A senior Greek researcher has outlined how artificial intelligence could reshape industrial production in the near future, with pilot applications expected as early as this summer.
Dr Dimitris Tzovaras, Research Director at the Information Technologies Institute of the Centre for Research and Technology Hellas (CERTH) in Thessaloniki, described a shift towards highly flexible manufacturing systems.
Speaking to Voria.gr's AI-focused vidcast, Homo sAIence, he said future factories "will have such great flexibility that they will be able to readjust the production line either for mass production or for highly specialised output", including the possibility of producing a vehicle tailored to a single customer.
A key component of this transition is the use of "digital twins", virtual replicas of factories that enable companies to simulate operations. According to Tzovaras, "you will be able to 'run' an entire industrial unit in a digital twin, through a simulation, so that you can test things and identify problems in its operation or diagnose maintenance needs very early, without even needing to be on site". He linked these capabilities to advances in AI, particularly "agentic AI", and to the integration of robotics across production processes.
AI agents are expected to play an increasingly important role in decision-making within factories. However, Tzovaras stressed that human oversight will remain essential. "In important decisions, no one will leave an AI system on its own," he said, adding that systems could handle less critical operational decisions while humans focus on "issues of ethics, creativity and deontology". He indicated that a pilot factory "managed to a significant extent by AI agents" is expected to operate in Greece this summer.
On industry readiness, Tzovaras said that manufacturing is no longer lagging technological developments, noting that digital transformation, including the digitisation of factory data, is already underway. This, he said, is the first stage and the basic prerequisite for applying AI at scale, enabling predictive maintenance, improved automation and AI-supported decision-making.
Dr Tzovaras said that Greece and its businesses have strong potential to utilise small language models (SLMs) rather than large, general-purpose models such as ChatGPT, Gemini, or Claude. He explained that SLMs are more targeted and specialised, designed for specific applications rather than broad use, creating opportunities for Greek companies, universities, and research centres to contribute. These models are trained using data provided by the company itself, meaning "data never leaves the company", which reduces the risk of industrial espionage and ensures ownership of the outcome. He added that if such models can be adapted through further training to suit other companies, "the field immediately becomes broader".
Dr Tzovaras identified three key challenges to the effective adoption of AI in Greece. First, he said the public sector appears more advanced, while the private sector lags due to risk aversion and reliance on state funding. Second, he highlighted a lack of education and specialised talent, noting that although companies may be able to develop AI products, their staff often cannot support or update them, especially as "almost everything changes every six months". Third, he pointed to a focus on short-term returns, arguing that private businesses need to make larger, forward-looking investments.
The discussion also touched on opportunities linked to infrastructure, including the "Daedalus" supercomputer, described as "the best thing that has happened in the country in recent years", and the "Pharos" AI factory, as well as broader questions about robotics and the potential timeline for artificial general intelligence.
