More than nine in 10 governments worldwide have deployed artificial intelligence in public services, but fewer than 30% measure whether it actually works, experts told a digital governance conference in Azerbaijan on Friday.
The SOCGOV 2026 forum, held in Khankendi in Azerbaijan’s Karabakh region, brought together government officials and policy specialists to examine whether digitalisation is improving outcomes for citizens or simply replicating existing processes online.
Nijat Hajikhanov, a data management specialist at Luxembourg’s National Data Service, cited OECD data showing that while more than 90% of countries had adopted AI in public services, only 28% measured its real-world impact. He called this an “AI adoption gap”.
“It is not about jumping into what AI tool, infrastructure, or technology to adopt,” Hajikhanov said. “It is more about what problem we are going to solve.”
He urged governments to follow four steps before deploying technology: identify the actual problem, establish a baseline of current performance, intervene, then measure the outcome.
Azerbaijan’s ASAN Service model, a public-facing digital services framework launched in 2012, drew sustained attention.
Ulvi Mehdiyev, chairman of the State Agency for Public Service and Social Innovations, said 52 countries were studying the model and that more than 600 service centres worldwide had been built on its principles, collectively serving roughly 800 million people.
Anar Aliyev, Azerbaijan’s minister of labour and social protection, said the DOST Agency, which provides 164 categories of services through a single-window system, had digitalised more than 90% of its offerings.
Turkey’s e-government gateway, launched in 2008, now has close to 70 million registered users and records more than one billion annual clicks on social security services alone, Labour Minister Vedat Işıkhan said.
Greece’s social security agency, e-EFKA, is using AI to clear administrative backlogs while keeping complex cases in human hands. Governor Alexandros Varveris said AI handles straightforward decisions while staff manage cases requiring judgement.
Several speakers warned that automation without oversight risks encoding existing inequalities into welfare systems. Mehmet Ali Varol, a data science manager at Doğuş Teknoloji, said AI inherits the biases of its designers.
“This technology comes from us,” he said. “It is also injecting some of the biases, some of our perspectives, into this technology.”
Christian Fillet, chair of the European Social Network, said digital transformation in welfare services, if poorly governed, “can reinforce inequalities, introduce bias in decision-making, reduce human oversight.”
“Smart governance is not only about smart systems. It is about fair outcomes, transparency, accountability, and human dignity,” he said.
Marcelo Abi-Ramia Caetano, secretary general of the International Social Security Association, said the integration of AI for DOST Social Awards 2026 into the main programme offered “a valuable platform to recognise innovative applications of artificial intelligence that can contribute to better social services.”