Evidence

What AI already does in Polish companies: five documented examples

AI in business is often discussed in the future tense. Here we describe five working deployments in Polish organisations, and all the figures come from public announcements. Current as of August 2026.

23 June 2026

Bank Pekao: over 1.5 million documents a quarter

The przeczytAI system at Bank Pekao reads and classifies banking documents such as contracts, income certificates and scans, then extracts data from them. According to the bank's announcement of 30 April 2025, it processes over 1.5 million documents a quarter. Programs the bank calls virtual workers perform over 15,000 tasks a day: handling product applications, registering credit insurance policies and configuring accounts. Previously, this work required manual reading and retyping.

PZU: undisputed claims paid the day after filing

In Samoobsługa NEXT, PZU's self-service channel, generative AI reads the customer's description of the loss and compares it with the terms of the PZU Dom policy. This covers straightforward, relatively low-value claims in selected categories of that policy. In undisputed cases, the customer receives the payout as soon as the day after filing. In its announcement of 3 February 2025, PZU calls this Poland's first production insurance process using GenAI, meaning a solution that handles real customer claims. The company has said it will verify around 12,000 claims a year after the first deployment phase and says it analyses around 250,000 property claims annually.

Comarch: over 10 million documents read

Comarch OCR reads invoices and accounting documents. It recognises the tax ID, document number, dates and amounts, then transfers the data to the ERP system. According to a Comarch announcement of 29 April 2021, the company's servers had processed over 10 million documents since the service launched in June 2019. In that same 2021 announcement, Comarch reported over 90 percent accuracy for the document formats it supports. The data goes to an accountant, who approves the posting.

TAURON: failure warning on a 460 MW unit

At the Łagisza power plant, the RSIMS system, built by the Kraków-based company ReliaSol, monitors a 460 MW power unit. It detects unusual equipment behaviour and warns of a possible failure many hours in advance. The system has been used in day-to-day operations since late 2018. The OPTI AI UNIT development project, which extended the system, was worth over PLN 21 million. According to ReliaSol, the system predicts failures of the monitored equipment 3 to 17 hours in advance.

Żabka: checkout-free stores in six cities

Żabka Nano stores use computer vision to recognise what a customer has taken off the shelf and settle the purchase without a checkout. In its January 2022 announcement, the chain reported 25 autonomous stores in six cities and was then the largest network of its kind in Europe. In a January 2023 deployment case study, Żabka reported nearly 50 Nano stores and around 2 million products sold using the autonomous technology within a year.

What do these deployments have in common?

These deployments have three things in common. In each one, the machine takes over the first and most labour-intensive step: reading documents, claim descriptions, sensor signals or camera feeds at scale. The decision stays with a person: the accountant approves the posting, the claims adjuster approves the payout and the engineer responds to the warning. In Żabka Nano stores, the purchase is settled without an employee. Each organisation has also published a concrete number that can be checked at the source.

The numbers in this article relate to other companies, so they do not determine the result in yours. The result for your company has to be calculated from its data and compared with the process as it stands today. Point us to such a process in the contact form or at kontakt@uruchom.ai. Within one working day, we will send three questions about the process and propose a 45-minute call. We will prepare the first-stage outline after that call.