# Challenges of implementing and applying artificial intelligence

In recent years, hardware segments have achieved remarkable processing capacities. Mobile devices now surpass the computing power of computers available on the market, while Data Centers operate on scales ranging from petabytes to yottabytes (equivalent to 1,000,000,000,000,000,000,000,000 bytes).

However, despite this processing power, there are still gaps that disrupt the continuity of necessary calculations in AI models. Depending on the scale of data, components, and ecosystems involved, a supportive infrastructure is required to ensure processing scalability and facilitate the rapid insertion and modification of components within a corporation's systems (ZHANG & DAFOE, 2020).

Conducting careful systems architecture analysis is essential for company projects. It helps predict the initial infrastructure needs and anticipate the evolutionary steps necessary to maintain AI solutions that assist in governance (GARG & GOEL, 2021). This analysis serves as a primary means of organizing the target functionalities of the solution and assessing whether AI is genuinely achieving positive results for the organization.

It provides a way to mitigate false alerts, identify specific limits, and guide the business's sustainable evolution. Solution architecture practices also address numerous challenges, including the four main points identified by the market for developing an AI solution (DAVENPORT & RONANKI, 2018):

* Bottlenecks: Technology and system architecture can create barriers to machine learning (ML) due to the lack of evolution in development languages and integration standards. The significant content differences between development versions can hinder ML progress.

* Scale: In some cases, a lack of content may not be the issue. However, high development productivity and the various versions of development languages available can become physical barriers, limiting the processing capacity within the solution infrastructure.

* Overestimated infrastructure: Productive stability and consistency of standards, such as development language version limits and utilization of capabilities, can be crucial for determining the infrastructure. Strategic alignment with business objectives becomes challenging, as it involves financial risks associated with providing either an excessively capable or inadequately forecasted infrastructure when service usage increases.

* Preparation of learning models: When information bottlenecks are not prevalent, scale and infrastructure are planned based on the needs of the business strategy. However, companies that maintain structural silos within their organization may face challenges if the data model does not align with development standards.

To enable the AI solution to meet these key capabilities, the use of visual means such as dashboard panels is suggested. These panels are designed to present metrics, decisions, and actions resulting from AI analysis conducted within the systems architecture framework, addressing the challenges of ML development.

## Conclusion

Seeking to bring a view that is positive for the maintenance of the life cycle of components, in the same way for the data scientist to understand and monitor the work performed by ML (ERIKSSON et al., 2020).

Once efficiency is found and this entire structure is delivering satisfactory performance and information, the additional capacity for AI that will deliver a lot of value to the company will be to predict in which business unit the focus should be on the evolution of systems within a financial limit or even for the technology objectives to be aligned with the metrics and the business objectives.