⚙️
🌟 Najlepsze dopasowanie
Wpisz szukaną frazę, aby zobaczyć propozycje AI.

PARP study: readiness for artificial intelligence adoption in companies

PARP study: readiness for artificial intelligence adoption in companies

 

Only one in five Polish companies (22%) has a formalized digital data management policy, and 34% do not intend to develop one at all, according to a report by the Polish Agency for Enterprise Development (PARP). This is one of the main factors slowing down AI implementation. Cybersecurity is also becoming an increasingly significant challenge in this context.

 

Although entrepreneurs have become accustomed to the idea that openness to innovation and an adequate budget are sufficient for a successful digital transformation, the AI revolution shows that the digitization process is not limited to the IT department but requires the involvement of the entire structure. This is especially true when the success of introducing new tools depends on factors such as data quality and how it is collected and processed.

To invest in AI, you must invest in data

As stated in the PARP report “Artificial Intelligence in Companies: Readiness for Adoption, Competencies, and Needs,” no more than 38 percent of companies have a data management policy, with only 16% of those being informal. Every third enterprise does not intend to implement such a policy at all, while 45 percent consider their data to be standardized, and nearly half consider it up-to-date (51 percent) and complete (49 percent). Interestingly, the authors of the study claim that many enterprises only begin to realize the quality of their data once they start using AI.

Michał Żelichowski, Director of Business Development and Product Management at PSI Polska, agrees with the thesis that a conscious approach to the quality of data used in an organization is a sign of digital maturity. He reminds us that artificial intelligence is only as good as the data it receives. Therefore, it must be structured, properly described, and managed so that algorithms can use it effectively.

The PARP study shows that Polish enterprises currently use this technology primarily for content creation, text generation, and supporting marketing activities. Meanwhile, the greatest return on investment in AI occurs when it supports daily operational decisions and business processes from production planning and resource management to real-time response to disruptions – advises Żelichowski. – An algorithm that predicts machine failure or indicates the optimal production sequence can generate significantly more value for a company than a tool used solely for content creation.

Implementations in industry do not exceed 26 percent

Guided by operational benefits, one might assume that industrial enterprises are among the biggest beneficiaries of AI today. In reality, paradoxically, they are among those with the lowest level of artificial intelligence implementation. In the PARP study, the manufacturing and equipment repair industry ranks last, with only 14 percent of companies using AI. The situation is slightly better in the production of other goods, where the percentage of companies using AI is 26 percent.

The low level of AI implementation in industry is not due to a lack of interest in the technology, but because these are highly process-oriented industries where every change must translate into concrete operational results. Therefore, investment decisions here are more cautious and based on hard indicators such as OEE, timeliness, efficiency, or resource availability. In practice, we observe that AI only begins to bring real value when it is directly linked to production planning and control systems, rather than functioning as a separate analytical tool – says the PSI Polska expert.

Logistics needs tailor-made AI

On the other hand, logistics companies that decide to invest in AI implement solutions relatively broadly. 26 percent of enterprises have their own proprietary systems based on artificial intelligence.

Operators managing a large number of warehouses, a diverse fleet, and dynamic order volumes reach the point where manual planning or simple system rules are no longer sufficient more quickly. In such places, AI is used, for example, for dynamic route planning depending on traffic conditions and customer priorities, automatic scheduling of delivery windows, or real-time optimization of warehouse space utilization. Order picking also remains a key area, where even small improvements in process organization can bring noticeable operational effects. Shortening the order picking path directly affects the speed of order fulfillment and allows for lower operational costs while increasing warehouse throughput – explains Michał Żelichowski from PSI Polska.

This does not change the fact that industries where PARP respondents signal an urgent need for advanced information processing, such as energy (86 percent), modern business services (81 percent), or finance and insurance (80 percent), place much greater emphasis on AI implementation. According to Jakub Stadnik, Head of Delivery at Scalo, these industries reached the level where competitive advantage comes not from data access itself, but from processing it in near real-time and combining multiple information sources into a single decision-making model, the earliest.

In these conditions, AI does not constitute a separate layer of innovation, but a natural extension of existing systems, such as analytical platforms, transactional systems, or process management solutions. The key factor here is the maturity of data and IT architecture, which enables the integration of predictive models with decision-making processes without disrupting their continuity – emphasizes Stadnik.

The further into AI, the greater the fear for data security

In this regard, the Scalo expert reminds us that in AI models, it is important not only what data is used, but also where it is processed and how it is protected against unauthorized access or leakage.

More and more companies are consciously designing an architecture where data is processed in specific, often local data centers, with full control over its replication and access – says Jakub Stadnik. Companies are increasingly limiting the use of public language models in favor of closed or locally deployed solutions, which allow for control over information flow and data usage in the model training process, while supporting regulatory compliance and minimizing operational risk.

Information about the study and the main conclusions can be read in the PARP publication.