At a roundtable organised by Deloitte and ABSL, Global Business Services (GBS) automation experts reported that most companies still do not have a clear artificial intelligence (AI) strategy. The discussion focused on practical experiences and strategic issues: how to turn AI into reliable, business-ready solutions and which pitfalls organisations typically face.
Core issues: absent strategy and governance
Participants said a common mistake is enthusiasm for AI without clearly defined objectives. Many organisations wrestle with the tension between broad use of general-purpose models and the need for solutions tailored to specific business problems. This often results in fragmented pilots and unrealised potential, preventing interest from translating into measurable business value.
Speakers emphasised the importance of governance. Several organisations have set up dedicated AI committees—typically composed of IT, compliance and automation specialists—to ensure consistent strategic principles and responsible AI use. These bodies also handle complex topics such as data ownership and ethical deployment.
“Looking back, it is clear that many organisations did not define a clear AI strategy: they either pursued mass adoption without a purpose or tried to solve everything with a single tool. The fundamental question of what we want to achieve with AI was often neglected.” – Lukács Eszter, GBS Advisory and Finance Transformation Director at Deloitte Hungary.
Implementation approaches and tangible results
In practice, both “top down” executive-driven programmes and more organic “bottom up” initiatives were discussed. The latter include internal AI accelerators, hackathons and pilot initiatives that promote experimentation and reduce user anxiety.
Examples of successful applications presented at the roundtable included:
- optimisation of pricing and promotion strategies,
- systems to detect abusive product returns,
- automatic processing of incoming emails and documents,
- AI-driven simulations for process optimisation,
- intelligent inventory management solutions.
These cases demonstrate AI’s potential to increase operational efficiency and improve customer experience. Major AI investments typically deliver payback within 18–24 months.
Risks: not all tasks are suitable for LLMs
Experts warned against treating AI solely as a cost-cutting tool, which can divert attention from revenue growth, capability building and fixing operational weaknesses. They highlighted differences between large language models (LLMs), specialised smaller language models (SLMs) and traditional machine learning solutions. While LLMs excel at interpreting and processing language content, they can generate inaccurate or fabricated information; hence tasks requiring high precision—such as financial forecasting or contract analysis—often need more targeted solutions.
“AI offers enormous opportunities, but we must recognise the important role of human expertise in verification and content validation. Incorrect expectations and lack of user knowledge carry serious risks, which underlines the value of experienced specialists.” – Páll Zoltán, AI Manager, Technology & Transformation, Deloitte Hungary.
Talent development, public incentives and the human factor
Talent development and reskilling were recurring themes. Although Hungary has strong foundations in analytics, data science and engineering skills, rapid AI advances often outpace formal education and internal training programmes. As a result, organisations increasingly invest in internal training that covers technical skills as well as soft skills such as resilience, problem solving and proactivity.
Participants also noted the role of R&D incentives and public grants: in Hungary, AI-related development—especially projects that involve experimentation, uncertainty or novel combinations of existing technologies—often qualify for R&D tax incentives and non-repayable grants. Eligible costs can be recovered in the range of 10–65%, improving project economics and enabling experiments that might otherwise struggle to secure internal approval.
The human dimension was emphasised as well: both excessive trust in AI and outright rejection can undermine well-designed projects. Experience shows that AI primarily augments human capabilities rather than fully replaces them, making the active involvement of domain experts in validation and continuous tuning indispensable.
Data strategy and alignment
A strong data strategy and data governance were identified as foundational pillars. Even though modern AI solutions, particularly LLMs, can handle unstructured data, standardised, high-quality data significantly increases accuracy and reliability. Aligning AI strategy with data management frameworks—often run by different teams—remains a challenge and requires close collaboration across functional areas so that data become truly fit for business use, not just AI-compatible.
Looking ahead: adaptability and continuous learning
GBS leaders agreed that integrating AI successfully requires continuous learning and organisational adaptability. The pace of AI development demands infrastructure that can evolve with the technology rather than only addressing current needs. Organisations that treat AI as a strategic growth lever—not merely a cost-reduction tool—and combine strategic vision with agile execution and human-centred design are best positioned to succeed in the rapidly changing AI landscape.
