More than 71% of companies are utilizing generative artificial intelligence (AI) in at least one business function, according to the latest global research conducted by the consulting firm McKinsey. AI is most commonly applied in marketing and sales, product and service development, customer support, and software engineering, where it can deliver the greatest potential value. Although domestic companies express a desire not to fall behind in the AI revolution, experience shows that many abandon the process halfway—often due to lack of knowledge, lengthy implementation, and the delayed arrival of tangible results. To avoid this, experts suggest that the easiest way to enter the world of artificial intelligence is “through small doors,” by introducing AI at the most basic level and integrating smart tools into smaller sectors.

This, of course, depends on the size of the company, the industry in which it operates, and the need for applying smart technology. While larger organizations and those in technology-intensive sectors adopt AI more quickly and comprehensively, smaller firms encounter numerous challenges from the outset—such as employee readiness, the availability of skilled staff, and the company’s strategic goals—which is why AI use often remains limited to pilot projects or tools for handling specific tasks.

One of the biggest concerns for all companies is the security of data shared with artificial intelligence. As generative models become more sophisticated, risks are also increasing. Many real-world cases show that adversarial attacks are no longer just hypothetical but occur daily, even targeting the largest companies, which is why this area requires serious oversight.

Still, while some companies wait for the right moment, others are already reaping the benefits. The adoption of artificial intelligence across industries is growing by up to 20% annually. From 2023 to 2024 alone, the use of generative AI jumped from 55% to 75%. According to Maks Belov, technical director at IT company Coherent Solutions, artificial intelligence does not need to be revolutionary—it first needs to be practical. Too often, companies rush to adopt the latest shiny tools as part of their transformation plans, only to realize they have overspent money, time, and energy on systems and processes without clear goals or execution paths.

Nebojša Matić, owner of Mikroelektronika, a company that develops hardware and software solutions and has a department dedicated to artificial intelligence, told NIN that the easiest way to apply AI is by using ChatGPT. He emphasizes that this application should not be seen as a tool to finish the job for you, but rather as one that facilitates your work and helps complete tasks more quickly.

The ChatGPT mobile application has generated two billion dollars in revenue since its launch in May 2023, with an average revenue per installation of $2.91, according to the latest report by Epfigurs. In the first seven months of 2025 alone, ChatGPT earned $1.35 billion from installations—an increase of 673% compared to the same period last year. Its closest rival, “Grok,” earned only $25.6 million from installations during the same period, meaning ChatGPT’s revenue was 53 times greater.

Matić notes that ChatGPT primarily takes over routine clerical tasks, such as drafting a professional email, preparing contracts between parties for rental agreements, or even comparing data. Still, he stresses that its outputs should not be taken at face value, as the technology is not yet fully mature, and users must ultimately rely on their own expertise to ensure accuracy.

This finding is reinforced by McKinsey, whose research shows that, despite the enthusiasm surrounding the potential of artificial intelligence and the impressive capabilities of smart tools, the first generation of large language models faced limitations that significantly restricted their use at the enterprise level. Large language models (LLMs) can produce inaccurate outputs, making them difficult to trust in environments where accuracy and consistency are critical. They are inherently passive—unable to act independently or initiate workflows and decisions without human input. Moreover, they struggle to manage complex processes that require multiple steps, decision points, and logical reasoning.

AI poses the greatest challenge for small and medium enterprises

According to a domestic survey published earlier this year by ICT Hub, conducted in collaboration with the Serbian Association of Managers (SAM) and Reprezent Communications, only 34% of companies in Serbia utilize artificial intelligence (AI) in their operations, while the majority have yet to integrate this technology into their processes. The highest application of AI technologies has been recorded in the information, communication, and professional services sectors, where 60% of firms employ artificial intelligence.

Market observers note that domestic companies with 200 to 500 employees, which have managed to cope with digitization to a certain extent, feel the need to catch the “AI wave”, especially out of fear that competition will overtake them. However, the biggest issues include inadequate knowledge of the topic, underdeveloped infrastructure, and, most importantly, a lack of understanding of how AI can aid in their business processes.

Filip Karaičić, CEO of the domestic company “Kvantoks Technology,” states that in practice, the initial process of implementing AI is usually quite “painful,” lengthy, and often ends unsuccessfully. He indicates that at the very start, a significant amount of initial consulting activities is required to understand what the company does, its processes, where the most time is wasted, where the most errors occur, and where there is the most room for improvement. Only after this is determined can a strategy for AI implementation be considered.

However, a frequent obstacle in the next step is disorganized data, which is often quite fragmented and scattered or duplicated across various services, servers, and vendors. This necessitates additional work and time to clean up, consolidate, and adapt everything for use in AI solutions. When this is all “put on paper,” it becomes clear why the entire process appears as a massive expense—both financial and temporal—for the firm, and it can often be an even greater problem for employees.

According to a domestic survey published earlier this year by ICT Hub, in collaboration with the Serbian Association of Managers (SAM) and Reprezent Communications, only 34% of companies in Serbia currently use artificial intelligence (AI) in their operations, while the majority have yet to integrate this technology into their workflows. The highest levels of adoption have been recorded in the information, communication, and professional services sectors, where 60% of firms employ AI.

Market observers note that Serbian companies with 200 to 500 employees—those that have managed to advance digitalization to a certain degree—now feel the need to ride the “AI wave,” driven largely by fear of being overtaken by the competition. Still, their main obstacles remain limited knowledge of the field, underdeveloped infrastructure, and, most importantly, a lack of clarity about how AI can actually support their business processes.

AI is easiest to introduce in repetitive jobs

Despite the fact that the largest investments in artificial intelligence come from the most valuable companies in the world, business transformation influenced by AI is no longer reserved solely for global giants and the technology sector. Companies of all sizes and industries are finding ways to leverage AI for business enhancement, the introduction of innovations, and understanding the needs of their customers.

Prof. Dr. Nebojša Bačanin Džakula, Deputy Dean for Teaching and Development in Artificial Intelligence at the University of “Singidunum,” states that depending on the nature of the company’s business, the easiest way to enter the world of AI is to start with the application of advanced chatbots and the introduction of virtual assistants. These tools are now taking first contact with customers in numerous companies, whether in banking, telecommunications, or e-commerce, all aimed at providing better customer service.

The same is true for applications in call centers, where advanced deep learning models can automatically recognize the user’s intent and classify their request. They help to understand why the user is calling, and can even resolve certain issues without involving an operator. This significantly reduces pressure on employees, and customers receive quicker responses.

“AI can also be easily utilized in translation and proofreading tasks, as AI-based tools provide excellent real-time translations that are contextually appropriate. Text proofreading programs can assist in correcting grammatical, stylistic, and other errors, offering significant support in journalism, education, and administration in general. AI can be smoothly integrated into administrative tasks, such as sorting work documents, and can be applied in calculations and task assignments. Additionally, its support in marketing is already evident, where AI easily creates personalized messages and develops specific content,” explains the professor.