Source: in-cyprus.philenews.com
By Alex Rashkovan*
I am curious by nature, and being the CEO of a recruitment tech company gives me the professional excuse to indulge it. A few days ago, I opened LinkedIn Jobs, set the location to Cyprus, typed “AI” in the search bar, and hit enter.
At first, the results looked encouraging. LinkedIn returned page after page of vacancies from established companies, many of them with impressive titles.
My disappointment came when I started reading the job descriptions. The more I read, the less useful the “AI” label was in telling me about the work involved or the capabilities each employer needed from a candidate.
One recent vacancy from a professional services firm asked for experience in areas like “Generative AI,” “language models,” “enterprise applications,” and “data engineering.” Another, posted by an IT company, was for a Machine Learning Engineer role with a focus on predictive models and automation. An MLOps Engineer vacancy listed responsibilities including work on “deployment pipelines,” “cloud infrastructure,” and “model monitoring.”
On their own, the vacancies are admittedly pretty clear. However, when read together, they show just how broad and generic the category “AI professional” has become. Someone suited to one of these positions does not automatically mean they are suited to another.
This is the hiring problem Cyprus risks encountering as it begins implementing its AI strategy.
The draft National AI Strategy 2032 recommends a target of more than 3,000 “AI professionals” in the market by 2032. But who does that figure actually include?
The target gives us a sense of the scale of the workforce Cyprus is hoping to build. Cyprus’s success in reaching it will depend heavily on whether employers can break down the umbrella term “AI talent” into the constituent capabilities each position requires, and then recognise those capabilities in the candidates applying for their vacancies.
AI
The Cyprus Statistical Service’s 2025 ICT survey helps to show just how different those capabilities can be. Among Cypriot enterprises using AI, over 70% purchased ready-to-use commercial technology, while just under 30% reported that their own employees had developed company AI systems. In the survey, businesses were allowed to report more than one route, so we can’t conclude that these are opposing groups, one or the other.
But it helps with our argument. So, an organisation implementing a commercial system may need someone who can integrate it with the existing infrastructure, prepare the data, and then oversee its use, especially when it comes to a regulated environment. But a company developing models internally will usually require employees with more in-depth knowledge and experience in machine learning, software engineering, and model evaluation.
Both approaches require “AI talent,” although the capabilities required of each employee are quite different. The trouble begins, therefore, when an employer (or recruitment team) starts with the label “AI” and puts together the on-trend keyword technologies in the job description without fully understanding or deciding what the person in the role will actually be expected to deliver.
The same problem affects candidate screening. A search based exclusively on candidates’ previous job titles that match the job offering may miss a software engineer who has deployed language-model applications, a data engineer who has built machine-learning pipelines, or even a researcher who has implemented a commercial AI system. Their experience may be relevant even if their last title did not contain the keyword “AI.”
I do not have the stats to show how many suitable candidates are lost through title and keyword screening (no Cyprus national data is available for this). However, Eurostat reports that more than half of Cypriot enterprises with at least ten employees that recruited (or tried to recruit) ICT specialists in 2023 had difficulty filling those vacancies. Cedefop’s 2026 Cyprus jobs and skills spotlight states that one in three highly skilled Cypriot workers aged 25 to 34 was overqualified for their job in 2023, describing this as evidence for mismatched or underused potential.
Now, neither of these figures measures AI recruitment per se, and neither proves that it was the hiring practices that created the mismatch. However, they do highlight that hard-to-fill vacancies and underused skills can coexist in the same labour market. Before Cyprus concludes that it lacks AI talent, employers should first ask whether they are defining and recognising the talent (and the capabilities for the specific roles) with enough clarity.
Start with the work
The process begins before hitting publish on a new job vacancy.
Before advertising for an AI role, an employer (or recruitment team) should answer two questions. What must this person be able to do when they join? And which capabilities could reasonably be developed after their appointment?
Once there is a clear answer to both questions, the employer can then decide which parts of a CV are relevant and what needs to be assessed on an individual basis.
For instance, specific degrees and professional credentials may help establish whether a candidate has foundational knowledge or satisfy regulatory requirements. Previous roles and completed projects could provide evidence of experience of similar work, even when a candidate’s job title in previous roles is different from the one listed in the vacancy. Next are the work samples and technical exercises that match the actual job to show whether a candidate can, in fact, perform these relevant tasks. That is not to say a previous job title is not useful context. It should not be treated, however, as a filter or eligibility test.
This approach requires more thought before a vacancy is published. Employers have the responsibility to separate essential capabilities from preferences and decide what they are prepared to teach on the job. That effort gives recruiters a more focused brief and candidates a fairer opportunity to demonstrate whether they can do the work.
The same principle should be applied to the technology used to screen applicants. Under the EU AI Act, certain systems used to filter applications or evaluate candidates are classified as “high-risk” because they can affect a person’s access to employment, career prospects, and, ultimately, their livelihood. The classification does not stop organisations from using these systems, but they must do so under stricter requirements, especially regarding risk management and transparency. Those requirements are due to apply from December 2027.
But employers should not wait until December to understand what their screening tool is assessing and which information affects its recommendations, and where human oversight enters the process. If they struggle to answer those questions, the tool should not be used to determine which candidate proceeds or not.
Measure where the skills go
The proposed National AI Skills Observatory could help employers by tracking which capabilities are in demand and which roles remain difficult to fill. Employers should then be able to use that information when defining their roles and drafting job descriptions.
The reports need to move away from the broad category “AI professional.” Instead, they should set out the actual capabilities as connected to priority roles, and then show how the requirements for the roles are evolving. When it comes to training, providers should state which capabilities are developed during each course and the kinds of tasks or problems those capabilities can be applied to.
However, it doesn’t stop there. Course completions and credentials show who took part in the training. The Observatory should also track whether the participants then entered work that uses what they have learned, how long vacancies requiring these priority capabilities remain open, and in which areas underemployment continues.
These measures could help policymakers understand why some vacancies looking for specific capabilities remain open. Some of those capabilities may be scarce in the population, suggesting a need for more training opportunities or potentially outsourcing to international recruitment. In other cases, suitable candidates may already be available but overlooked because their job titles, career paths, or current credentials do not match what an employer (or system) expects.
The target of 3,000 professionals gives Cyprus a number to work towards. Whether it is ambitious or not, only time will tell. What can be done now is work out the details. One is ensuring the Observatory tracks whether those professionals find work that uses their capabilities and whether employers are filling their vacancies. Otherwise, Cyprus could still meet the target while qualified candidates are left underemployed and unemployed and businesses continue to report shortages.
Of course, a national strategy cannot manage every recruitment decision. What it can do is show whether investments in training and talent attraction are leading to suitable employment, and whether qualified and capable candidates are being missed.
Cyprus needs 3,000 AI professionals. Whether it recognises them will be decided one vacancy at a time.
*Alex Rashkovan is the co-founder and CEO of Atalef, a Cyprus-based recruitment technology company focused on skills-first hiring for technical roles.