Andrzej Horawa, general director of AWS in Poland: We are 20 years old and we still feel like a startup

Wprost: It’s been 20 years of the AWS cloud. It’s a long time coming, but today we are talking about new challenges – on the one hand, the great AI revolution and problems with its implementation, and on the other hand, issues of sovereignty, which are becoming more and more prominent as a result of EU policy. Can we talk about it today, or are these topics not very comfortable for you?

Sure we can. We see no problem in relating to everything that is happening in the world of technology today.

So let’s start with companies and AI implementations. Why is AI adoption relatively slow?

We must clearly state that when generative artificial intelligence models were introduced 3-4 years ago, we were all happy that we would prompt, talk and that many business processes would be able to improve. However, the fact that agent solutions have been introduced and we are giving these solutions more autonomy has undoubtedly made companies realize that they need to prepare well for this change.

Indeed, currently in Poland – according to our latest study “Unleashing the potential of AI in Poland 2026” – only 16% of enterprises implement complex AI solutions.

On the other hand, nearly half of all enterprises, as many as 48%, already use AI – chatbots and other solutions, e.g. for writing e-mails.

Today we see that more and more companies are experimenting and looking for advanced solutions where they want to give the model complex functions to perform in a fully secured way. But to achieve this effect, they must have well-written programs that must have good context, access to properly structured data, and adequate memory. So there are challenges and that’s why the adoption process is progressing, maybe not as quickly as we would imagine.

However, many companies we work with have implemented many interesting advanced AI solutions, for example for analyzing and handling insurance applications or for managing customer relationships in contact centers. There are many such examples.

The study you mentioned shows a big difference between the implementations of artificial intelligence in companies and the implementations of this advanced AI. For example, the implementation of AI in the following sectors: finance is at the level of 58%, in telecommunications – 52%, energy – 43%. In terms of the use of advanced AI, the largest share is held by financial services, telecommunications and information technologies and services. But there are fewer companies that have implemented advanced AI. 26%, 23% respectively. and 21 percent What are the causes of these differences?

Companies from the industries you mentioned operate mainly on regulated markets. And in such industries, we focus more on regulatory compliance. When we think about compliance, innovation fades into the background. This is one reason for the differences between the adoption of simple tools versus advanced ones.

The second one is definitely knowledge and skills.

AI is a new technology and it needs to be learned and understood. You need to have people who want to develop. And here we have a real problem. In conversations with our clients, it turns out that organizations need to create teams that need to be given space for change and good development programs. Lack of skills is today the biggest barrier to the development of artificial intelligence.

AWS has trained – only in Poland – over 100,000 people in AWS technology. But we will not make a developmental leap in the entire society on our own.

So what needs to happen to increase the number of AI experts in Poland?

We need to build educational programs so that resources can be drawn to companies. Precisely for creating these more complex projects using advanced AI.

We currently cooperate with almost 30 educational institutions and universities in Poland. But greater programmatic support for technological change is needed.

What would the state do?

The government should certainly raise society’s awareness in the area of ​​AI and recognize the fact that there is a real shortage of competences. For 43% of companies, the obstacle to implementing AI solutions today is the lack of appropriate competences in their teams for future solutions.

This is actually number 1 in the discussion about AI in Poland today.

You say we have a skills problem and the opportunities to use artificial intelligence are increasing. During the main speech at the May AWS Summit 2026 in Warsaw, several dozen names of your products were mentioned. Dozens of services were discussed at the stands and panels.

Today, we have the largest offer of ready-made cloud solutions and services – over 240. Each of them is a separate IT field, e.g. computing services, mass storage, network services, but also many solutions in the field of AI and the construction and operation of agents. Each service includes several hundred functionalities. For example, we offer our clients several dozen databases, and it’s not just about SQL databases.

AWS, like all of Amazon, gives you a choice. This is our approach.

First, the customer selected books on Amazon, then he could search for any product in the Amazon store, and today he can choose various services within our cloud.

Your cloud service is 20 years old today – do you feel that the number of products and opportunities you provide are the result of these years of work.

Of course, but on the other hand, today we have a change driven by AI, so we are constantly creating new services – maybe that’s why we still feel like we are a startup (laughter).

Let us now move on to the issue of sovereignty, i.e. compliance of companies’ cloud policies with the latest EU directives.

We are very well prepared for the topic. Our entire cloud offering is sovereign – in the sense that only customers decide where their data is stored, no one other than the customer has access to this data.

At the same time, due to the needs of selected organizations from strictly regulated markets, we have recently launched the European Sovereign Cloud Region – it is a separate cloud region, physically and logically separated from other AWS regions. We have already invested EUR 7.8 billion in this project.

What about legal jurisdiction issues? According to the latest EU guidelines, data should be stored within the EU.

As I mentioned, for each AWS cloud region, it is the customer who decides where their data is stored. In the case of the Sovereign Cloud Region, it is operated only by European citizens or European residents.

We also established a new parent company and three subsidiaries (GmbH) registered in Germany. Europeans sit on the management board and supervisory board.

Customers therefore retain full control over their data, they also have their own encryption keys, and the system is secured by a solution that isolates customer data at the hardware level, preventing even AWS employees from accessing the files – this is ensured by our software: AWS Nitro System.

Finally, I will ask about your LLM and your plans in this area. Amazon Bedrock is a platform that hosts your own models, but also acts as a middleware to access third-party models. You recently announced your cooperation with OpenAI, and then expanded it. You intend to continue this path of development in the field of LLMs.

Yes, Amazon gives you choices. And we also have this approach here.

We cooperate not only with OpenAI, but also with Anthropic and many other companies offering fundamental models, such as AI21 Labs, Meta and Mistral AI. All of these models are available through Amazon Bedrock, our fully managed service.

We have already been trusted by over 100,000 customers who use our AI and machine learning services – companies such as adidas, Boeing, Booking.com, JP Morgan Chase, Pfizer, Ryanair and Salesforce. Why? Because we give them three things. Firstly – the widest selection of leading models, available through one API. Secondly, a security guarantee: customer data is not used to train base models or shared with model providers. Third – simplicity: it is a fully managed service, so customers can focus on building their solutions, not on managing the infrastructure.

We also invest in tools for developers. We recently launched Kiro, a new AI-powered development environment that helps developers go from prototype to production. Kiro enjoys unprecedented interest – the demand exceeded our expectations.

As for OpenAI, it is a long-term strategic partnership and a large investment on our part. We will definitely want to develop them.