Public familiarity with AI has increased over the past year, largely because tools became widely accessible and many people tried them in everyday contexts, rather than as a result of deliberate education efforts. According to Tiszavölgyi Péter (Stylers), while baseline user knowledge has grown, true AI literacy is still lacking: the ability to recognise when models make mistakes, to probe answers more deeply and to check outputs.
Tiszavölgyi compares the current situation to handing out high‑powered cars without driving licenses or traffic rules — unrestricted use of a powerful new technology carries significant risks.
Risks behind rapid adoption
Experts point to workforce transformation as a direct consequence: several companies have carried out 20–30% headcount reductions particularly in administrative roles without thoroughly reconsidering organisational workflows. Tiszavölgyi warns that society does not yet have effective responses to many of these developments.
Leadership communication: slow down and listen
Szerémi Péter (communication consultant) observes that an enthusiastic leader often ‘‘rushes’’ colleagues who are more uncertain about AI. The recommended approach is not to flood them with arguments about why AI is the future, but to slow down and listen to their fears. A typical mistake is to argue with someone who says they don’t need AI — that statement often masks deeper anxieties about status, competence or job security.
If leaders themselves are unsure about the future, Szerémi advises acknowledging uncertainty honestly while providing concrete short‑term guidance. For example, promising no major changes until the end of the year can reduce employee anxiety.
Building an AI strategy that creates real business value
Laczkó Gábor (Stylers) proposes three building blocks for strategy: a steering guideline from management (a lane‑keeping assist), a baseline assessment of organisational maturity (automation, data, organisation, regulation), and a deep analysis of current processes to identify bottlenecks and manual work. That yields a prioritized set of use cases — a ‘‘treasure map’’ — from which implementation can begin.
Typical challenges and remedies
-
Organisational and communication issues: management pressure for rapid rollouts without a clear strategy, and failure to explain to employees why AI is needed. Often the goal (improving work quality rather than replacing jobs) is not communicated, which would immediately reduce resistance.
-
Technical obstacles: legacy systems, data silos, inconsistent formats and tacit knowledge (undocumented expertise held by employees) complicate integration. Another common problem is unrealistic management expectations — for example, expecting 100% accuracy from a system that achieves 98%; human workers likely erred at a similar rate.
-
Risk management: organisations must define where errors are acceptable. In clinical decisions, zero tolerance for mistakes may be required; for an occasional misposted invoice item, a low error rate might be economically tolerable. Effective integration requires clear rules about where human work and oversight are necessary.
SME vs. enterprise: common and differing challenges
Laczkó notes that both SMEs and large companies need extensive employee education. SMEs can be faster innovators due to lower regulatory and organisational friction, but often start from a more chaotic automation/data readiness. Large enterprises are slowed by parent‑company constraints, organisational complexity and regulatory uncertainty; in practice, it’s rarely technological immaturity that blocks progress, but organisational, regulatory and ROI ambiguity.
How to evolve an existing AI strategy
AI transformation is among the most complex organisational changes companies will face. Laczkó recommends initial external consultancy to push the effort forward and set directions, then building in‑house core competencies to continue development. Consultants can provide breadth of perspective and prioritisation across value‑complexity tradeoffs; organisations typically gain that experience only over multiple transformation cycles.
Technological trends: multimodality and natural language interaction
Tiszavölgyi expects a major advance in multimodal models that can interpret images, video and text simultaneously, with strong implications for robotics. In software development, he foresees a shift where traditional programming languages may give way to interactions driven increasingly by natural language.
Conference participation and practical workshops
The experts will take part in the 2026 HVG AI Talks conference. Their workshop will reenact real situations to illustrate organisational conflicts that arise during AI adoption and demonstrate how to resolve tensions from communication, business and technology perspectives. They expect the event to produce valuable professional discussions and practical ideas for company practice.
Ticket information: tickets for the 2026 HVG AI Talks conference are available on the aitalks.hu website.
This content was produced within the framework of AI Talks, the HVG event, with the cooperation of HVG BrandLab; the editorial offices of HVG hetilap and hvg.hu did not participate in the production or editing of the content.
