Page 120 - Azerbaijan State University of Economics

Ilkin Seyidzade, Rudnak Ildiko: General Implementation Processes of Artificial
Intelligence and Its Economic Effects in Hungary

immediately in the event of a problem. In both cases, continuous education and
training is required, or outsourcing a given task can also lead to results.

And last but not least, business issues should be mentioned. The lack of business
alignment can even be linked to the people issue, as it means that the adoption of
technology could not be incorporated into the corporate culture itself. Not only
employees but also managers need to be required to understand AI so that it can be
incorporated into corporate goals and strategy. The absence of AI knowledge may
hamper implementation in several organizations. Choosing the right service provider
is at least as important from a business standpoint, as the system can only be
effective if it is used for actual business tasks and has the support. Before
implementing AI development, it is advisable to examine what AI solutions
providers can propose to the company and then evaluate and select based on that.
And to overcome the integration challenges, it is worthwhile to go step by step
during the implementation, because if the integration takes place at all levels of the
company at the same time, it can even be detrimental. And finally, legal issues can
be mentioned here. It is also worth clarifying from a legal point of view before
implementation that e.g. what damage can the AI cause, whose liability is it? How to
collect data? How to manage data? How to protect data? As there are no clearly
adopted or enacted rules on them yet, it is necessary to clarify the legal aspects and
fix them in a contract or agreement (Polachowska, 2019).

Concerning international security and international law, AI is currently an immature
field that is now critical to nations, too. Since there are no limits to algorithms and
there is no standard and universal control at the moment, there is no clear boundary
between ownership and the origin of the data. Thus, for the time being, there is no
consensus on all business models and their applications (Pandya, 2019).

According to Doucette (2019), according to a Gartner survey, around 85% of AI
projects up to 2022 will produce flawed results due to rudimentary data management
and algorithms. Some critics believe that AI will not deliver the hopes it brings and
the expected values. On the one hand, initial models are not scalable or too
experimental for use by any customer. Normal development processes receive
insufficient emphasis, and transition and transition times are not suitable for
finalization. Although some companies follow the project transfer steps closely,
there is no proper workgroup and production support workflow. On the other hand,
the transfer process may be inadequate, that is, innovation teams that are sharply
separated from, and not accepted by, non-digital teammates. While an AI solution
can meet the company's goals, the challenge is to use it within the company.

120