Orders arrive as PDF, as email, as a photo of a handwritten note. Someone on your team types every one into the system.
Every customer sends orders in their own format. Your team translates them, line by line, into yours. It works, until the volume grows, the rush order jumps the queue, or a wrong quantity slips through.
You know this day.
Orders arrive as PDF, as email text, as a photo of a handwritten note, and someone types each one into the system.
- Orders arriving as PDF and retyped
- Every customer orders in their own format
- Article numbers translated from their catalogue to yours
- Rush orders jumping the queue by phone
- A mistyped quantity discovered at the loading dock
How many orders a week does someone type in by hand?
How it gets fixed.
Capture and read.
Orders keep arriving exactly as they do today: email, PDF, portal, scan. Every field, customer, positions, quantities, dates, references, is read and checked against your article and customer data. Nobody changes how your customers order.
Check and route.
Clean orders go straight through. Anything uncertain, an odd quantity, an unknown article, a price that does not match, goes to your team with the doubt highlighted. People handle exceptions, not everything.
Into your system.
The order lands in your ERP the way a careful person would have entered it, with every value traceable back to the document it came from. Your system, your workflow, unchanged.
One order, read and checked.
Nordmann & Sohn GmbH
Industrievertrieb · Hamburg
| Article | Description | Qty |
|---|---|---|
| NM-2205 | Hex bolt M8x40, zinc | 500 |
| NM-1180 | Washer M8, DIN 125 | 500 |
| NM-3320 | Threadlocker 243, 50ml | 12 |
deliver Thursday if possible
Read and checked
Customer
Nordmann & Sohn GmbH → C-1047
Position 1
NM-2205 → BX-88401, Hex bolt M8x40 zinc, qty 500
Position 2
NM-1180 → BX-88112, Washer M8, qty 500
Position 3
NM-3320 → BX-90455, Threadlocker 243, qty 12
Margin note
Captured as delivery remark
Requested date
“Thursday”, which week?
Sent to a person to confirm before the order posts.
Illustrative example with fictional data.
It fits around you, not the other way round.
Your customers will not change how they order. They do not have to. The system learns your document mix, including the ugly ones. Especially the ugly ones.
Your ERP stays. We build into the system you run, through its interfaces or around its missing ones. No migration, no new tool for your team to learn.
Accuracy is measured, not promised. Before anything goes live, it runs on your real orders, and the result is a number we agree on upfront. If the Value doesn’t hold that number over the proving period, the attempt was free.
What about sensitive customer data?
The most frequent question we get, and the most important one.
The assumption is usually that AI goes against compliance, especially by feeding client data into it, and that assumption can be true and dangerous at certain times, with certain solutions.
So how do we go about it, so that AI is okay to use for your clients and for the law?
- We start with the data, not the tool.What does the process actually touch: personal data, financial records, health records? The rules attach to the data first, and they already govern your email, your document system, your CRM. A system we build is one more processor under rules you already live by, not a new legal universe.
- We map the rules that apply, during discovery.By the data, by what the software is for, by AI usage itself where applicable (for example, the EU AI Act), and by where your company, the system, and its users sit. When a heavier regime applies, we tell you before anything is built, what it means for cost, and whether the Value survives it.
- We build with the rules as architecture, not paperwork.Those rules define the architecture. The most common ones: hosting where the data must stay. Only the data the process needs. Access controlled and logged. Contracts with every provider that touches it. Client data does not train public models. And decisions with certain effect on people route to a person by design; nothing decides silently.
- Everything is written down, and checked by the right people.The rules we found, the choices we made, the reasons why: documented. We research the rules and build to them, but the legal judgment belongs to legal experts: your counsel reviews our findings, and if it gets complex, we work with them directly. We do not offer legal advice. We offer everything a legal expert needs to give you theirs, quickly.
So, what does this mean for you?
Nothing to be scared of. These rules exist to protect people, your clients included, and that is a goal we share, not an obstacle. Should you drop your idea out of fear of complexity? It is often easier than it looks, and when it is not, you deserve to know early. That is one more reason discovery is free: we find out together how heavy the rules sit on your idea, before it costs you anything.
What would a build look like?
It depends on the process, and the range is wider than most people expect. At one end it is a single focused artifact your team opens in a browser, no integration, no infrastructure. At the other it is a system wired into the software you already run, with monitoring, an evaluation suite and a handover so your own people can own it. We scope to what's needed to deliver the value reliably.
How long does a build take?
Anywhere from about a week to a couple of months. A focused artifact can be in your hands inside a week or two. A system that has to read from and write into your existing software can take longer.
Who maintains it?
It ships with monitoring, an evaluation suite, and documentation your own team can run. We stay on if you want us; you are not hostage if you don't.
What does it cost?
Discovery is free of charge. That's intentional: discovery is where we both find out whether this is worth doing. After it, a fixed build quote with the thresholds written in. You pay only when the value holds the thresholds we agreed.
The people on this page are the people on your project.
Dion Kasumi
Solution Architect
Translating business goals into architecture.
Rina Bekolli
Delivery Lead
Owning delivery from scope to handover.
Qendresa Potoku
AI Engineer
Building reliable AI systems.
Blina Retkoceri
Systems Engineer
Standing the systems up and keeping them stable.
Elion Emrullahu
AI in Finance
Putting it to work inside finance teams.
No handover to a delivery team you never met. The engineers who map your process are the engineers who build it.
Find out if it’s worth it.
Bring one real order to the call. The ugliest one you have. We’ll tell you in 30 minutes whether it’s worth pursuing, and what it would have to deliver to be worth it.
Straight into a calendar, no email back and forth. Microsoft Teams.