KSeF in an online store: the order, invoice and return in one case
The system links the return with the order and the invoice and prepares a complete set for the corrective invoice. Accounting approves.
Automation, paid once
from €3,000 net
Test on your process: €0
You pay 50% after acceptance
How much does KSeF automation for an online store cost, and when does it pay back?
Short answer
Automating KSeF, Poland's national e-invoicing system, for an online store costs from €3,000 net. In our scenario it pays back in 1.9 months. Today, clearing up 800 returns and corrections by hand costs €1,861 a month. With the system, the same work costs €286. Accounting approves the corrective invoice. The system gathers the order, invoice and return into one case. The test on your process is free.
Statistics
KSeF and e-commerce statistics 2026
1 February 2026
According to Poland's Ministry of Finance (2026), every company in Poland has received its invoices through KSeF since 1 February 2026.
What clearing up returns and corrections costs you today
€1,861
a month for clearing up 800 returns and corrections by hand
Today€1,861
With the system€286
Payback: 1.9 mo.
In your company
Small store
400 returns · €931 a month
3.8 mo. payback
Mid-sized store
800 returns · €1,861 a month
1.9 mo. payback
Large store
2,500 returns · €5,817 a month
0.6 mo. payback
The team gets back about 45 hours of work a month, or 6 working days, and the company €18,900 a year. We count after the time a person spends checking and after the model cost.
800 × 4 min = 53.3 h × €34.90 = €1,861 a month. With the system, a person checks the output in 15% of the old time (€279) and the model costs €6.43. That leaves €286. The saving is €1,575 a month, so €3,000 pays back in 1.9 months.
Assumptions: 4 minutes per return or correction to clear up and checking in 15% of the old time are our assumptions. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat, for companies with 10 or more employees. Model cost: Claude Sonnet 5.5, about 2,500 + 400 tokens per return, at the ECB rate of 1.1205 USD per euro on 22 October 2026.
Customer service looks for the order and the invoice
Team
Writes to accounting about what to correct
Team
Accounting asks about the amount and the date
Team
The corrective invoice waits
Team
With the system
A customer returns an item
Client
Links the return with the order and the invoice
System
Prepares a complete set for the correction
AI
Accounting approves the corrective invoice
Person
The case closes with a full record
System
What stays with people: Accounting approves how the corrective invoice is issued. The system gathers the order, the invoice and the return into one case.
foap.com customer service: 95% of replies are sent automatically.
Users from several countries write to foap.com, in several languages. The system reads the entire conversation and the terms of that person, and then sends a response in that person's language. Questions about money and exceptions go to the team with a ready draft, so they have to check and send them.
the system sends the answers itself, without human typing
95%
the system sends the answers itself, without human typing
time from the arrival of the e-mail to the ready response (median)
30 s
time from the arrival of the e-mail to the ready response (median)
e-mails that require a response receive a ready-made draft
100%
e-mails that require a response receive a ready-made draft
less work for every 1000 answers (estimate)
47 h
less work for every 1000 answers (estimate)
Data from foap.com production, October 20-21, 2026. Estimate: 3 minutes for handwritten response.
Our invoice coding system checks the invoice against the order and the delivery confirmation and holds it when there is a difference, with a specific reason. In an online store that reason is often a return. Demonstrator on synthetic data.
“We started with my own AI agent. During the training that followed, I learned to build new applications myself. Today I am bringing this way of working with AI into my business. Working with Syntalith gave me exactly the skills and tools I needed.”
A complete set of data for accounting to issue the correction
A queue of returns with gaps
Integration with the order system the store already has
What happens to your data
You choose the test cases. We anonymise them on request.
They stay in the EU and we delete them after the test.
We sign the NDA and the data processing agreement before we receive anything.
After acceptance you keep the code, the test set and the documentation. You pay for licences and model usage directly to the provider. With several departments or companies, the engineer suggests a Process Audit from €600 net: a map of the flow and a fixed quote. We deduct the audit fee from a build signed within 3 months.
Questions
How much does automating returns and corrections in a store cost?
From €3,000 net, paid once. After the 5-day test the build takes 2 to 6 weeks. You get a fixed price before the contract: 50% on signing and 50% after acceptance on your own returns.
Would the link between our store and accounting software be enough?
Start with it if you have it. The system makes sense for returns that the link lets through incomplete: partial ones, late ones or ones from another sales channel.
When does it not pay off?
Below 130 returns a month the payback takes more than a year. Then ChatGPT or Claude for the store team is enough, from €1,150 net.
Who issues the corrective invoice?
Your accounting or sales software, as it does today. The system prepares the complete set of data, and accounting approves the correction.
What do we need to prepare?
Access to the order system, 20 to 100 returns from recent months with their corrections and the person in accounting who approves them today.
What our OpenAI and Anthropic partnerships mean for you.
OpenAI makes ChatGPT and Anthropic makes Claude. We belong to the partnership programs of both companies, so we work with their tools every day and learn from the creators themselves.
Claude Partner Network
01We will select a tool for the task
We implement both ChatGPT and Claude. We check both on your documents, and you choose the one that makes fewer corrections and saves more time.
02Knowledge straight from the creators
We learn from official OpenAI and Anthropic courses and materials. Your team learns about the features and settings that work today.
03Licenses without our margin
You buy accounts directly from OpenAI or Anthropic and receive an invoice from them. We do not add anything to the license.
04Certified engineers
On the team, we have an AWS certification in machine learning and an Anthropic certification in systems architecture on Claude.
Participation in programs does not mean that OpenAI or Anthropic endorses our services.
customer service
