Bottom line: AI projects deliver value only when business departments define use cases, IT acts as an enabler, and clear governance rules limit risks.
A Horváth study shows that many companies treat AI primarily as an IT issue, launching expensive experiments without clear benefits. Successful implementation requires business departments as initiators of use cases and IT as technical enablers.
Aggressive sales strategies from AI vendors meet high investment readiness on the part of companies. However, this often leads to technology taking centre stage before it is clarified what concrete benefit AI should provide for operational processes. The result is costly projects without added value.
The decisive mistake lies in role distribution: business departments know operational bottlenecks and inefficient processes best, but are often not involved in the initial requirements process. Instead, the IT department alone defines the requirements. A sound foundation is created only through close collaboration: business units as initiators of specific use cases, IT as technical enabler for integration, scalability and stable operations. If AI is treated solely as an IT issue, long-term process anchoring cannot succeed.
CDOs should proceed systematically: first analyse which use cases offer efficiency gains — for example through optimised processes or relief from repetitive tasks. Only then comes technology selection with clear criteria: realistic performance promises, the vendor’s process and industry expertise, data protection concept, traceability of AI results and quality assurance. According to the Horváth study, many companies perceive the price-to-performance ratio as unbalanced; hidden costs for data preparation or technical adjustments substantially diminish economic viability.
Another governance risk: four out of five surveyed companies fear that employees will adopt AI-generated results without question. Binding rules are required — no personal or internal data in public AI systems, systematic plausibility checks, confidence scores and parallel use of multiple models. The “Human in the Loop” approach ensures human control, provides feedback for model improvement and reduces error risks.
Source: www.it-daily.net · Published 26 July 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.3.