Page 115 - Azerbaijan State University of Economics

THE JOURNAL OF ECONOMIC SCIENCES: THEORY AND PRACTICE, V.77, # 1, 2020, pp. 113-132

contribution to unlocking its hidden resources, as well as providing significant help
in controlling data flow. Designing and executing AI models thus not only provide
insight into useful information for the company but also help leverage effective
business information and get the best quality data. Because AI can handle many
different types of data (data with different wording, style, content, and length), the
tagging process makes it easy to distinguish between these data sources. Once the
records from multiple data sources have been merged, the unification mechanisms
will then be able to update the data in time. Last but not least, there is a need for a
validation process consisting of validating the model and the indicators. Regardless
of the industry in which AI is wanted to be implemented and integrated, meeting
these requirements is an important part of the preparation phase and proves useful in
the practical implementation. (Kessler and Gwozdz, 2018).

However, it must not be forgotten that, in addition to assessing needs, problems, and
objectives, infrastructure requirements must also be met. On the one hand, the
transition to AI and integration will involve a significant financial investment, as it
will be able to change the overall corporate structure and workflows to date. As
technology becomes more complex, the cost of resources increases. Nowadays,
cloud solutions and cloud technologies are the foundation of AI, so when choosing
the right platform, the following should be evaluated.

 High computing capacity, which means proper performance computing, including
CPUs and GPUs that are suitable for AI integration.
 Storage capacity is also important due to the increase in data volume, but in some
cases, a company needs to consider whether it prefers a system with more capacity
but slower, or a system with less capacity but faster. For the most part, AI solutions
can work effectively when as much data and applications as possible are available.
 Network infrastructure, i.e. communication within the network, scalability,
bandwidth, and uniformity.
 The security issue, i.e. the confidentiality of sensitive data such as e.g. in health
care, the patient register, in the case of a financial institution, personal data, and
financial information. Poor integration of AI can cause system vulnerabilities, but AI
cannot work effectively.
 Cost-effective solutions. Nowadays, the more complex and advanced a system is,
the more expensive it is. For the infrastructural performance to bring out the best
possible and maximize, the company or unit performing the implementation must also
take into account the continuous increase in costs. But also, after a thorough decision,
it can be expected that the performance of the company will increase, which will even
result in cost reduction if the AI is applied effectively (Hofstee, 2019).

115