Skip to content
App Crafters
Book a call

For any company where people retype, copy and chase all day

If your team does it the same way every day, software can do it

Retyping supplier invoices into the ERP, copying orders out of email, chasing unpaid invoices, building the same Monday report from three spreadsheets. We automate that kind of work for companies in any sector - hotels, clinics, wholesalers, accounting offices, shipping offices, shops - on the systems you already use. A person still approves anything that moves money or goes to a customer.

Talk through a taskTry the invoice demo
extraction-benchworked examples
  • supplier invoice, clean PDF: 7 fields read, ΑΦΜ valid, totals add up
  • same invoice, sent as a fax: invoice number at 76%, sent to a person
  • delivery note, ΑΦΜ 080554493: check digit fails - number cannot exist, held
  • printed total €3,748.80: lines plus 24% VAT make €3,868.80, held
  • CV, notice period: read at 71%, under the threshold, to review
  • every field cleared or confirmed: only then marked ready to transmit

Below, one repetitive task running in your browser: supplier invoices read, checked, and the doubtful ones sent to a person. Give it a bad scan and see what it refuses to guess.

In plain language

What is business process automation?

Business process automation is software doing the repetitive steps of everyday work that people currently do by hand: reading a document and typing it into another system, copying data between two programs that do not talk to each other, sending the same reminder or confirmation, matching payments to invoices, putting together the same report every week. The rules stay yours - what gets checked, who approves what, when it stops and asks. Some steps need AI, for example reading a scanned invoice or understanding a free-text email; most only need integrations and clear rules. We use whichever is cheaper and more reliable for each step, measure it on your own data, and build it to stop and ask a person when it is not sure rather than guess.

Words that get used interchangeably and should not be

Workflow automation
A chain of steps that starts on its own when something happens: an order email arrives, the order is created in the ERP, the warehouse gets a pick list, the customer gets a confirmation. Most everyday automation is this.
Integration (API)
How two systems exchange data without a person copying it across. Where a system has no API we work from its export or its database, and we say so rather than calling it an integration.
RPA
Software that clicks through screens the way a person would. Useful for old systems with no other way in, fragile the day a screen changes. We reach for it last.
AI in automation
A model for the steps fixed rules cannot handle: reading a scan, understanding an email written any which way, drafting a reply. One tool among several, measured like the rest, and it has to say when it does not know.
Approval by a person
The software prepares, a person confirms. The correct design for anything that moves money, signs something, or talks to a customer.

One repetitive task, automated: supplier invoices

Typing supplier invoices into the accounting software is the repetitive task we hear about most, so here it is running. Each value lights up where it is printed, with the confidence it was read with. Two checks are real arithmetic, the ΑΦΜ check digit and the invoice totals, and a failed check or a doubtful field goes to a person instead of into your books. Switch the scan to a phone photo or a fax and watch the confidence drop instead of staying green.

  1. Pick a document and run the extraction.
  2. Switch the scan to a phone photo or a fax and run it again.
  3. Accept or correct a field from the review queue.

0/3

extraction-benchinteractive

Document

Scan quality

Fields under 90% go to a person. Drag it and watch the page.

Run it and the values light up where they sit.

Change the scan quality first if you want to see it struggle.

Runs entirely in your browser. Nothing you click here is sent anywhere. The ΑΦΜ and totals checks are computed for real; the confidence scores are a simulation of the real pipeline's behaviour, not a live model call.

Nine in ten Greek companies have not started with AI

AI is only one part of automation, but it is the part that reads documents and emails, and it is the part with official numbers: Greek companies use it at well under half the EU rate, and the EU rules for it already apply. Whoever takes the repetitive work off their people first gets those hours back first.

8.2%
of Greek companies with ten or more employees use any AI system at all - 3,868 out of 47,246, counted by ΕΛΣΤΑΤ in December 2025
Source: ΕΛΣΤΑΤ, ICT use in enterprises 2025
20.0%
of EU enterprises used AI in 2025, nearly two and a half times the Greek rate
Source: Eurostat, 20% of EU enterprises use AI
2 Aug 2026
most of the rest of the EU AI Act applies from that date, transparency rules included - an automation that writes to your customers is something you have to be able to declare and document
Source: EU AI Act, implementation timeline

The daily tasks we automate most often

Six kinds of task come up in almost every company, whatever the sector. If yours is none of them we will say so on the first call rather than bending your problem to fit what we already know how to do.

  • Documents typed into your systems

    Supplier invoices, delivery notes, receipts, onboarding paperwork - whatever arrives as a PDF, an attachment or a phone photo. We read the fields, check them against what you already hold (ΑΦΜ, totals, the purchase order) and post them to the ERP or accounting software. A wholesaler with a few hundred supplier invoices a month or an accounting office with a client's shoebox of receipts is the typical case. Anything doubtful waits for a person, and myDATA documents are validated before they go out.

    • Invoices
    • ERP
    • myDATA
    • OCR
    • Review

  • Orders and requests out of the inbox

    Orders that arrive by email or Viber and get retyped into the system, quote requests, group booking requests at a hotel, appointment requests at a clinic. The automation reads the message, creates the order or the booking as a draft, and flags what is missing - a product code it does not recognise, a date that does not exist - instead of guessing.

    • Email
    • Orders
    • Bookings
    • ERP

  • Payments, reminders and reconciliation

    Bank statements matched against open invoices, card payouts matched against sales, reminders to customers who are late paying, sent on a schedule you set. The matches it is sure of are marked; the rest go to a list for a person. Reminders go out only in the wording and to the customers you approved.

    • Reconciliation
    • Reminders
    • Banks
    • Viva Wallet

  • Reports that put themselves together

    The Monday report someone builds by hand from three spreadsheets and an export - sales, occupancy, stock, open invoices. It arrives by email or on a dashboard at the time you choose, built from the same sources every time, so the numbers stop depending on who made it that week.

    • Reports
    • Excel
    • Google Sheets
    • Dashboard

  • Systems that keep each other up to date

    Stock that has to match across the shop, the e-shop and a marketplace. Customers entered twice, once in the CRM and once in the ERP. Appointments that need a reminder the day before. We connect the two systems so the update happens once, and log every change so you can see what moved and why.

    • Integrations
    • API
    • Stock
    • Sync

  • Answers to the questions you get every day

    Opening hours, prices, what documents to bring, where an order is, booking confirmations and the follow-up after a visit. Answers come from your own material, in Greek and English, and are prepared as drafts or sent only where you allowed it. When it does not know, it hands the conversation to a person instead of making something up.

    • Customer service
    • Greek
    • Confirmations
    • Handover

What it connects to in a Greek company

An automation that cannot reach your systems is a demo. These are the connections that come up in Greek companies over and over.

myDATA / ΑΑΔΕ
Documents validated against the schema before transmission, so a rejection is caught here rather than at the tax authority.
ERP and accounting
SoftOne, Entersoft, Epsilon, SAP, or the bespoke thing built in 2009 that everything still depends on.
Gmail, Outlook and shared inboxes
Where most Greek business processes actually live. Orders, invoices and requests picked up straight from the inbox.
Excel and Google Sheets
The spreadsheets the business really runs on. We read them, write to them, or replace the copy-paste between them.
Banks and payments
Bank statement exports, Viva Wallet and card payouts, matched against what was invoiced.
Viber and WhatsApp
Viber is where Greek customers reply. An automation that only lives on your website is answering the wrong channel.
E-shops and booking systems
WooCommerce, Shopify, marketplaces, hotel and clinic booking software - wherever stock, orders or appointments live.
CRM and your own database
HubSpot, Zendesk, Postgres, MySQL, or a read-only copy that keeps us well away from production.

What it costs, starting with one task you can stop

Naming numbers because the alternative wastes a call. These are the bands we quote in; the scoping engagement fixes yours before any code is written.

  • First automation, one task

    €3k – €6k

    One task - supplier invoices into the ERP, say, or payment reminders - running on your real data for a month with a person still checking it. You see what it gets right, what it sends to review and how much time it takes off your team, then decide whether to go further. If the answer is no, that was the cheapest way to find out.

  • Automation in daily use

    €7.5k – €17.5k

    The first automation made solid: monitoring, a log of every action, a review queue your team owns, and the integrations wired properly rather than held together for the demo. Often a second or third task on the same systems.

  • Several processes, or data that cannot leave

    €12.5k – €22.5k

    A connected set of automations across departments, or a deployment on your own infrastructure or an EU data centre when the data must not leave, open-weight models included. Hardware and hosting are yours and are not in this number.

Most of this spend qualifies under ΕΣΠΑ Ψηφιακός Μετασχηματισμός ΜμΕ, which subsidises 50–60% of software costs. We are not consultants for the application and we will not pretend to be, but we will write the technical description your consultant needs.

What we measure before you trust it

Every page on this site argues that a number measured on your own data beats a vendor's promise. That is cheap to say, so here are the four we would actually put in front of you.

Accuracy on your own data
Scored on your real invoices, orders or emails, not ours. You get the number and the failures together, because the failures tell you whether the number is usable.
Share going to a person
How much someone still has to look at. If it falls too fast it usually means the threshold got loose, not that the automation got better.
Errors that got through
The expensive ones. Found by sampling what went through without review, because by definition nobody flagged them.
Hours given back
Time per task before, measured with your team over a normal week, against time after, including the minutes spent on review. Leave the review minutes out and the number flatters itself.

What we have shipped

Work for clients and tools we run in public, each labelled as which it is. Where the tool is the demo above, you can check it yourself instead of taking our word for it.

  • Client work

    Wholesale distribution · 2026

    The catalogue that builds itself, and the invoices behind it

    A seller between producers and buyers was spending most of the week rebuilding catalogues, writing the same outreach by hand and typing invoices out of orders. We built the portal that does all four from the same data, so the week goes on selling instead.

    • Next.js
    • TypeScript
    • Postgres
    • myDATA
    • Resend

    Read the build

  • 3

    Validations before anything is written: ΑΦΜ check digit, line-item sum, VAT re-derived at the stated rate

    Our own tool · Wholesale and distribution · 2025

    The supplier invoice queue, read once by a machine

    Somebody in accounts opens a few hundred PDFs a month and retypes seven fields out of each one. The expensive part is not the typing - it is the transposed ΑΦΜ that nobody catches until ΑΑΔΕ does. This is the queue that reads them and refuses to guess.

    • Python
    • Claude API
    • Postgres
    • myDATA
    • Next.js

    Read the buildIt runs in the demo above

What we can build

None of these exists. Each is a system we have specified for this sector at the level of detail we would start from: the parts that are hard, the decisions that would have to be made early, and what would be worth measuring once it runs.

Questions we get asked about this

How much does it cost to automate a task?
A first automation on one task is €3k – €6k and runs for a month on your real data with a person checking it. Putting it into daily use, with monitoring, a log and a proper review queue, is €7.5k – €17.5k. Several connected processes, or a deployment on your own infrastructure, is €12.5k – €22.5k. You get a band on the first call and a fixed price in writing before any code is written.
Which tasks should we automate first?
The ones done often, the same way, by rules you can explain to a new hire, with the data already in an email, a PDF or a spreadsheet. Do the arithmetic before calling us: a task that takes 20 minutes and happens 15 times a week is 5 hours a week. Retyping invoices, copying orders, reconciling the bank and building the weekly report are usually near the top. Tasks that need judgement every time are usually the wrong place to start.
Will it replace our staff?
It replaces tasks, not judgement. The person who retypes invoices today usually knows best where they go wrong, and becomes the one who handles the exceptions the automation sends to review. What you do with the hours, grow without hiring or move people to better work, is your decision, and we will not pretend it has nothing to do with how you plan headcount.
What happens when it makes a mistake?
It is built to stop and ask rather than guess: a doubtful field, a failed check or a case it has not seen goes to a person. Anything that moves money, signs something or writes to a customer stays behind an approval until you decide otherwise. Every action is logged, so you can see what it did and undo it.
Does it work with our ERP and accounting software?
Usually. SoftOne, Entersoft, Epsilon and SAP come up regularly, as does bespoke software nobody has documentation for. Where there is no API we work from the export or the database, and we say so plainly. Documents bound for myDATA are validated before they are sent.
How long until it is running?
It depends on the systems more than the task. One task on systems with a usable API or export is a matter of weeks, not months, followed by the month it runs with a person checking. The date is in the written scope, along with what happens if we miss it.
Do we need AI for this?
Often not. Most of the work is integrations and clear rules, which are cheaper and more predictable. AI earns its place where a step means reading a scan or understanding an email written any which way, and there we test it on Greek material specifically, because models still do worse in Greek than in English.
Where does our data go, and what about GDPR?
The data stays yours. We pick providers and settings that keep it in the EU and do not train on it, and where it must not leave your infrastructure at all we run everything there, open models included. Which of the two applies to you, and who can see what, is written into the scope.
Is this eligible for ΕΣΠΑ funding?
Software and equipment spend under Ψηφιακός Μετασχηματισμός ΜμΕ is subsidised at 50–60%, and custom software development generally qualifies. We are not ΕΣΠΑ consultants and will not advise on your application, but we will write the technical specification and costing your consultant needs.

The other four

Bring us one task, not a brief

The most useful first call is about one thing somebody on your team does every day or every week that software could do instead. Twenty minutes, and you will leave knowing whether it is worth automating and roughly what it costs.

Or write to hello@appcrafters.gr.

What it does, who it is for, and when you need it.

Rough budget

A rough range saves us both a call. Optional.

Or grab a slot in the calendar