AI is becoming a part of life, appearing at the forefront of the research agendas of countries worldwide while leading the technological landscape as the most powerful tool for businesses today. Assessing how businesses can harness this power and profit by automating their internal processes is a topic massively relevant in the competitive corporate world. As AI progresses and gets used worldwide, it’s imperative for businesses to not only consider the technical limitations of adopting AI, but also the ethical implications that are part of all AI use. This paper examines the prospect of AI as a strategic tool used by businesses today while giving a comprehensive outlook on its technical limitations and a detailed discussion of the various ethical considerations related to its use in the business world.


Artificial intelligence, or AI, is a widely thrown-around term in today’s technology-driven world. But what does it actually mean? Artificial intelligence is a broad umbrella term that covers any task associated with intelligent beings that is performed by a computer-controlled system. AI is observed in countless tasks, from everyday routines such as using Google Maps to help us navigate to big industrial tasks such as production or market analysis. In this systematic review, we plan to narrow down the focus to the world of business, more specifically where artificial intelligence fits into the current corporate and financial scenario, what are the shortcomings, and what are the ethical controversies that follow the widespread usage of AI.

Artificial intelligence has given firms around the world multiple opportunities to optimise their production and services while simultaneously augmenting efficiency and productivity. There are various methods by which firms can utilise the facilities of AI, ranging from creating forecasts of demand, enhancing inventory control, refining customer service platforms, personally catered marketing, and innovation in the production processes or the product itself. 

AI encompasses a vast array of technologies; thus, this paper will start by breaking down the major branches of AI. The first branch is known as machine learning (ML). ML, although itself a broad term, is any algorithm that is produced that attempts to imitate the way that a human brain thinks and behaves to produce results and output that would emulate the predictions that a human would make (IBM, 2023). The second main process is known as deep learning (DL), which is a subset of machine learning. In this process, the algorithms take inspiration from the human brain and thinking techniques. This involves techniques such as ANNs, or artificial neural networks, which are layers of nodes that imitate the neurons in human brains and process input data using weights and biases to create predictions or verdicts. The third branch is Natural Language Processing (NLP), which is a tool that implements algorithms that can understand speech and text written by humans and then internalise this information to produce valid and accurate outputs. NLP is ubiquitous in the area of voice-to-text AI models, AI chatbots, and other voice assistants. 

Moving to the physical realm, we have widespread advancements in robotics (SoCi, N.D.). Robotics is a boundless AI branch that involves physical robots that can replicate almost exactly the physical processes that humans carry out, semi-automatically or completely on their own. Robotics are extensively used in the secondary sector during manufacturing processes. 

Another important process that is visible in everyday life is the use of expert systems. This involves a user entering an industry-specific query, and then the algorithm attempts to formulate a logical answer from the information stored in its knowledge base. Lastly, a vital tool in our futuristic world is fuzzy logic. In a world of binary tools involving a true or false output, sometimes there can be an in-between or a partial truth. Fuzzy logic utilises algorithms that can determine the degree of truth or the degree of uncertainty during decision-making (Blocchi, 2023). Other simpler tools have also been enhanced by AI, such as the simple linear regression models used in ARIMA (Autoregressive Integrated Moving Average), which are linear graphs that are modelled based on data points. These linear models allow firms to understand the correlation between data points as well.

This paper will first discuss the different methods by which firms have enforced the plethora of AI tools in the literature review. Second, it will provide an in-depth analysis of the strengths and limitations of AI in business while maintaining a critical outlook on the ethical concerns raised by the usage of AI in business. Finally, it will bring everything together in the conclusion.

Literature review

Demand Forecasting

AI has paved its way into the realm of business by allowing producers to optimise their production processes with numerous practical applications. To begin with, AI has given producers the ability to carry out a process known as predictive analysis using machine learning and data mining techniques (Bharadiya, 2023). Predictive analysis allows firms to gain insight into the future of demand and supply variables by organising and modelling historical data. Not only has AI given firms the chance to carry out the processes, but it also offers high attitudes of accuracy (Bharadiya, 2023). An important application of predictive analysis is observed in demand forecasting. When carrying this process out, businesses use a tool called machine learning. ML involves a technique known as ANN. The inputs to this neural network consist of past data that the company has collected from POS (point of sale is the point at which customers purchase the goods, e.g., cashiers or online payment portals) or market research (IBM, N.D.). Then weight is attached to each variable input, which is the importance of that specific factor in the prediction of demand. Finally, an output is predicted after the data is processed.

AI also allows businesses to execute a process known as market basket analysis. This is a process that analyses historical data from POS transactions using the association mining rule. This process entails algorithms that analyse the relationship between the different products that customers buy at a POS. It allows the firm to understand the way consumers think and buy products, allowing them to predict demand more efficiently (Kurnia, 2019). Additionally, firms also make use of a process known as outlier analysis. This involves two unsupervised ML algorithms. Firstly, the algorithm is given input data in which it has to observe different patterns or relationships, and then it can cluster the data according to those patterns. Then, if the algorithm can find data points that do not fit into these categories, it can mark them as outliers (Kang and Kaur, 2016). 

However, as and when we use these AI models, we must understand that the outcomes and predictions that the AI models are making are heavily dependent on the data that is fed into the algorithms. So companies will have to spend time and financial resources ensuring that the data is precise, accurate, and free of any discrepancies (Cederberg and Elliot, 2020). Moreover, AI has difficulties taking into account all the factors that play a role in demand forecasting. It is also important to take the qualitative factors into account that would require human thinking capabilities.

Supply Chain Optimization

Supply optimization is a key part of any business, and the myriad of AI tools that were discussed in demand forecasting play an influential role on the supply side as well: the sorting and grouping of data, the identification and removal of outliers, and overall gaining a more meaningful insight into the real-time and static data with greater accuracy and reduced costs (Helo and Hao, 2022). 

A model known as ARIMA allows firms to analyse time-series data and create future predictions. This process, with the aid of AI, makes linear regression models on past-lag data and then uses moving averages to smooth out the noise in the time series data. This allows much more errorless predictions, ensuring that firms do not overstock or understock (Hayes, 2023). Controlling the inventory has become much simpler and more efficient as a result, minimising stock waste and inventory control costs. Furthermore, a struggle that many firms face is countering the bullwhip effect. The bullwhip effect is a phenomenon that occurs when small fluctuations in demand lead to slowly repelling effects on the supply side of the supply chain. These effects get carried over and lead to greater supply-side fluctuations than necessary (Blanco, 2016). This increased precision in utilising AI allows firms to experience fewer shocks from demand-side change. A study has shown that there was almost a 6.4% increase in the accuracy of demand forecasting as a result of implementing AI, boosting capital efficiency, and cutting down on inventory control costs (Cederberg and Elliot, 2020). The role that AI plays in this is that it allows for better analysis of non-linear patterns as well. This can enable the firm to identify non-linear, possibly exponential relationships.

A significant drawback that firms can face when creating and implementing ML models to carry out this process is the issue of “overfitting.” This occurs when the algorithm performs swimmingly on the training data set but then fails to give similar results on the test data set, which is unseen data, defeating the very purpose of the algorithm (Ying, 2019). This may occur due to different data types on the training set, or it may memorise outliers or noise from the training set, reducing its ability to adapt to newer data sets (IBM, 2023). Yet, we must discern the models of AI that are constantly receiving upgrades and developments.

Personalised Marketing

Methods utilising the data-processing and predictive capabilities of AI in marketing have only recently been adopted by various businesses (Wierenga, 2010). One of the primary methods of automated marketing processes includes using AI to analyse customer data, behaviour patterns, and preferences to outline specific customer segments. Analysing vast amounts of qualitative data, ML algorithms in AI can extract patterns in customer behaviour, outlining customer preferences and allowing businesses to offer personalised marketing campaigns, product recommendations, and tailored customer experiences. A pioneering example today includes Netflix, which saved $1 billion in customer retention costs (Gomez-Uribe and Hunt, 2015) by recommending TV shows to customers based on their preferences or the preferences of other customers with similar purchasing behaviours (Dwivedi, R., et al., 2020). These personalised recommendation systems allow businesses to improve their cross-selling and upselling opportunities, promoting conversion rates and customer retention when customers consistently receive the customer service and products they prefer.

Despite the opportunities that ML algorithms provide in personalised marketing, the extent to which businesses can capitalise on these IT-driven marketing strategies depends largely on the data used to train the AI. Inaccurate data can provide inaccurate results, leading businesses to funnel resources towards an unprofitable avenue that can raise various financial risks (Szilágyi, R., and Tóth, M., 2023). Due to the importance of data in AI, businesses must ensure they use quality data in their AI-based marketing projects to reap benefits and avoid risks. 

Data becomes another issue in personalised marketing when the data used infringes upon human rights. Personal data is often embedded within big data, and the methods used to originally capture this data aren’t always transparent (Okorie, G.N., et al., 2024). Businesses have to be careful to avoid unlawfully and unethically using the personal data of individuals who did not consent to their data being used. A further exploration of how data becomes a significant technical and ethical limitation for businesses using AI will take place in the discussion section of the paper.

Automated Customer Service

Businesses today have capitalised on the capabilities of AI by improving the physical and real-time buying experience of customers. One such way is by automating their customer service with AI chatbots. Appearing as “staff” available to answer customer queries online on company websites, AI chatbots make use of natural language processing (NLP) techniques to understand and answer customer queries through an online messaging platform. Additionally, machine learning algorithms allow the AI to train with large data sets, continuously improving its understanding and responding capabilities (Bharadiya, 2023a). By adding an AI-powered chatbot to a business’ website, businesses can provide personalised customer support by directly resolving the inquiries and issues of clients. This automates the job of customer service staff, who can now use their saved time to address more customers with complex issues requiring human counselling, allowing businesses to address a wider number of customers, prevent call queues, and expand their outreach (Bharadiya, 2023b).

Integrating AI chatbots that automate customer service lines can also decrease labour costs and allow businesses to reach wider audiences that speak a different language and live in different timezones, permitting businesses to expand to different demand sources in the form of foreign audiences (Mariciuc, D.F. 2023; Kadasah, E.A. 2023). Extended conversations between the AI chatbot and customers can also provide businesses with valuable personal data from the customer, feeding into the machine learning mechanisms involved in personalised marketing that can tailor advertisements to the preferences of customers, increasing engagement and sales. An example of a business adopting AI chatbot solutions to improve its customer service includes Decathlon, which offers a digital customer assistant on the platform “Heydey” that automates 65% of its customer inquiries (Heydey by Hootsuite, 2021). 

Organizationally speaking, the implementation of chatbots in marketing processes may be a challenging internal change. Firms may experience hurdles in executing and planning the IT project, especially depending on the size and goals of the operation. Key requirements for successful chatbot implementation include personnel who are knowledgeable in conducting AI and chatbot processes, as well as an internal culture that is accepting of digital innovation changes (Zhang, Følstad & Bjørkli, 2023).

When using large language models like ChatGPT for customer service chatbots, ethical concerns regarding the methods used to collect the data to train the chatbot arise. Since the personal data of customers may be used by businesses, corporations must be transparent with how they collect and use data. In the case where there were to be data-related issues with the chatbot, it would be ethically necessary for businesses to also conduct self-accountability and corrective action in their marketing practices (Rivas, P., & Zhao, L., 2023.). Organisational challenges and data-related factors are significant considerations involved in the different strategic uses of AI. These issues will be inspected in detail in the later sections of the paper.


Although AI has paved its way into the world of business, making processes such as predictive analysis, market analysis, personalised marketing, and customer service easier than ever before. We shall also consider how impactful it has been in the innovation aspect of business. While the role of AI has become increasingly prevalent in business ventures, they must become familiar with their benefits and setbacks so that they can adapt accordingly. There are countless implications of AI in the world of business, such as automated repetitive undertakings, data analysis, and self-driving automobiles, all of which have proved to have substantially benefited from the integration of AI. (Pierce Denning, M. 2023). In the following paragraph, the main specific implications are explored in detail.

Tesla is one of the pioneers of self-driving automobiles, where artificial intelligence plays a major role. By leveraging this technology, Tesla can make their vehicles safer and more fuel-efficient, as they run on electricity, which makes driving a more expedient experience (Amira, V.Y. 2023).

Similarly, Amazon has found multiple ways of integrating AI into its operations, starting with personalised product recommendations. This is when customers open the Amazon app or website and exhibit a range of different products based on the history of their previous purchases and searches on Amazon. (Seasia, 2024). These recommendations allow Amazon to attain a higher number of sales, and customers are more likely to buy products from Amazon as the recommendations are relevant to their interests. Amazon also uses AI to detect fraudulent activity, a prime example of which is when a customer’s account is attempted to be accessed from another location. Artificial intelligence will detect the operation and timely inform the customer (Seasia, 2024). Ethical issues with innovation and the prevailing integration of AI may be the amount of autonomy that it can be given to make everyday decisions otherwise made by humans and the extent to which the principles of autonomy, independence, and democracy apply to artificial intelligence, especially in life-threatening situations such as Tesla’s automobiles, where granting full autonomy to AI can bring about extreme safety hazards. These ethical concerns will be discussed in detail in the discussion section of this research paper.


AI is revolutionising the world of recruitment, with 43% of human resources professionals using AI in their hiring processes (Agouridis, A. 2023). The reason AI makes this process so much more efficient is its ability to interpret large amounts of data while also being able to pick up on the patterns in the selection process by their human creators. Allowing it to effectively analyse job resumes, conduct pre-employment assessments, and all other application processes much more efficiently as compared to humans. It allows businesses to maximise recruitment efficiency. The average recruiter spends about 30 hours a week on all recruitment processes, decreasing business productivity, whereas AI-based recruitment tools can source through hundreds of thousands of job resumes in no time (Agouridis, A. 2023). On top of that, AI can also prevent human biases from disrupting selection processes; 50% of human resource directors admit to being at least somewhat unconsciously biassed during their selection processes, which can be overcome with artificial intelligence now taking over the recruitment of new employees (Agouridis, A. 2023). While it can prevent some obvious biases, it also retains many of the biases that were a part of its training data. Which, if not recognised by the recruiter, may lead to biassed recruitment that ultimately defeats the purpose of using AI for recruitment.


AI Implementation Concerns

How seamlessly AI solutions can be implemented into the structural and organisational frameworks of a company is a large factor affecting how effectively businesses can leverage the capabilities of AI. Similarly, whether the introduction of AI elevates the strategic capabilities of a business or not depends on the scale of IT specialisation and infrastructure in both the personnel and capital of the business. Not to mention, the process of planning and executing digital innovation projects also involves numerous financial costs, which businesses must account for. These factors outline clear technical limitations in the implementation of AI, which businesses have to consider before introducing AI into their processes.

In terms of organisation, businesses navigating AI implementation will require strong communication of the change with employees before, during, and after the implementation of AI. The intrinsic nature of AI technology mirrors that of a “black box”, and the lack of a foundational understanding of the inner workings of AI within a business can lead to employee scepticism or resistance to AI-based changes. There may be a considerable shift in the work activities of employees in businesses, meaning AI projects will require strong leadership and executive support (Zhang, Følstad & Bjørkli, 2023). Seamless AI integration requires an internal culture that is accepting of digital innovation, and several internal commitments, such as executive pilot projects and AI teams, are often needed to achieve this (Reim, W., Åström, J., and Eriksson, O., 2020.). Particularly due to the specialised expertise needed to effectively use AI, businesses will also require the necessary personnel and IT infrastructure to be able to efficiently adopt AI processes. Whether it be acquiring AI expertise via internal recruitment or capitalising on the benefits of outsourcing human IT capabilities, firms greatly benefit from the strong IT-led analytical frameworks of AI-specialised individuals, especially when AI is still a novel technology. 

Technologically speaking, AI is a regularly-run technology requiring maintenance, training, and continuous supervision. With AI strategically deployed in the marketing and human resources aspects of businesses, the recurring and comprehensive upkeep of AI technologies is significant for businesses to maintain individual stakeholder relationships and brand reputation, demonstrating the importance of IT and AI-specialised human resources (Wamba-Taguimdje, S.L., 2020). Similarly, businesses will require modern IT infrastructure that can effectively run and be compatible with modern AI technology. This would include compatible and modern software and hardware that have the updated systems to efficiently run the AI training process and store the vast amount of data (Kim, J.B., 2019.). Naturally, the overarching costs involved with developing the right organisational frameworks, whether it be in terms of cultural, leadership, or communication factors, are significant financial risks businesses must consider. In the same way, the numerous costs of training, outsourcing, and recruiting AI-specialised labour significantly affect the ability of businesses to adopt AI solutions. Costs regarding the acquisition of the correct software and hardware required to run and store AI are also notable financial considerations. This array of costs associated with implementing AI, appearing at every step of the implementation process, limits the ability of businesses to strategically integrate AI solutions into their business models.

Computing limitations

It is no secret that AI models require extensive hardware power to train and run their heavy data-driven algorithms. This leads to a large demand for graphical processing units that help compute these models and support the vast number of threads that work cohesively and simultaneously to predict and create outcomes (Jeon, 2021). Millions of calculations need to be carried out in a matter of seconds in models like ANNs, where different weights and biases need to be included to find out the relative probability of a certain outcome. This computational power is granted by a process known as GPU acceleration. In GPU acceleration, CPUs are paired with the graphical processing power of a GPU to accelerate various computational processes (Merrit, 2021). GPU acceleration is especially important in the case of applications such as chatbots. The input database for AI chatbots is quite vast, so to select the appropriate data, carry out calculations, and give relevant, logical answers, firms use many interconnected GPUs. 

The pivotal issue with GPUs is their accessibility and cost. In recent years, as the implementation of AI rises and users invest thousands in processing power, demand has surged, resulting in prices skyrocketing (Griffeth, 2023). Suppliers were also unable to keep up with the demand for GPUs, which led to a lack of accessibility. Moreover, firms have to invest a lot of financial capital in the initial purchase of the GPUs to drive high-power processing. Although, in the long term, it helps firms improve their efficiency and overall productivity, leading to technical economies of scale.

Accuracy, Generalisation, and Transparency

Another important flaw of AI is the possibility of inaccurate predictions and results. The realm of AI and its various branches are still in the rudimentary stages of development. As models are being worked on and improved every day, it is possible that the current models lead to inaccurate predictions and outcomes. For example, in deep learning, during the training period, the algorithm needs exposure to more than 10,000 different example data sets to understand how an accurate and precise response is formulated (Zohuri and Moggaddam, 2020). Furthermore, the training data sets are often static, whereas the real-time data that the algorithms tend to deal with is much more dynamic and constantly evolving, reinforcing the extent to which the inaccuracy develops. AI also tends to lack a human’s common sense and logical approach. In the context of business, the AI may not be able to interpret consumer behaviour and qualitative inputs to the extent a human can. This may lead to incorrect forecasts of consumer demand, which may repel demand onto the supply side, exacerbating the bullwhip effect as explained earlier (Zohuri and Moggaddam, 2020). Furthermore, firms often release unsupervised ML models, such as Microsoft’s bot TAY (Thinking About You), but this bot saw a major flaw; it didn’t have the common logic to fact-check, so it ended up learning incorrect statements and other derogatory terms as well (McKendrick and Thurai, 2022). The lack of logic leads to incomplete arguments and predictions, which could adversely affect the forecasting abilities of a firm since the model is solely dependent on the data that is input into it. Thus, it is important for a firm to carefully consider the predictions and outputs of an AI model when utilising it.

Another limitation of AI models is the generalisation of their learning. This refers to the ability to carry over their knowledge and capabilities from one experience to another, which may have similarities but are distinct. This type of transfer learning is immensely useful in the areas of chatbots and marketing, as firms may require the AI model to use its knowledge of one market and apply it to another (Chui, Manyika, and Miremadi, 2018). However, the AI models have difficulty associating the new data with the one they have been trained on or learned from. So this again contributes to poor levels of accuracy. Oftentimes, the models also face difficulties in explaining how they have derived a certain outcome. This is not a new issue, and it persists, especially in the world of business. Firms may be keen on an explanation of the different forecasts of demand or the marketing decisions that the AI model has taken, but the lack of explainability does not allow firms to understand and record these outcomes for the basis of future decision-making. 

An effort that firms are implementing to overcome this challenge is a model known as local-interpretable-model-agnostic explanations (LIME). This model observes which parts of the input data have been used and to what extent, which permits firms to have a more concrete and refined explanation for the outcomes that they receive from AI models (Chui, Manyika, and Miremadi, 2018). Yet, in the current corporate scenario, many firms avoid using algorithms that create transparency, as it may lead to legal consequences as a result of the data and output being more public. The biases and anomalies in the data may be pointed out, so firms may not use such models to avoid risks.

Data limitations

Data issues, particularly the massive amount of data that is required, data bias, and the inaccuracy and untrustworthiness of qualitative data, are among the main technical limitations of AI. Starting with the need for large sums of data to hire suitable candidates for a particular job. It will be essential because if AI lacks adequate data detailing the precise methods its human creator used to manually recruit new employees before transitioning to AI, it won’t effectively recognise suitable candidates for the job (Harmon, A., no date). 

Then come data biases when using AI for recruitment. The first is sample bias, in which a certain type of population may be overstated or understated, not precisely overlaying the real layout of various populations around the world. Followed by algorithmic bias, which is more concerned with the depth of neural networks or previous information imperative for the algorithm. Representation bias is similar to sample bias, and it arises when irregular data is collected, or, in other words, it does not consider “outliers” or “anomalies”. It is also likely to occur when all the demographics of a particular population are not taken into account during the recruitment process. Lastly, measurement bias occurs when inaccurate outcomes are drawn during the development of the initial data set; this can lead to biassed results against various genders, races, and ages. (Team, I.D., and A. 2023). 

Since AI and its use for recruitment are becoming more prevalent, the impacts and consequences are also becoming increasingly widespread. For instance, via AI face recognition, a method recognised within AI circles for exhibiting accuracy biases across various ethnicities, this not only brings about legal and statutory risks but also puts business ventures at risk of losing capable and talented employees and is also likely to lead to a loss of capable employees (Elcock, S. 2023).

Data concerns aside, general AI bias refers to AI systems that produce biassed results that are inherited from their human creators, including current and historical social inequality. There are two types of AI biases: cognitive and algorithmic. Cognitive bias occurs because it is natural for humans to process a particular type of information and then form a judgement. These judgements creep into AI systems since they use our data to autonomously make their decisions, causing bias. Using flawed training data is usually the leading cause of algorithmic bias. For example, a data interpreter may consciously or unconsciously weigh in biassed factors that disrupt the algorithmic decision-making system (Team, I.D., and A. 2023).

Such biases bring about serious ethical concerns. For example, AI models used in recruitment have been shown to discriminate based on biases captured in their training data. An example of this caught sight in 2014 when Amazon launched an AI-driven recruitment tool that rated candidates from one to five stars. However, the programme was shut down next year as it was found to have been systematically downgrading women’s CVs, particularly for tech-oriented jobs such as software developers.

Although AI has proven to be capable of pulling off the task of recruitment quite successfully for simpler job responsibilities, the lack of human touch or “gut instinct” is a crucial aspect of hiring and identifying capable employees effectively and cannot be proven to be provided by AI. It may not also be able to detect whether or not a certain candidate will fit into the culture or the work environment, and is not capable of performing an in-depth review of a person. Also, since AI itself relies on human inputs and data, if the data or input fed is itself insufficient, biassed, or corrupted, it will defeat the purpose of having AI recruit new employees as it will make errors similar to those of standard human employers (Council, F.C. 2019).

Accountability and liability

In the case where the use of AI leads to harm or accidents, outlining the individual or organisation held accountable for such impacts and enforcing adequate redress are essential tasks that ethically and legally must be undertaken. In facing accountability, businesses are to comply with enforced standards, agreeing to the ethical and legal structures put in place. Actions relating to this agreement include the business’ actions being thoroughly recorded by the business itself and other agents, allowing for the business’ behaviour to be tracked and for evidence to be collected for enforcement processes. In agreement with these regulations, if businesses were to face enforcement as a result of their actions, they would incur sanctions, authorizations, or prohibitions depending on the evidence about their actions that is gathered (Novelli, C., Taddeo, M., and Floridi, L., 2023.). 

The allocation of accountability remains a difficult task today. In the current literature, the issue of whether the organisation creating the AI algorithm or the agents using it is responsible for AI-related harm is still a widely researched topic (Millar, J., et al. 2018). The European AI Act is an example of AI regulation, where businesses must follow guidelines including identifying the level of social risk their AI can have and declaring it, keeping transparent practices involving the use of data and the use of AI in customer service, and following laws that ban certain AI practices (European Commission, n.d.). Businesses must adapt to the different and changing regulations involving the use of AI, changing business processes to adhere to the laws established by authorities. Formally and informally, institutions supervise the use of AI, and businesses must execute ethical behaviour according to social and legal guidelines.


Artificial intelligence, or AI, is quite a broad term with countless implications in our everyday lives. However, in this research paper, we assessed AI’s role in business, its impacts, limitations, and ethical concerns. The first implication is demand forecasting, or more specifically, predictive analysis, which allows firms to efficiently predict future demand and supply variables. Followed by market basket analysis that analyses historical data from POS transactions to predict their product demand more accurately. Since these predictions are based on data fed into the algorithms, it is crucial to ensure the accuracy and precision of this data. 

Supply optimization is also a key part of any business, ensuring firms do not face overstock or understock and making inventory control easier, which may result in firms facing the problem of “overfitting”. Personalised marketing allows business ventures to enhance cross-selling and upselling prospects, but brings about ethical concerns regarding the storage of the personal data of various stakeholders. The use of automated customer service allows businesses to cut down on labour costs and reach wider audiences, but often businesses struggle with the execution of this IT project. The implications of AI surrounding business and innovation are also countless, including self-driving automobiles. A leading pioneer in the field is Tesla, which allows the company to make their vehicles safer, more comfortable, and more fuel efficient, but also brings about the horrors of cyber security concerns and unreliable driving leading to car accidents. 

AI has also revolutionised the process of recruitment, making processes such as data interpretation and job application processing much more expedient. However, ethical concerns are no exception here either, as AI recruitment may cause discrimination and racism-driven employment due to corrupted data. Organisational concerns and issues that businesses may need to address with AI implementation include communication of the change with employees, shifts in staff work activities, the need for strong leadership, and the costs associated with this AI implementation. 

AI does require extensive hardware to operate its heavy data-driven algorithms, which has resulted in a lack of GPU supply in recent times. Despite all that, there are issues with inaccurate predictions and results produced by AI, as well as the generalisation of their learnings, passing knowledge and data onto experiences and scenarios that are similar to one another. It is also crucial to outline the individuals and firms responsible for these consequences and for them to face the enforcement that has come as a result of their actions. As for accountability, whether or not an organisation creating a certain algorithm is responsible for the harm it causes is a matter of debate and a widely researched topic today. 

To conclude, there is no doubt that AI has revamped the landscape of how business ventures operate in today’s world, and there is little doubt that this fact will become the most powerful tool for these ventures to use in the times to come. However, these businesses ought to be cautious of the shortcomings of AI, especially considering that their algorithms are fed by data, provided by humans, which might be inaccurate and unreliable in almost any implication of AI in business.


Agouridis, A. (2023a) Ai in recruitment – everything you need to know, Jobylon. Available at: https://www.jobylon.com/blog/how-ai-is-transforming-the-world-of-recruitment#:~:text=AI%20allows%20recruiters%20to%20use,candidate%20success%20and%20cultural%20fit  (Accessed: 25 February 2024). 

Amira, V.Y. (2023) How tesla is using AI and Big Data Analytics in their self driving cars?, Medium. Available at: https://medium.com/dare-to-be-better/how-tesla-is-using-ai-and-big-data-analytics-in-their-self-driving-cars-7072e410c1b8 (Accessed: 25 February 2024). 

Bharadiya, J.P., 2023a. Machine learning and AI in business intelligence: Trends and opportunities. International Journal of Computer (IJC), 48(1), pp.123-134.

Bharadiya, J.P., 2023b. Artificial intelligence and the future of web 3.0: Opportunities and challenges ahead. American Journal of Computer Science and Technology, 6(2), pp.91-96.

Brown, S. (2021). Machine learning, explained. [online] MIT Sloan. Available at: https://mitsloan.mit.edu/ideas-made-to-matter/machine-learning-explained

Cederberg, L., Elliot, J., Engdahl, A., Hedenborg, F., Sikström, J., Wranå, L. and Åberg, J., 2020. Towards Better Demand Forecasting Using Artificial Intelligence.

Chui, M., Manyika, J. and Miremadi, M., 2018. What AI can and can’t do (yet) for your business. McKinsey Quarterly, 1(97-108), p.1.

Council, F.C. (2019a) Council post: 10 downsides of using artificial intelligence in the hiring process, Forbes. Available at: https://www.forbes.com/sites/forbescoachescouncil/2019/08/14/10-downsides-of-using-artificial-intelligence-in-the-hiring-process/?sh=3ffeb681685e (Accessed: 25 February 2024). 

Dwivedi, R., Anand, A., Johri, P., Banerji, A. and Gaur, N., 2020. Product Based Recommendation System On Amazon Data. Researchgate.

Elcock, S. (2023a) Data bias in AI: Causes, consequences and solutions, CEO.digital. Available at: https://ceo.digital/guest-articles/ai-data-bias-causes-consequences-solutions#:~:text=Organisations%20are%20investing%20heavily%20in,outcomes%2C%20and%20cause%20erroneous%20decisions (Accessed: 25 February 2024). 

European Commission. (n.d.). Regulatory framework for AI. [Online] Available at: https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai#:~:text=The%20AI%20act%20will%20be,%2C%20safety%2C%20and%20ethical%20principles.

Gomez-Uribe, C.A. & Hunt, N., 2015. The Netflix Recommender System: Algorithms, Business Value, and Innovation. ACM Transactions on Management Information Systems, 6(4), Article No.: 13, pp. 1–19. Available at: https://doi.org/10.1145/2843948

Griffith, E. (2023). The Desperate Hunt for the A.I. Boom’s Most Indispensable Prize. The New York Times. [online] 16 Aug. Available at: https://www.nytimes.com/2023/08/16/technology/ai-gpu-chips-shortage.html

Hayes, A. (2023) Autoregressive Integrated Moving Average (ARIMA) Prediction Model. Avaliable at: https://www.investopedia.com/terms/a/autoregressive-integrated-moving-average-arima.asp.

‌Helo, P. and Hao, Y., 2022. Artificial intelligence in operations management and supply chain management: An exploratory case study. Production Planning & Control, 33(16), pp.1573-1590.

Heyday by Hootsuite. 2021. How a Global Sport and Fitness Retailer Used Conversational AI to Grow Sales and Lower Costs. [online] Available at: https://heyday.hootsuite.com/customer-stories/decathlon-uk/ [Accessed: 19/02/2024].

IBM (2023). What is Machine Learning? [online] IBM. Available at: https://www.ibm.com/topics/machine-learning.

IBM (2023). What is Overfitting? | IBM. [online] www.ibm.com. Available at: https://www.ibm.com/topics/overfitting.

infotech (2024a) Latest News, Updates and Events. Available at: https://www.seasiainfotech.com/blog/how-amazon-uses-artificial-intelligence/ (Accessed: 25 February 2024). 

Jeon, W., Ko, G., Lee, J., Lee, H., Ha, D. and Ro, W.W., 2021. Deep learning with GPUs. In Advances in Computers (Vol. 122, pp. 167-215). Elsevier.

Kadasah, E.A., 2023. Artificial Intelligence Powered Chatbot for Business. International Journal of Information Technology and Business, 4(2), pp.61-66.

Kaur, M. and Kang, S., 2016. Market Basket Analysis: Identify the changing trends of market data using association rule mining. Procedia computer science, 85, pp.78-85.

Kim, J.B., 2019. Implementation of artificial intelligence system and traditional system: a comparative study. Journal of System and Management Sciences, 9(3), pp.135-146.

Kurnia, Y., Isharianto, Y., Giap, Y.C. and Hermawan, A., 2019, March. Study of application of data mining market basket analysis for knowing sales pattern (association of items) at the O! Fish restaurant using apriori algorithm. In Journal of Physics: Conference Series (Vol. 1175, No. 1, p. 012047). IOP Publishing.

Malil, A. (2023). AI Bias In Recruitment: Ethical Implications And Transparency . [online] Forbes. Available at: https://www-forbes-com.cdn.ampproject.org/v/s/www.forbes.com/sites/forbestechcouncil/2023/09/25/ai-bias-in-recruitment-ethical-implications-and-transparency/amp/?amp_gsa=1&amp_js_v=a9&usqp=mq331AQIUAKwASCAAgM%3D#amp_tf=From%20%251%24s&aoh=17080932715925&referrer=https%3A%2F%2Fwww.google.com&ampshare=https%3A%2F%2Fwww.forbes.com%2Fsites%2Fforbestechcouncil%2F2023%2F09%2F25%2Fai-bias-in-recruitment-ethical-implications-and-transparency%2F [Accessed 9 2023].

Mariciuc, D.F., 2023. A Bibliometric Analysis of Publications on Customer Service Chatbots. Management Dynamics in the Knowledge Economy, 11(1), pp.48-62.

McKendrick, J. and Thurai, A. (2022). AI Isn’t Ready to Make Unsupervised Decisions. [online] Harvard Business Review. Available at: https://hbr.org/2022/09/ai-isnt-ready-to-make-unsupervised-decisions.

Merritt, R. (2021). What Is Accelerated Computing? [online] NVIDIA Blog. Available at: https://blogs.nvidia.com/blog/what-is-accelerated-computing/.

Millar, J., Barron, B., Hori, C.D.K., Finlay, R., Kotsuki, C.K. and Kerr, I., 2018. Accountability in AI. Promoting greater social trust. CIFAR. Available at: https://cifar.ca/cifarnews/2018/12/06/accountability-in-ai-promoting-greater-social-trust.

Novelli, C., Taddeo, M. and Floridi, L., 2023. Accountability in artificial intelligence: what it is and how it works. AI & SOCIETY, pp.1-12.

Okorie, G.N., Egieya, Z.E., Ikwue, U., Udeh, C.A., Adaga, E.M., DaraOjimba, O.D. and Oriekhoe, O.I., 2024. LEVERAGING BIG DATA FOR PERSONALIZED MARKETING CAMPAIGNS: A REVIEW. International Journal of Management & Entrepreneurship Research, 6(1), pp.216-242.

Pierce Denning, M. (2023a) The impact of artificial intelligence on Business Operations, LinkedIn. Available at: https://www.linkedin.com/pulse/impact-artificial-intelligence-business-operations-pierce-denning-bs/ (Accessed: 25 February 2024).

Ponte Blanco, B., 2016. Bullwhip effect reduction through artificial intelligence-based techniques.

Reim, W., Åström, J. and Eriksson, O., 2020. Implementation of artificial intelligence (AI): a roadmap for business model innovation. AI, 1(2), p.11.

Rivas, P. & Zhao, L., 2023. Marketing with ChatGPT: Navigating the Ethical Terrain of GPT-Based Chatbot Technology. AI, 4(2), pp.375-384. Available at: https://doi.org/10.3390/ai4020019.

Sison, A., Ferrero, I., Pablo García Ruiz and Tae Wan Kim (2023). Editorial: Artificial intelligence (AI) ethics in business. Frontiers in Psychology, 14. Available at: https://doi.org/10.3389/fpsyg.2023.1258721.

‌SOCi. (n.d.). The Basic Components and Branches of AI. [online] Available at: https://www.meetsoci.com/resources/knowledge/localized-marketing/branches-of-artificial-intelligence/

Szilágyi, R. and Tóth, M., 2023. Use of LLM for SMEs, opportunities and challenges. Journal of Agricultural Informatics, 14(2).

Team, I.D. and A. (2023a) Shedding light on AI bias with real world examples, IBM Blog. Available at: https://www.ibm.com/blog/shedding-light-on-ai-bias-with-real-world-examples/ (Accessed: 25 February 2024).

Wamba-Taguimdje, S.L., Fosso Wamba, S., Kala Kamdjoug, J.R. and Tchatchouang Wanko, C.E., 2020. Influence of artificial intelligence (AI) on firm performance: the business value of AI-based transformation projects. Business Process Management Journal, 26(7), pp.1893-1924.

What are Neural Networks? | IBM (no date). Available at: https://www.ibm.com/topics/neural-networks.

Wierenga, B., 2010. Marketing and artificial intelligence: Great opportunities, reluctant partners. In Marketing intelligent systems using soft computing: Managerial and research applications (pp. 1-8). Berlin, Heidelberg: Springer Berlin Heidelberg.

www.linkedin.com. (2023). The 6 Branches Of Artificial Intelligence. [online] Available at: https://www.linkedin.com/pulse/6-branches-artificial-intelligence-paul-blocchi/.

Www.recruiter.com (no date a) Pros and cons of using AI in recruiting, Recruiter.com. Available at: https://www.recruiter.com/recruiting/pros-and-cons-of-using-ai-in-recruiting/ (Accessed: 25 February 2024). 

Ying, X., 2019, February. An overview of overfitting and its solutions. In Journal of physics: Conference series (Vol. 1168, p. 022022). IOP Publishing.

Zhang, J.J.Y., Følstad, A. & Bjørkli, C.A., 2023. Organisational Factors Affecting Successful Implementation of Chatbots for Customer Service. Journal of Internet Commerce, 22(1), pp.122-156. DOI: 10.1080/15332861.2021.1966723.

Zohuri, B. and Moghaddam, M., 2020. Deep learning limitations and flaws. Mod. Approaches Mater. Sci, 2, pp.241-250.