five stages, and the first one is measuring, not building.
- 1
Measure where the time goes
Two weeks of writing down who does what and how often. The most expensive routine is often not the one people complain about; it is the one everybody treats as perfectly normal.
- 2
Choose the first routine
We work through effort and return for two or three candidates and take the one with the clearest answer. If none of them clears the bar, we stop here and tell you so plainly.
- 3
Build it
Rules, connections to your existing systems and, where it is needed, a model to read or sort. Built with your real documents rather than with sample data.
- 4
Run it alongside the old way
The new routine runs next to the old one for a while and you compare the results. The manual way is switched off only once you believe the new one.
- 5
Add the next one
The second routine costs less than the first, because the connections, the approvals and the log are already standing. The whole thing grows in steps you decide on one at a time.
When we advise against it
- If the routine changes every few weeks: whatever has not settled down gets automated along with it, and the constant rebuilding costs more than doing it by hand.
- If it happens three times a year: two hours of manual work a year will not carry a routine that wants looking after.
- If nobody on the team wants to change how they work: a routine everyone walks around costs more than no routine at all.
- If the real problem is missing data: then too much manual work is not the trouble, a gap is, and no automation closes it.
What does an automation cost?
It depends on how many steps the routine has, how many systems it connects to and whether something has to be read that looks different every time. That is why we do not name a figure before we know. At the free initial consultation we work out with you what the manual way costs today, and you then receive a fixed price for exactly that scope. On top of that come the running costs, and the usage charges of a model if one is doing the reading.
Free initial consultationHow long before something is running?
A single, clearly bounded routine is usually working within a few weeks, because we deliberately start with one rather than with ten. We name a date only once the scope is settled and we know how your systems can be connected. Anything earlier would be a guess.
Will anyone lose their job over this?
That is not how we sell it. What disappears are steps, not people: the retyping, the copying, the chasing up. Where automation gets started, the reason is usually the opposite one: the work has grown and nobody can keep up. Whoever enters documents today reviews the cases that call for judgement afterwards.
What happens when the automation gets something wrong?
That is exactly why it runs alongside the old way at the start and you compare the results. After that, three things protect you: sign-off wherever money, deadlines or the outside world are involved, the log that records every step, and the pile for the cases the routine itself flags as unclear. No technology is faultless, and anyone promising you that is promising too much.
Do we have to replace our existing systems?
As a rule, no. We build on what you have and connect it through APIs: accounting, till, stock, shop. Where a system offers nothing outwards at all, there is usually still a way through files or an export. And where there genuinely is not one, we say so beforehand rather than afterwards.
Are the off-the-shelf tools not enough for this?
Often they are, and we will tell you so. For simple chains between two services, tools such as Zapier or Make are quicker to set up and cheap to keep. Building it properly earns its keep where many steps hang together, where documents have to be read, where approvals and a log belong to the job, or where the data must not leave your own house.
Tell us which task comes round most often in your business.
At the free, no-obligation initial consultation we work out with you how much time sits behind it, what you save and how much more the same team can handle. If it does not pay off, we say so. We reply within 24 hours.
Book an initial consultationJust a quick question? The AI assistant in the bottom right answers straight away.
Ready for the summit?
Tell us about your project. We reply within 24 hours with an honest assessment. Free and without obligation.
hallo@aurphi.chAurPhi MoukrimMarktstrasse 18, 8853 Lachen SZ, SwitzerlandPostal address, no walk-in service. Appointments by arrangement.