AI is quietly entering the office
How is AI being integrated into the work of Kosovo's public?
By Gentiana Paçarizi | May 28, 2025
In late 2022, the company OpenAI launched the ChatGPT platform for mass testing. Within the first five days, about a million people around the world had used it. This number grew rapidly, and by January 2023, ChatGPT had reached 100 million active users. That same year, I conducted my first experiment in working with artificial intelligence (AI). I co-wrote an article with ChatGPT, asking the platform to explain how it works, what its origins are and what impact this technology could have on work processes, especially on the integrity of information.
At the time, when I was researching with the help of AI, some of ChatGPT’s limitations included the lack of updated information beyond 2021. When I asked it to read some of my recent writing in order to adapt to my writing style, it failed to access web pages or documents. And yes, it would interrupt me every hour to inform me that I had asked too many questions within that timeframe. At times, it wouldn’t work at all due to high demand. I couldn’t write the article in Albanian either, as ChatGPT’s primary language was English and the information it provided in other languages wasn’t at the same level of accuracy. When it came to reliability, ChatGPT had little credibility as a source of information.
Today, two years later, these limitations are almost non-existent. Although content generated by ChatGPT still leaves subtle traces, they are far less noticeable than they were two years ago.
While ChatGPT became the symbol of AI in the eyes of many users, a number of other platforms with similar features also emerged as alternatives. Although none achieved the same reach as ChatGPT, these generative AI (GenAI) tools — a specific branch of AI focused on creating new content based on existing data.It made routine tasks, such as writing emails, almost fully automated.
One of the memes circulating on social media about the use of AI shows Sisyphus sitting in a chair reading, while a robot carries the stone uphill. In Greek mythology, Sisyphusis condemned by the gods to spend eternity rolling a stone to the top of a mountain, only for it to roll back down each time — has often been used as a metaphor for modern work systems. This is especially true in corporations, bureaucratic jobs or industries where workers perform repetitive tasks day after day, tasks which, through sheer repetition, often become exhausting and frustrating.
But beyond automating repetitive tasks, these tools are also gaining ground in the creative sectors. Generative AI is now being used to produce near-realistic videos and images, to generate creative advertising concepts in minutes, and to realize them just as quickly. GenAI’s ability to create within fields that traditionally rely on human creativity also raises concerns about its potential impact on people working in the arts and culture sectors.
For example, as media outlets increasingly use AI-generated images instead of professional photography, it is becoming more difficult for photographers to make a living from their craft alone. In March 2023, artist Boris Eldagsen won first prize in the open competition’s creative category at the Sony World Photography Awards. However, Eldagsen declined the award after revealing that the image had been created using the GenAI image generator DALL·E 2, and he did not consider it his own artistic work. By submitting the image, Eldagseb intended to spark a discussion about the role of AI in art and photography.
Today, however, with around 123 million people worldwide using ChatGPT daily, GenAI models are no longer just tools for experimental use in the workplace. They have quietly, and sometimes enthusiastically, entered offices across the globe, becoming an inseparable part of everyday life.
Moreover, GenAI’s capacity to perform increasingly complex tasks has grown with each new model, managing to complete in minutes what would take humans hours. In some areas, GenAI’s performance has even matched or surpassed that of humans.
An analysis by the METR organization, published in the scientific journal Nature in March 2025, highlights AI’s rapid progress toward handling more complex and time-consuming tasks, where GenAI still lags behind human performance.
The analysis uses a logarithmic scale to measure the time it takes humans to complete a task that GenAI can perform with 50% accuracy — a way to clearly understand how capable the models are compared to human performance, especially on longer tasks.
In 2019, models like GPT-2 could only complete basic tasks, like answering simple questions, and couldn’t perform tasks that would take humans more than a minute to complete. But since then, the models have advanced at a rapid pace.
GPT-4, and especially more advanced models like Anthropic’s Claude 3.7 Sonnet, along with similar platforms such as ChatGPT, Google’s Gemini and xAI’s Grok, have pushed the boundaries even further. This new generation of AI has managed to accurately complete tasks that would take humans up to 59 minutes — such as building new website features or analyzing complex data. According to METR, GenAI models will be able to perform tasks that would normally require a month of human effort, with a 50% confidence level, by 2029 — or perhaps even sooner.
Meanwhile, according to data from Stanford Human-Centered AI, a research institute founded by Stanford University and included in an April 2025 report, GenAI’s performance on a number of tasks has already reached or exceeded human levels. In areas such as image classification, English language comprehension and visual interpretation, GenAI has surpassed the human benchmark by 100%. For more complex tasks, such as competitive-level mathematics or PhD-level scientific questions, GenAI is also approaching expert standards.
The development of AI will have a fundamental impact on the way we approach the concept of work, and the first effects are already beginning to surface.
How do Kosovo citizens use AI at work?
To understand how citizens in Kosovo use GenAI for work, we at K2.0 conducted a survey in April 2025. Over the two weeks the survey was open, we received responses from 204 users.
More than 50% of respondents were in the 35-54 age group, a demographic that often holds leadership or managerial positions and can use GenAI to accelerate complex tasks such as project reporting and data analysis, explaining their high representation in the survey.
Age of respondents
Meanwhile, about 21.1% of respondents were over the age of 55. The high representation of this age group, although generally less inclined to adopt new technologies compared to younger groups, can be explained by the fact that they often occupy senior positions or specialized roles in sectors such as education, finance or civil society, where GenAI can be useful for tasks like reporting, data analysis or translation. Additionally, the availability of free versions of GenAI platforms, such as ChatGPT, may have lowered the barriers to adoption for this age group.
The least represented age group was those under 18 and those aged 18–24. This lower representation may be due to their engagement in the retail or hospitality sectors — where GenAI has not yet been widely integrated.
Which industry do you work in?
Since the survey targeted users who use AI for work, the data shows that AI has spread across various industries.
Almost half of respondents — 48% — use GenAI platforms for work several times a day, ranging from three to 10 times. However, a significant portion — 34.3% — reported using it only once or twice a week.
How many times a week do you use these AI tools for work?
The vast majority of respondents — 95.6% — use ChatGPT, while 8.3% use Gemini and 6.9% use Grok, along with a range of other applications (respondents were allowed to choose more than one option). ChatGPT’s high usage can be attributed to its easy integration and its availability in free versions. The same trend holds true globally: data shows that by February 2025, ChatGPT had nearly 400 million weekly active users and about 122.58 million daily users.
Which AI platform do you use most often for work?
Respondents indicated that they use GenAI most often for research and content creation, such as texts, articles, descriptions and social media posts.
Using ChatGPT for research has become more useful since the platform has integrated referencing through links to information sources. For many users, conducting centralized research on these platforms is easier than searching through Google.
However, the accuracy of information generated by ChatGPT and other GenAI models remains questionable. ChatGPT’s journey in building credibility is particularly illustrative of this challenge.
What do you usually use AI for in your work?
There have been concerns about ChatGPT’s fabrication of sources and information. In a 2023, The Guardian reported receiving frequent requests from readers for articles that ChatGPT had attributed to the outlet, but which in fact did not exist. Despite its high accuracy across many tasks, ChatGPT can produce what are known as “hallucinations.” This term, borrowed from psychology — where hallucinations refer to false perceptions of reality — is used in the context of GenAI to illustrate the nature of such errors.
These hallucinations occur because GenAI models are built on large-scale data systems and generate responses based on statistical predictions, rather than direct fact-checking. If the data is limited or the prompt is unclear, GenAI may fill in the gaps with fabricated information.
For example, a study published in the Journal of Medical Internet Research in 2024 analyzed the accuracy of citations generated by GPT-3.5, GPT-4 and Bard (PaLM version 2.0, now Gemini) for 33 queries based on 11 systematic reviews in four medical fields. GPT-4 had a citation error rate of 28.6% (34 out of 119 citations), meaning that about a third of the citations generated were incorrect, with at least two elements — title, first author or year of publication — being wrong. In comparison, GPT-3.5 and Bard had higher citation error rates of 39.6% and 91.4%, respectively.
AI Overview, launched by Google in 2024, has also faced similar problems. AI Overview uses AI to generate concise, immediate answers to users’ queries on Google Search. In some cases, it suggested that users looking for ways to make cheese stick better to pizza could use “non-toxic glue.” It also claimed that geologists recommend eating a rock a day. Some of these answers appeared to be based on comments from Reddit or articles from the satirical website The Onion, highlighting that, while these may be isolated cases, the use of GenAI platforms and models for search requires caution.
How has the use of AI affected the way you work?
In the K2.0 questionnaire, many respondents mentioned using GenAI for spelling and translation into other languages, compiling lesson plans (particularly among those in the education sector), writing reports and coding. Others noted using GenAI to improve overall work processes.
According to respondents, the four main benefits of using GenAI at work are: it saves time, enhances creativity, improves organization and boosts productivity.
A marketing respondent said:
“A concrete example of when AI significantly helped me in my work was during the analysis of a collaboration process on the Monday.com platform. I was working to better understand the flow of tasks, team roles and opportunities to improve efficiency in the project. To speed up the analysis, I used ChatGPT to structure my thoughts, identify critical questions to ask and suggest possible ways to optimize the process. This helped me save time and arrive at a clear, well-reasoned solution more quickly. I also wrote this answer you’re reading with the help of ChatGPT.”
Although GenAI is widely integrated into the work process in Kosovo, only 36.3% of respondents said that their workplace encourages and supports its use. 31.4% said there is no official support or clear policy, while 10.8% reported that their workplace has restrictions or policies against the use of GenAI. This indicates that GenAI is largely being adopted on an individual basis rather than systematically by workplaces.
Is your workplace friendly and open to using AI?
However, employers will soon need to strategically invest in workplaces that take into account the changing ways workers approach their work today.
New standard of work?
Predictions that GenAI could, within a few years, perform tasks that take humans a month, suggest that it may set a new standard for work — one that will require more adaptive skills rather than deep expertise in a specific field. This could lead to many individuals taking on roles for which they do not necessarily have formal qualifications. For example, a journalist could write code without being a programmer, and a project manager could create beautifully designed visual content without having any knowledge of graphic design. However, this also means that workers who do not have access to these tools or skills will need to quickly adapt to the demands of the new standard GenAI may introduce.
AI could support exactly this kind of transition according to the results of an experiment published in 2024 by the BCG Henderson Institute, in collaboration with researchers from Boston University and the OpenAI Economic Impact team, The experiment showed that employees who previously lacked the knowledge to perform technical tasks, such as statistical analysis or coding, were able to accomplish them with the help of GenAI.
The study involved 480 BCG general consultants and 44 data scientists to assess how GenAI affects the ability of non-technical employees to complete tasks that require technical skills. Participants were divided into two groups: one used ChatGPT-4 with advanced data analysis features, while the other used traditional resources such as Stack Overflow. Both groups were assigned complex tasks, such as coding and statistical modeling, that GenAI alone could not solve. The final analysis compared accuracy, time and performance between the two groups and against standards set by BCG experts.
The experiment showed that GenAI significantly expands people’s capacity to complete tasks outside their field. Consultants with no experience in coding or statistics, with the help of GenAI, were able to write code, apply machine learning models and correct statistical errors.
For example, in a coding task, participants with GenAI assistance achieved 86% of the performance level of data scientists — a 49 percentage point improvement over those who did not use GenAI. Even participants who had never written code before reached 84% of the expert standard, while the group without GenAI achieved only 29%. The GenAI-assisted group also completed the tasks about 10% faster than the data scientists. Meanwhile, those who worked without GenAI often progressed no further than opening folders and performing initial data cleaning; they achieved only 29% of the standard set by the data scientists.
One of the respondents to the K2.0 questionnaire, who works in civil society, supported these findings:
“As someone who has no technical coding skills or deep knowledge of arithmetic, AI has helped me, instead of writing code, formulate prompts that translate into programming formulas, allowing me to perform data analysis or design in Python.”
However, GenAI only seems to improve user performance when the model itself has expertise in the relevant field. In a previous experiment in 2023 by the BCG Henderson Institute, in collaboration with researchers from Boston University and the OpenAI Economic Impact team, it was found that workers supported by GenAI performed significantly better on tasks within their area of knowledge — as long as the technology was also competent in those areas. But when GenAI lacked expertise in the relevant domain, users often relied too heavily on it and performed worse than if they had worked independently.
Do you use AI for personal purposes?
Meanwhile, the use of ChatGPT and other GenAI platforms goes beyond the workplace. Their integration has become a kind of personal assistant, capable of providing answers to almost anything — as illustrated by the response of one of the respondents, who works in civil society:
Of the 204 respondents, 82.4% said they also use it for personal purposes outside of work. The majority, 53%, said they use it for learning, followed by 47% who use it for personal advice (such as decisions, relationships or mental health), and 34.8% for food-related purposes.
Can AI replace human work?
Technological revolutions have always brought with them the automation of certain jobs previously done by humans. With the introduction of machinery during the Industrial Revolution, many tasks once performed by hand were transferred to machines. The cost of production dropped significantly, pushing workers into roles that could not be automated — such as technical maintenance, creative work or management. In the 20th century, computerization gradually replaced manual office work such as form-filling and financial calculations. With the advent of the internet, many professions, including postal workers, archivists and telephone operators, have been transformed.
Similarly, with the rapid acceleration of AI development, several possible scenarios for the labor market have been predicted. The most pessimistic scenarios envision a world in which, within a decade, almost every creative job — such as writing, analysis, graphic design, basic programming and legal support — could be rendered obsolete by full automation. Meanwhile, the most optimistic scenarios suggest that AI will serve as an auxiliary tool, increasing productivity and allowing people to focus on more complex, creative and strategic tasks.
Some critics express concern about how AI has taken over jobs that, until now, were considered fundamentally human, such as writing and painting. While social media often simplifies complex discussions, it occasionally circulates content that captures deep human dilemmas. For example, a quote attributed to the author Joanna Maciejewska, which circulated widely in recent months, reads: “I want AI to wash my clothes and dishes so that I can do art and writing — not for AI to do my art and writing so that I can wash my clothes and dishes.”
AI will also impact economic inequality — according to an analysis by the International Monetary Fund in 2024, AI is expected to affect nearly 40% of jobs globally. The impact of AI varies significantly across economies. Developed economies, such as the U.S. and the U.K., face the highest exposure, with around 60% of jobs potentially affected by AI’s ability to automate high-skill tasks. Roughly half of these jobs could benefit from increased productivity, favoring higher-wage workers, while the other half could see reduced demand for labor, leading to lower wages or job losses, thereby widening income inequality.
In contrast, developing economies face lower exposure, at 40%, and low-income economies at 26%, due to less developed digital infrastructure and fewer workers capable of using AI.
Meanwhile, by 2030, according to estimates by the McKinsey Global Institute published in 2023, 30% of working hours in the U.S. will be automated, resulting in approximately 12 million professional transitions. Workers may need to move into new roles or industries — to be displaced — due to the automation of tasks such as data processing and customer service. AI will change the nature of tasks performed by knowledge workers, increasing productivity without fully replacing them. However, low-wage workers in the U.S. (earning $30,800–$38,200 per year), according to the study, are 14 times more likely to change professions compared to higher-income workers.
While many jobs across generations could be automated, the future remains especially uncertain for younger generations. AI is advancing so rapidly that people at the beginning of their careers could soon face an almost entirely automated job market. Anash Raman, a LinkedIn executive, wrote in an opinion piece for The New York Times that there are growing signs AI is replacing entry-level jobs that have historically served as a stepping stone for younger generations entering the workforce.
The positions most at risk are those at lower levels: in the tech sector, AI is replacing basic programming and software testing; in law firms, it is reviewing documents faster than paralegals; and in retail, chatbots and automated systems are taking over tasks like customer service.
While almost all fields will feel the impact, office jobs are expected to be hit the hardest — especially those that require advanced degrees but do not involve specific technical or creative skills. Although AI is expected to create many new jobs in the future, according to Raman, the lack of entry-level roles could slow down the career development of young people and deepen social and economic inequalities.
Despite the ongoing debate sparked by AI’s rapid development and predictions that technology will remain the most defining force in the job market until 2030, few respondents to the K2.0 survey expressed concern that AI will replace them in their current jobs.
Has the use of AI made you think that your job could be replaced by technology?
However, respondents still remain concerned about some of the effects of using GenAI platforms. They mention addiction, passivity, reduced creativity, plagiarism and unoriginal work, as well as the loss of personal skills if GenAI platforms are used excessively. Some express specific concerns about information privacy and the protection of personal data. Respondents also mentioned problems relating to overreliance, errors and the low quality of automatically generated content.
In February 2023, when I first used GenAI in my writing, I also had to deal with generic and often error-prone responses . But despite this, GenAI tools, especially ChatGPT, and more recently Grok, have become an integral part of my work. It took some time to train ChatGPT by asking it to remember how I prefer it to respond (no emojis, please), and by feeding it context about my work and the way I think.
AI often resembles a workaholic to me — unfettered by the 9-to-5 schedule and adapted to my unscheduled work pace and way of thinking in parallel across many tasks. Unlike in the beginning, I no longer use these tools to write first drafts, but for dozens of other tasks: from drafting concepts, reports and emails, to summarizing studies. I often even ask it for a first opinion on my drafts. Overall, I’ve found that GenAI has helped me complete less enjoyable and time consuming tasks more quickly, freeing up space for me to focus on work that truly fulfills me.
As AI develops at a rapid pace, and many of the skills I’ve built over the years begin to lose relevance, I find myself thinking about the need to reconsider my professional plans in the near future. It seems that few people expected such a sudden interruption in the course of their life plans — but in the end, the history of technological revolutions shows that humans are remarkably capable of adapting to new roles. And this, rather than worrying me, makes me deeply excited about the future to come.
Feature Image: Dina Hajrullahu / K2.0.
Want to support our journalism?
At Kosovo 2.0, we strive to be a pillar of independent, high-quality journalism in an era where it’s increasingly challenging to maintain such standards and fearlessly pursue truth and accountability. To ensure our continued independence, we are introducing HIVE, our new membership model that offers an opportunity for anyone who values our journalism to contribute and become part of our mission.
Become a member of HIVE or consider making a donation.