The AI Adoption Barriers That Make Businesses Hesitate to Implement AI

AI adoption barriers remain a significant challenge for organisations that handle sensitive data or operate in regulated environments. Although businesses recognise the potential of artificial intelligence, many delay implementation because they lack sufficient control over data storage, access, compliance and the underlying AI infrastructure.

This article examines why external AI solutions often fail to meet enterprise requirements and how internally deployed, customisable AI can provide greater security, flexibility and operational control.

Why Data Security Is a Major Barrier to AI Adoption

“It may seem that there are many AI solutions on the market – you just need to choose one and start using it. However, functionality is not the most important issue for businesses. Many of these solutions are built as a layer on top of other models that operate and store data outside the organisation.

This means that the organisation essentially outsources part of its data processing – often without having full control over where that data goes or how it is used. This is why many companies have deliberately delayed implementing AI; even though they understood its value – they simply lacked the ability to fully control it,” comments Renata Špukienė, head of the language technology company, Tilde IT.

This situation is also confirmed by Gediminas Radavičius, a member of the BTA Insurance board:

“We recognised the potential of AI long ago, but in practice, everything comes down to very specific questions – where the data is stored, who has access to it, and how compliance is ensured. When working with customer information, you cannot afford to use solutions you do not fully control.

For a long time, the situation was simple: we saw the value, but we couldn’t adopt it because there was no technological foundation that met our security requirements,” he says.

Risks may also arise in the future due to U.S. policy

Špukienė adds that security is only one of the AI adoption barriers, while another is dependence on external factors:

“One issue is where the data ends up, but another is that if we manage our organisation using AI solutions built on, for example, U.S. large language models, we are dependent on U.S. decisions. For instance, if the U.S. president were to decide to cut off all of Europe from the U.S. AI, all of our tools would stop working. If they are systematically implemented in business, the business ceases to operate because AI agents are no longer working in Europe.

There have been examples of this; let’s recall the case of China: When the U.S. banned the use of Google and Apple app stores on Chinese phones, users were deprived of all the essential apps, and people in Lithuania were unable to download, for example, the app of a bank operating in Lithuania. Political decisions in the U.S. affected us here in Lithuania because we were using external equipment. The exact same thing could happen with AI,” says Renata Špukienė.

How Internally Deployed AI Improves Data Control

According to Špukienė, these AI adoption barriers are no longer a dead end — organisations no longer have to choose between security and the use of artificial intelligence.

“Today’s technology makes it possible to have both. The key difference arises when the large language model on which the AI tool is based operates within the organisation – either on its own servers or in a trusted European infrastructure. In this case, the data never leaves the organisation – it remains where the organisation maintains control.”

This fundamentally changes the entire situation: Artificial intelligence ceases to be an external tool and becomes an integral part of the system, which can be integrated, just like any other IT system.

In addition, this also addresses the issue of quality. The model can be adapted to a specific business, its terminology, language, and unique characteristics. This is particularly important for smaller markets, where global models are often not sufficiently accurate,” she explains.

From adoption to creation

According to Špukienė, business is currently entering a new phase of artificial intelligence.

“Until now, most solutions have focused on usage – how to make use of existing tools. However, we are now talking about a different level – one where organisations want to have a technological foundation upon which they can build their own solutions.

This means not only more control but also increased flexibility. The organisation can decide for itself how to integrate AI, which processes to automate, and how to further develop the solution. Over time, this becomes not just a matter of efficiency, but also a competitive advantage,” she says.

According to the head of Tilde IT, it is precisely this shift that will determine how AI will be used in the coming years: “Having your own technological infrastructure today means not only security but also the opportunity to grow. Businesses that realise this sooner will be able to use AI to both optimise processes and also create new value for their customers.”

How Can Businesses Overcome AI Adoption Barriers?

Businesses can reduce AI adoption barriers by choosing solutions that provide clear control over data storage, access and deployment. Models hosted on an organisation’s own infrastructure or in trusted European cloud environments can help reduce reliance on external providers and support stricter governance requirements.

Customisation is equally important. AI systems adapted to an organisation’s terminology, internal knowledge and operational processes are more likely to deliver reliable results than generic tools implemented without sufficient integration or oversight.

Successful enterprise AI adoption therefore depends not only on the model itself, but also on secure infrastructure, defined data governance and ongoing control over how the solution is developed and used.