SellerFirma Demo Sp. z o.o. NIP 123-456-78-90BuyerKlient Sp. z o.o.NIP 987-654-32-10
ItemQtyNetAnnual licence16 400,00Implementation12 950,00Technical support2796,34
Net10 146,34VAT 23%2 333,66Total due12 480,00 zł
Process automation, AI and OCR — a studio from Gdańsk
Somebody in the company re-types the same things every day: invoices into accounting, form data into a spreadsheet, orders from email into the warehouse. The expensive part is not the hours — it is the typos that surface only when the month is closed.
Automation starts with one process, not with a revolution. The system reads the document, validation rules stop a wrong entry before it reaches the database, and the person stays where they are genuinely needed — on the exceptions and on the decisions.
What the scope covers
- Analysis of documents and rules on your real files, not on samples
- Document reading (OCR) instead of manual re-typing
- Validation rules that stop a wrong document before it is written to the database
- An exception path: what happens to a document the system does not understand
- The AI module as a separate component — it can be switched off or replaced without rebuilding the rest
- Processing in agreed infrastructure, with no content passed to external services
- Integration with ERP, warehouse, accounting and email
- An operation log: what the system did, when, and who approved it
- Hosting, SSL certificate, backups and monitoring configured within the implementation
- The technical side of GDPR — company documents almost always contain personal data
- Documentation of the rules, handover of accesses and full ownership of the code
Hosting, the SSL certificate, backups and the technical side of GDPR duties are configured as part of the implementation — they belong to the agreed price, not to a separate line item. Ongoing care of a running system, meaning updates, monitoring and support, is a separate subscription, quoted at handover.
How we work
Analysis of documents and rules
We take real files from the past weeks and look at what repeats and what is an exception. That decides whether automation pays off at all — and sometimes the answer is that it does not.
What we need from you A full set of sample documents, including the ugly ones: scanned crooked, annotated, with a stamp across the amount.
Prototype on your files
We show the reading on your documents, not on demonstration ones. It becomes visible straight away where the system is wrong and what has to be tightened in the rules.
What we need from you The person who re-types those documents today — they know where the traps are.
Deployment on one process
We launch a single document type, running in parallel with the current work. Until the results agree, the automation replaces nothing.
What we need from you Agreement to a period of parallel work and a person who will compare the results.
Extension to the next ones
Once the first process runs, we add further document types. Rules and integrations are already in place, so each next one is cheaper than the first.
What we need from you A decision on which process is next in line.
Handover and care
Documentation of the rules, training and handover of accesses. The rules are written so that they can be changed without us.
What we need from you The name of the person who will manage the rules on your side.
What work done by hand costs over a year
Work out what the hours spent re-typing data by hand cost per year.
The full employer cost, not the net rate — otherwise the result comes out too low.
This is your assumption. We have no number to put here — it depends on your process, not on your industry.
Fill in every field — without them there is nothing to calculate. We substitute no default values, because each of them would be our assumption rather than yours.
This is arithmetic from the assumptions you entered, not our forecast. Every number in the result comes from the fields above — there are no market averages and no estimates of ours in it.
Common questions
Do our documents go to an external AI provider?
By default no: the module processes documents in agreed infrastructure and the content does not leave it. If a particular project required an external model, we say so before the contract is signed — that changes your own GDPR documentation and is not a technical detail.
What happens when the system reads a document wrongly?
Validation rules stop it before the entry reaches the database. The document goes onto the exception path and waits for a person. We design it so that an error costs a check rather than a correction in the accounts.
How many documents does it take to be worth it?
There is no single threshold — repeatability matters, not volume. A hundred identical invoices a month automate better than a thousand documents that each look different. The answer comes from the analysis on your files, not from a table.
Will it replace people?
It replaces re-typing, not decisions. It usually ends with the same person handling exceptions and checking results instead of entering everything one by one.
Does the AI make decisions for us?
No. The module reads and proposes; the entry passes through the rules and, where it matters, through a human approval. Every operation stays in the log, so it is visible what the system did and on what basis.
What if the document format or the regulations change?
The rules are described separately from the code precisely for that. A change to the invoice format is a change to a rule, not a rebuild of the system, and after handover you can make it without us.
Can this be plugged into our current system?
Usually yes — we integrate with ERP, warehouse, accounting and email. If the system exposes no API, we look at other ways of exchanging data and say plainly when one of them is fragile.
Let us talk about your project
Pick a time for a call or leave your number — we call back during working hours.