Workflow Automation
Replace the Manual Hours. Keep the Valuable Work.
Every business has a handful of jobs that cost hours and create nothing new. Retyping the same figures, chasing the same replies, moving the same rows between two screens. We find them first, then build the systems that take them over, so the hours go back into the work that actually earns.
Examples
What This Looks Like.
Invoices that book themselves
Supplier PDFs arrive all week. Someone opens each one, retypes the lines into the accounting system and files it. Instead: the numbers get read, checked against the order, and booked. You only look at the ones that don't match.
Enquiries that never sit unanswered
A form comes in at ten on a Friday night. Rather than waiting for someone to notice on Monday, it gets an answer straight away, the company details looked up and attached, and it lands with the right person with a reminder on it.
An inbox that sorts itself before you open it
Half your mail is the same five questions. They arrive grouped by what they're about, with a reply already drafted in your own wording. You read it, fix what's wrong, send.
The Monday report, already made
The numbers you assemble by hand every week — from the till system, the accounting, the ad accounts — collected, cross-checked and sitting in one place before you sit down.
Research that takes minutes instead of days
Our Supplier Agent does exactly this: a sourcing brief goes in, a ranked and verified comparison comes out. 40+ hours of desk research, done in minutes.
The copy-paste between two systems that don't talk
An order in one place has to become a job in another and then an invoice in a third. Right now a person retypes it twice and occasionally forgets. The automation moves it, every time.
In Every Build
Every automation ships with this.
A screen you can actually use. Somewhere to see what ran, what it skipped and why — not a black box that either works or doesn't.
Runs on your data, in the EU. Servers in Frankfurt, nothing sitting with a US provider.
A human approves anything that leaves the building. The system prepares, you release.
Edge cases mapped before we build, and flagged for review after.
Error handling and alerts, so you hear about a failure from us and not from a customer.
Documentation, and a walkthrough with the people who use it.
How We Work
Fixed Scope. Fixed Price.
The same shape every time. Nothing gets built before the scope is written down and the price is agreed, because that is the only way a fixed price is safe for both of us.
- 01
Workflow audit
1 week · freeWe sit with the people who actually do the work and map where the hours go. You keep the map whether or not you hire us.
- 02
Scope and price
Before we buildWhat it does, what it does not do, and what it costs. Fixed, in writing, agreed before anything starts.
- 03
Build and pilot
2–4 weeksIt runs beside the manual process first, on real data, until the output is trusted. Then the manual version stops.
- 04
Live, and ours to keep running
OngoingMonitoring, fixes, and changes as your process changes. Part of the partnership, not a support tier you buy.
Questions
The Ones People Ask.
The same handful comes up in every first conversation. Here they are answered before you have to ask them.
What does it cost?
Most single automations land between EUR 500 and 1,500. You get a fixed price before anything is built, and we do not bill by the hour, so the number you agree is the number you pay. What moves it: how many systems have to be touched, whether they have a usable API, and how many exceptions the process really turns out to have. The audit answers all three, and the audit is free.
How long before it is running?
A single automation is usually two to four weeks from agreed scope to live, including the period where it runs alongside your manual process. The audit that precedes it takes about a week.
Where does our data go?
Into infrastructure we run for you in the EU, with server functions in Frankfurt. Nothing is stored with a US provider, and nothing you send is used to train anyone's model.
What happens when the AI gets something wrong?
The design assumes it will. Models are used for the parts they are good at, like reading and drafting, and the decisions that matter run on ordinary deterministic code. Anything that leaves your business waits for a person to approve it, and anything the system is unsure about is flagged rather than guessed.
Do we need someone technical on our side?
No. You need the person who currently does the work to spend an hour with us explaining it. We build, host, monitor and maintain the rest.
What if the process changes after it is built?
Small changes are part of the partnership and are not separately invoiced. If the change is genuinely a new piece of scope, we quote it fixed-price the same way we quoted the first one. The line is whether the thing already existed.
Are you going to automate someone's job away?
In our experience the work that gets automated is the work nobody wanted: the retyping, the chasing, the copying between two screens. The job usually stays and gets better. If you are automating in order to cut a role, say so up front, because it changes what we should build.