Skip to main content
Back to News

The Rubber Stamp and the Silicon Chip: Inside Armenia’s AI Revolution *

By Narek Tovmasyan, PMP®

February 12

AI leader.jpeg


By Dr. Narek Tovmasyan


For decades, the soul of Armenian public administration has been captured by a single, authoritative sound: the rhythmic thud of a rubber stamp hitting paper. In this post-Soviet corner of the Caucasus, legitimacy has long been measured by the wet ink of a signature and the physical heft of a dossier. In his recent research Dr. Narek Tovmasyan defines this as “paper loyalty” - a deep-seated cultural belief that if a process isn’t printed and stamped, it isn't real.

Yet, walking through the high-ceilinged government corridors of Yerevan today, a quieter sound is taking over: the soft click of keyboards as deputy ministers and policy analysts prompt large language models. Dr. Narek Tovmasyan have found that Armenia is becoming an unlikely laboratory for one of the 21st century's most urgent questions: how does artificial intelligence reshape the way a state is led?

In his recent study, which included a methodical review of the Digital Armenia Strategy (2022) and in-depth interviews with the "strategic echelon" of the government, Dr. Tovmasyan discovered that we are not witnessing a total automated takeover. Instead, we are seeing the birth of a new paradigm: the Augmented Leader.

Strategic Pockets and Persistent Shadows

To understand AI in Armenia, one must discard the idea of a top-down technological tsunami. Global trends suggest that AI adoption is inevitable; indeed, according to the OECD (2023), 45% of governments globally have already deployed AI-powered decision support systems. In Armenia, however, the journey is a patchwork of "strategic pockets."

A 2023 survey by the Armenian National Center for Innovation and Entrepreneurship revealed that while 32% of Armenian public sector entities have begun experimenting with AI, only 12% report actively using it in decision-making processes. These "innovators" are often driven by cold necessity. The Ministry of Justice, for instance, used an LLM-based system to digitize 15 million civil records - a task that would have consumed human lifetimes. As one Deputy Minister told me during my research, “We’re implementing AI where we think it makes sense.”

However, the ghost of the machine remains. The "paper loyalty" I observed is a classic example of "institutional inertia" - the resistance of long-standing routines to change (DiMaggio & Powell, 1983). Even when digital systems provide perfect efficiency, officials often default to paper to feel "secure" in their authority.

The "Human Firewall"

Perhaps my most striking finding is how Armenian leaders are protecting their own agency through a purposeful bifurcation of the decision-making process. I have termed this the "Human Firewall."

In my study, 100% of participants insisted that strategic judgment must remain a human monopoly. They use AI as an "efficiency engine" to synthesize international legal precedents or manage "Information Synthesis," but they strictly bar it from the final judgment. As Wirtz et al. (2020, p. 1052) argue, AI causes a "cognitive shift" from purely human-led judgment to a model of data-assisted judgment.

This Armenian "Firewall" is a safeguard against what scholars call "algorithmic satisficing" - the temptation for a leader to accept an AI’s answer simply because it is believable and attractive (Simon, 1972). One Department Head I interviewed summarized this verification culture with a local twist on an old proverb: “Measure seven times, cut once… it is AI, and that is why we are human.”

The Specter of Cognitive De-skilling

Beneath the technical optimism lies a profound anxiety. About 93% of the leaders I interviewed expressed a fear of "cognitive de-skilling." The concern is that by outsourcing analytical "muscle" to machines, public officials may lose the capacity to think critically (Goddard et al., 2012).

One Deputy Minister shared a haunting story of a relative who abandoned a career in data analysis, feeling they "cannot compete with this automated toolkit." This aligns with the warnings of Floridi and Cowls (2019), who caution that phronesis (practical wisdom) may be undermined by excessive dependence on AI. If we stop synthesizing information ourselves, we risk falling into an "accountability vacuum" - a state where no one feels responsible for an AI-informed error (Bovens, 2007).

The Path to an Augmented State

Armenia currently ranks 76th globally in the Government AI Readiness Index (2023). To move forward, my research identifies three imperatives for policy-makers:

  1. Closing the Training Gap: Currently, learning is "organic" and ad-hoc. We need a National Leadership Development Program based on the 70-20-10 model: 10% formal instruction, 20% social learning, and 70% experiential learning through "innovation labs" (Ministry of High-Tech Industry, 2023).
  2. Establishing Ethical Stewardship: Leaders must move from being "Information Processors" to "Purpose Setters." As Davenport and Kirby (2016) argue, the main benefit of AI is augmenting, rather than replacing, human cognitive capacities.
  3. National Governance Frameworks: We must establish an independent AI ethics council to oversee high-risk projects and ensure "Contestability by Design" - allowing citizens to challenge algorithmic choices (Dignum, 2019).

Armenia’s journey suggests that the most technologically sophisticated public administration of the 21st century will not be the one with the fastest algorithms, but the one with the most resilient human leaders. By embracing "augmented leadership," we can build a state that is both incredibly efficient and fundamentally human.

The thud of the rubber stamp may eventually fade, but the weight of human responsibility must remain absolute.


References & Further Reading

  • Berryhill, J., et al. (2019). Hello, World: Artificial intelligence and its use in the public sector. Paris: OECD Publishing.
  • Bovens, M. (2007). 'Analysing and assessing accountability: A conceptual framework', European Law Journal, 13(4), pp. 447-468.
  • Davenport, T. H. & Kirby, J. (2016). Only Humans Need Apply: Winners and Losers in the Age of Smart Machines. New York: Harper Business.
  • Dignum, V. (2019). Responsible Artificial Intelligence. Cham: Springer.
  • DiMaggio, P. J. & Powell, W. W. (1983). 'The iron cage revisited: Institutional isomorphism and collective rationality', American Sociological Review, 48(2), pp. 147-160.
  • Floridi, L. & Cowls, J. (2019). 'A unified framework of five principles for AI in society', Harvard Data Science Review, 1(1).
  • Goddard, K., et al. (2012). 'Automation bias: a systematic review', Journal of the American Medical Informatics Association, 19(1), pp. 121-127.
  • OECD (2023). The Strategic and Responsible Use of AI in the Public Sector. Paris: OECD Publishing.
  • Simon, H. A. (1972). 'Theories of bounded rationality', in Decision and Organization. Amsterdam: North-Holland.
  • Wirtz, B. W., et al. (2020). 'An integrative public management model of smart government', Public Management Review, 22(7), pp. 1052-1075.


* Note: This article is based on the original research Dr. Tovmasyan conducted as part of his Executive MBA studies through the Pan-European Executive MBA, a dual degree programme offered by the University of York (United Kingdom) and the University of Strasbourg (France).