30
.
01
.
2025
Source: Rasa Gulbe, Member of the Board, DATI Group SIA
In our previous article, we looked at the four main steps to implementing artificial intelligence (AI) in the enterprise. However, to be successful, you also need to be aware of the major obstacles that can hinder progress. Once these challenges are clearly articulated and understood, it becomes easier to find effective solutions. As a result, AI integration no longer seems an unattainable goal. At present, four major challenges can be identified as the most common obstacles to progress.
The potential of AI remains untapped if it is only introduced in one isolated project. To ensure maximum benefit, it is important to develop a strategy before implementing AI that allows it to serve the whole organisation. For example, ChatGPT is implemented to generate marketing content, but other non-marketing staff have no idea how else the tool could be used.
Large companies mostly have digital transformation strategies, where the MI strategy is simply a section of the overall company strategy. In SMEs, the MI strategy should be addressed separately. Sometimes the management of the company is not sufficiently informed about the possibilities of technological solutions, so the implementation of AI is not a high priority. This can be changed by demonstrating even a small but pilot project, by sharing success stories between departments and even between companies, and by simulating different scenarios.
It is often assumed that AI solutions can be implemented without training, but in reality, they require at least a basic understanding of analytics, statistics and mathematics to be used successfully. This knowledge is important not only to implement the technologies, but also to be able to critically evaluate their responses and recommendations.
Similar challenges can also arise at the management level. To implement AI effectively, it is essential to clearly communicate the objectives of the change, the added value and its long-term relevance. SMEs often lack a specific person or team to set the strategic direction, develop a tactical plan and clearly define the next steps in innovation.
Three approaches can be used to overcome these barriers. Developing dynamic capabilities - it is important for a company to develop the ability to identify changes quickly and integrate them into its daily work. This means adapting dynamically. Such skills can only be acquired through practice. Implement small-scale pilot projects and experiments where employees can gain practical experience. The most important thing here is to understand the roadmap and that this is only the beginning of more complex solutions. Training and mentoring, with support from industry partners, suppliers and other experts to facilitate knowledge sharing and strengthen the skills needed.
SMEs most often cite lack of financial resources and time to implement new solutions as a barrier to implementing AI. However, it should be said that first tests and prototypes can be carried out with simple and now inexpensive solutions, such as Microsoft's built-in Copilot. In reality, therefore, most SMEs lack strategic vision and priorities.
If, however, lack of finance is a barrier, then one of the support programmes for digitisation and digital transformation available in Latvia can be used.
Most companies that have implemented AI today have actually implemented supervised machine learning. The biggest barrier is the lack of appropriately labelled data to use in machine learning. Supervised machine learning means that the algorithm has to be given pre-labelled data sets that are used to train it. If this is not available, then supervised learning is generally not possible.
The alternative is unsupervised learning, which is used very little. If we give unlabelled data and the algorithms themselves define the groups or pattens, it becomes quite difficult to explain and justify the decisions taken or recommended at that point. In such cases, it is usually decided not to use MI because the risks are too high and it would be unethical to use AI. The choice of each algorithm and model must be able to be explained.
For SMEs, a good data management strategy starts with knowing what the data sources are, where they are, who is responsible for them and who manages them. Then, with those responsible, we can start to analyse which data is usable for training algorithms, which data has duplicates in other databases, etc. If the data is scattered, e.g. some in a private cloud, some in a public cloud, and not connected to each other, then the data sources need to be integrated. There is no other way around it. SMEs most often outsource this work.
_________________________________________________________________
Learn More
For an overview of AI implementation stages, read our article:
"4 Stages of Implementing AI in a Company."
Ways DATI Group Can Assist Your Project
We offer secure system integration for implementing data management solutions, including:
.svg){kind=icon}
