Sanjeev Bansal , first of all I believe a good introductory seminar explaining what is AI will help, as many people goes with the "AI hype" thinking that is some kind of super smart technology that will solve everything... in reality the AI currently being used is "weak" AI (https://www.investopedia.com/terms/w/weak-ai.asp), so basically what you will be doing in finance is probably implementing some machine learning algorithms and bots, while for audit/accounting probably some text-mining algorithms. Depending on the computational needs, this can be done in a (good) single computer or in a cloud-based HPC environment. So I believe in organizational terms, it implies for everyone to understand what AI is and is going to imply, who is going to be in charge of it, and under which environment computations will be made

Sanjeev Bansal ,

On the human side

Before implementing any AI it is good to determine whether it is going to automate a task currently being carried out by personnel in the department. The process of automation is usually stressful for employees and needs to be addressed appropriately (such as reassuring employment and improvement towards more important tasks).

Be sure that the technology is not intrusive. Some implementations are more burdensome than helpful. Some implementations become so complex that cause errors nobody knows how to fix or trace. This will frustrate users and the technology will be abandoned.

On the data side

In terms of implementing AI, it is good to know the state of the data. Sometimes it is assumed that the data is already processing friendly when in fact it is not. It is also necessary to check to see what is the error percentage in the data.

Be sure that the technology works. Simple enough statement, but often overlooked. It is not the same to implement AI on a toy problem data set than on real data.

You can find this study helpful:

Chapter Applications of Blockchain Technology in the Finance and Ban...

Further you can consider the following profile for your research:

https://www.researchgate.net/profile/Sitara-Karim/research

Thanks a lot

My own paper in this area is :

Article Blockchain: The Future of Accounting

A company should follow these steps in introducing and then implementing artificial intelligence (AI). (1) Get familiar with AI; (2) Clearly identify the problems the firm wants AI to solve; (3) Prioritize concrete value (what value are you going to create using AI?); (4) Acknowledge the internal capability gap; (5) Bring in experts to set up a pilot AI project; (6) Form a taskforce to integrate data; (7) Start small (begin applying AI to a small sample of your data, use AI incrementally to prove value, collect feedback, and then expand accordingly); (8) Include storage as part of your AI plan; (9) Incorporate AI as part of your daily tasks; (10) Build with balance (build in sufficient bandwidth for storage, the graphics processing unit, networking, and understand the data to be used). A Finance Department should introduce data science in the finance curriculum as well as machine learning algorithm and advanced programming (Python/R/Julia). In teaching AI to students, Professors should always check data quality, biased data, and the dimensionality reduction of the data. Principal component analysis, linear discriminant analysis, t-distribution stochastic neighbor embedding, or uniform manifold approximation and projection can be used to address these data problems. Q.E.D.

Does the creativity in accounting really need AI?

:-)

If applied, what will happen then...

This article discusses Artificial Intelligence in Finance and Accounting favorably while providing essential information about downside risk, why digitization has been delayed in this upper echelon of firms, and resource references listings:

https://www.dpublication.com/wp-content/uploads/2020/11/11-8011-1.pdf

Nancy Ann Watanabe ,

Thanks for sharing this paper!

I wish to highlight an interesting clash (and I take this was the author's intention) between the theory of limited rationality and the efficient market hypothesis stated in the paper that:

If we take away from the section on limited rationality one does not have perfect information to make rational decisions, then the statement in the next section that "by incorporating more artificial intelligence , the more efficient the market have become" is not true.

I tend to believe based on the statement above that, the theory of limited rationality does trump hypothesis for efficient markets when it comes to AI. This makes the perspective theory the same to that of humans when it comes to AI.

Even if one closes the window to make AI as efficient as possible in terms of market efficiency, its knowledge will be based on imperfect information making it only as good as the information received. Further, the AI will not be able to predict black swans when they occur and their response might be even worse since they do not have the ability to rationalise that it is in fact a black swan event.

AI is a wonderful technology, but one must always assess its potential as well as its limitations.

According to my survey, which was addressed to firms' finance and accounting directors, the main barriers to AI adoption so far have been poor data quality and a lack of technical knowledge. Therefore, my suggestion would be to fix these two issues first and only apply AI algorithms afterwards.

You can check my publication in German and translate it if necessary:

Article Predictive Analytics im Rechnungswesen – Eine empirische Bes...

1. Management's decision.

2. Organization of databases that store information about business processes and their results.

3. Organization of automated programming interfaces (APIs) between business process databases and an intelligent system.

4. The presence of an intelligent system, the capabilities of which provide a solution to the task.

5. Creation, financing and training of a group of specialists responsible for solving tasks according to certain regulations.

Everything is as described here: http://lc.kubagro.ru/aidos/aidos96/index.htm in the section: CHAPTER IV. THE INFRASTRUCTURE OF THE APPLICATION OF THE “EIDOS” SYSTEM.