AI automation, piece by piece
Some work in your business comes round every week and nobody wants to do it: copying orders, typing up delivery notes, cross-checking spreadsheets. We automate it with AI and clear rules, and someone on your team still has the final say.
Which tasks make sense
Not everything is worth automating. These usually are:
- Orders that arrive by email or WhatsApp. They’re read, sorted and entered into your management software without anyone typing them again.
- Supplier delivery notes and invoices. They’re read from the PDF or the photo, matched against the order, and anything that doesn’t fit is flagged.
- Reconciliation with your ERP. What the paper says against what the system says, line by line.
- Management reports. Sales, margins and stock every Monday, without someone spending the morning building the spreadsheet.
- Alerts. An item below minimum stock, a customer who stops buying, an invoice coming due.
If you recognise two of these, it’s worth a conversation.
How it works inside
A supplier sends you a delivery note as a PDF. This is how the machine handles it.
Drag the windmill to turn it
The wind: your data
The delivery note arrives the way it does today: attached to an email. A photo taken on a phone, a WhatsApp message or an export from your software work too.
The sails: capture
The document is collected and stored with its email, its supplier and its order. Everything that comes in sits in one place, with the date and time.
The wheel: AI and rules
The AI reads the delivery note: supplier, number, date, items, quantities, prices. Then your rules come in. Does it match what was ordered? Is the price the one on the price list? Do the lines add up? Anything that fails a rule doesn’t go through on its own.
The stones: the process
Once checked, it goes into your ERP as a goods receipt. The stock updates and, if anything is missing, whoever does the buying gets an alert.
The flour: the result
A delivery note recorded without typing, an alert when something doesn’t add up and, at the end of the month, a summary of everything that came in.
The AI suggests, you decide
The AI reads well, but not always. That’s why the machine is built to show when it’s unsure.
- Doubtful fields are flagged. If a figure is smudged, you’ll see the possible readings and why it leans towards one. In our demo, a delivery note has a quantity that could be 21 or 27: only 21 matches the line amount, and even so it asks you to confirm.
- Nothing goes into your system without a rule or a person approving it. You decide what goes through on its own and what always gets checked.
- Every change is logged. Who approved it, when, and what was corrected.
We’d rather give one warning too many than enter one wrong figure. If you want to know how we use AI on this website, it’s in AI use (in Spanish).
With what you already use
The automation connects to your management software, your email and your spreadsheets. You don’t have to change software.
If your software has no way to connect, we look at it during the diagnosis and tell you what the options are. Sometimes the answer is that it isn’t worth it, and we’ll tell you that too. When you need a piece your software doesn’t have, we build it: that’s what we do in custom software.
What we won’t promise you
We won’t tell you that you’ll save a certain percentage. We don’t know, and neither does anyone who tells you before seeing your business.
What we do is measure. During the diagnosis we look with you at how long the task takes today, how often it repeats and where errors creep in. Once the automation is working, we measure again. That before and after is the only number we’ll give you.
We won’t tell you the AI never gets it wrong either. It does, and that’s why the machine shows its doubts.
Try it in the demo
We’ve built a made-up electrical supplies distributor so you can see it working: it reads delivery notes and invoices, answers questions about its sales citing where each figure comes from, and has a management dashboard. Fictional company and data, real AI.
If your business is a distributor, also see automation for retail and distribution. If it’s an accounting firm, AI for accounting and law firms. And if what you need is something that answers your customers, AI chatbots.
Delivery note A26/1587 · what the AI read
| Code | Item | Qty | Amount |
|---|---|---|---|
| CAB-0010 | H07V-K 2.5 mm² cable, 100 m reel | 5 | €202.54 |
| CAN-0030 | Surface-mount junction box 100×100Doubtful | 21 or 27? | €41.15 |
| CUA-0027 | DIN-rail feed-through terminal 4 mm² | 62 | €39.09 |
The quantity is smudged. Only 21 matches the line amount: the demo suggests 21 and asks you to confirm it.
Questions about automation
Does my data leave the company?
It depends on each piece, and we tell you in writing before we start: which service handles each piece of data, and where. If something mustn’t leave, it’s designed so it doesn’t.
What if the AI misreads something?
Anything doubtful is flagged and checked by a person. Whatever fails your rules doesn’t get into the system. And every change is logged.
Who owns what you build?
You do. The code, the documentation and the service accounts are in your company’s name.
How long does it take?
We give you the timeline in the proposal, after seeing your case. We work in short deliveries, so the first piece that works arrives before the whole project does.
Tell us what you’d like to automate
A 30-minute call is enough to know whether it makes sense to keep talking.