Bias in Artificial Intelligence-Enabled ERP Software Customization: Say whaaat?!

Bias in Artificial Intelligence-Enabled ERP Customization

This article scratches the surface of, “Understanding algorithm bias in artificial intelligence-enabled ERP software customization” (Parthasarathy, S., & Padmapriya, S. T. (2023). Understanding algorithm bias in artificial intelligence-enabled ERP software customization. Journal of Ethics in Entrepreneurship and Technology, 3(2), 79-93.)

The authors share;

“The possibility of algorithmic prejudice during the ERP customization process has not, however, been studied to date.” (p. 90).

But, why does this matter? What is the value of this paper to ERP organizational change practitioners and researchers such as Nestell & Associates?  Why should organizations care about such a topic?

Please take the time to read the paper for yourself. But in the meantime, well, let’s explore together!

First, Why is ERP Customization Important to Understand?

Enterprise Resource Planning (ERP) systems are complex software solutions designed to integrate and manage core business processes such as finance, human resources, procurement, manufacturing, supply chain, and more. While ERP systems offer standardized modules to support these processes, businesses often find the need to customize them to better fit their unique requirements and workflows.

Understanding ERP customization is crucial for several reasons:

• It can ensure needed alignment with business processes
• Provide a competitive advantage
• Ensure that industry-specific requirements are met
• Improve end-user adoption
• Allow the organization to scale and adapt to accommodate growth
• Support Integration with other systems
• Lead to long-term cost savings by streamlining processes
• Allow businesses to address specific risks and challenges unique to their operations
• And on and on.

Understanding needed ERP customization empowers businesses to leverage their ERP investment fully, optimize operational efficiency, and stay agile in a dynamic business environment. However, it’s essential to approach customization strategically, balancing the benefits of tailoring the system with the potential complexities and costs involved. Therefore, as we will further discuss, understanding how artificial intelligence enabled ERP can impact your organization is important to understand.

The Next Question, Why Are ERP Customizations so Challenging and Risky?

ERP customizations can indeed be challenging and risky due to several factors, below are just some of the main considerations:

• Complexity of the ERP System: ERP systems are inherently complex, with numerous interconnected modules and functionalities. Customizing these systems often requires a deep understanding of the underlying architecture, data models, and business logic, which can be challenging to grasp, especially for non-technical users.
• Impact on System Stability and Performance: Modifying core functionalities of an ERP system can introduce instability and performance issues. Customizations may inadvertently disrupt existing processes, leading to system crashes, data corruption, or degraded performance. Ensuring that customizations are thoroughly tested and properly integrated with the ERP system is crucial to mitigate these risks.
• Upgrades and Maintenance: ERP systems typically receive regular updates and patches from the vendor to address security vulnerabilities, bug fixes, and new features. Customizations can complicate the upgrade process, as they may need to be re-implemented or modified to remain compatible with the updated version of the ERP software. Failure to properly manage customizations during upgrades can result in downtime, data loss, or functionality regression.
• Vendor Support and Compliance: ERP vendors may not provide full support for heavily customized systems, as customizations can deviate significantly from the vendor’s standard configuration. This lack of support can pose challenges when troubleshooting issues or seeking assistance from the vendor. Moreover, customizations may introduce compliance risks if they result in non-compliance with industry regulations or internal policies.
• Dependency on Skilled Resources: Successful ERP customizations require skilled resources with expertise in both the ERP system and the specific business requirements. However, such resources may be scarce or expensive to hire, leading to challenges in finding and retaining qualified personnel. Additionally, reliance on a small pool of experts can create knowledge silos and dependency risks within the organization.
• Scope Creep and Requirement Changes: Customization projects are prone to scope creep, where additional requirements are continuously added throughout the development process. Managing scope creep effectively requires clear communication, robust change management processes, and disciplined project governance. Failure to control scope creep can result in project delays, budget overruns, and dissatisfaction among stakeholders.
• Data Integrity and Integration Challenges: Customizations often involve manipulating and integrating data from various sources within the ERP system. Ensuring data integrity and consistency across customizations and standard modules can be challenging, especially when dealing with large datasets or complex data structures. Inadequate data validation and error handling mechanisms can lead to data discrepancies and compromise decision-making processes.
• Long-term Maintenance and Total Cost of Ownership: Customizations increase the total cost of ownership (TCO) of an ERP system by adding complexity to ongoing maintenance and support efforts. Over time, the accumulated maintenance and support costs associated with customizations can outweigh the initial benefits, especially if proper documentation and knowledge transfer practices are not in place.

While ERP customizations offer opportunities to tailor the system to specific business needs, they also entail significant challenges and risks that must be carefully managed to ensure project success and maximize the return on investment.

“Knowing that customization is inevitable, ERP adopters must strike a delicate balance between company value and risk when determining how much customization is necessary. If customization is in line with strategy and aids the adopter in achieving its business objectives, it is regarded in ERP literature as a value-adding activity in a project”(Parthasarathy, S., & Padmapriya, S. T., 2023).

What impact does or will artificial intelligence-enabled ERP have on these challenges and risks?

Let’s continue.

A Good Third Question: How do Organizations determine “Out-of-the-Box” versus Needed Functionality Gap?

One way is good ole’ fashioned analysis in which subject matter experts who understand the functionality of a given ERP platform perform a manual process in which they compare, discuss, and track the platform functionality to their knowledge of the business (we’ll discuss this more in upcoming blog posts). Another option for such an assessment is via the use of computer programs. With this article, we focus on the latter since that is what Parthasarathy, S., & Padmapriya, S. T. (2023) refer to in their work.

This discussion aims to share how ERP practitioners can comprehend the presence of algorithmic bias when employing algorithms (i.e. computerized tools) to assess the extent of required ERP customization. The utilization of AI in these algorithms is contingent upon the practices of ERP vendors and may differ between projects.

Moreover, in their article, Parthasarathy, S., & Padmapriya, S. T. (2023) have structured their study to tackle algorithmic bias in two different “computerized program” scenarios: those utilizing algorithms without AI and those supported by AI. The authors, “present the relevance of AI to ERP software solutions from a customization perspective”, “provide a brief overview of ERP software customization, followed by our research statement”, discuss  “the algorithmic bias during ERP software customization”, “draw a roadmap for managing algorithmic bias during ERP customization in practice”, and lastly, share their conclusion and provide thoughts on future research of this topic.

Let’s explore further artificial intelligence-enabled ERP!

Due to inadequacies, companies seek more advanced ERP solutions to scale and adapt to evolving business needs.

How Will Artificial Intelligence-Enabled ERP Change The ERP Landscape?

Artificial intelligence (AI) and machine learning (ML) advancements are prompting companies to overhaul their traditional ERP software. Today, customized systems are proving inadequate, leading companies to seek more advanced ERP solutions to scale and adapt to evolving business needs. However, the challenge lies in maintaining production consistency amid rapidly changing customer requirements during ERP implementation and beyond.

By integrating AI and ML, new ERP systems can adapt and improve over time by learning from patterns and behaviors. These next-generation ERP models are expected to be highly customizable and flexible, enhancing performance and usability across various organizational divisions. The emergence of self-learning algorithms is set to challenge established software developers in the industry. Innovative programming techniques are necessary to integrate big data generated by outdated ERP systems with AI and ML, enabling algorithms to accumulate knowledge rapidly. Despite the potential benefits, the strategic incorporation of AI into organizational management practices, particularly within ERP systems, remains relatively unexplored. Consequently, the ERP industry is witnessing a surge in AI-enabled software solutions.

Artificial intelligence-enabled ERP is also impacting customization support tools. ERP vendors are also exploring AI’s potential to streamline ERP customization options during implementation, reducing time and costs associated with aligning software with customer business requirements. This shift towards AI integration is expected to drive greater efficiency and innovation within ERP systems. These AI-enhanced customization tools may have a bias that could have a consequential impact on organizations.

What does “Bias” Mean?

In this text, “bias” refers to systematic inaccuracies or unfairness in computer algorithms or data handling processes. Algorithmic bias specifically pertains to recurring errors in algorithms that lead to errors in outcomes. This bias can result from the learning process of machine learning algorithms, where exaggerated claims or patterns lead to consistently biased findings. Similarly, data bias occurs when certain data sources or types are intentionally or unintentionally treated differently from others, potentially skewing the results or conclusions drawn from the data. Again, this “bias” could have a consequential organizational impact (time, money, and effort). As Parthasarathy and Padmapriya demonstrate, artificial intelligence-enabled ERP customization tools are certainly prone to bias.

Introduction and Definition: Prioritized Requirements Customization Estimation (PRCE) Algorithm

The Prioritized Requirements Customization Estimation (PRCE) algorithm is a method used for estimating and prioritizing customization requirements in information technology (IT) projects, particularly within the context of enterprise software development. Unlike algorithms utilizing artificial intelligence (AI) techniques, PRCE does not rely on AI but instead focuses on organizing and prioritizing customization needs based on predefined criteria. However, it’s noted that even though PRCE doesn’t employ AI techniques, it can still be influenced by biases stemming from its data inputs or from alterations made to the algorithm’s procedures at certain points during execution. This implies that the accuracy and fairness of the results produced by PRCE may be affected by biased data or by intentional modifications made to the algorithm during its implementation. PRCE serves as a systematic approach for managing and prioritizing customization requirements in IT projects, but it’s essential to be mindful of potential biases that may impact its effectiveness and fairness.

“The PRCE method aims to calculate the level of customization needed for ERP software to integrate seamlessly into a customer organization’s needs. It breaks down customer requirements into application, process, and design levels, assessing the degree of modification needed for ERP software.” (Parthasarathy, S., & Padmapriya, S. T. , 2023)

Introduction and Definition: KNN-ERP

In their work, the authors developed an AI version of PRCE using KNN. The “KNN-ERP” algorithm, short for k-nearest neighbours-enterprise resource planning, is an AI-driven approach developed to predict the level of customization needed for ERP software during its initial implementation phase. It utilizes the KNN machine learning algorithm to match the requirements of a new ERP project with those of previously completed projects in the same domain. By comparing features such as the number of required changes (ARs, PRs, DRs) with historical data, the algorithm estimates the degree of customization necessary for successful implementation. The process involves steps like data collection, feature selection, defining similarity metrics, determining the value of K (the number of nearest neighbors), training the model, and testing its performance. However, biases may affect the algorithm’s accuracy, stemming from issues like biased data collection, inappropriate feature selection, anonymity constraints, and biased similarity metrics.

The Main Take-A-Way: Interesting and Applicable Work

Parthasarathy, S., & Padmapriya, S. T. (2023) share in their work the need for comprehensive research to address bias in ERP customization algorithms (PRCE and KNN-ERP) and proposes a systematic approach to evaluate and manage bias.

• Comprehensive Research Need: Parthasarathy, S., & Padmapriya, S. T. (2023) emphasize the necessity for in-depth research to confront bias in ERP customization algorithms, specifically PRCE and KNN-ERP.
• Systematic Approach for Evaluation: They propose a methodical strategy to examine and manage bias, spotlighting the crucial role of understanding bias in both traditional and AI-enabled ERP customization algorithms.
• Focus on PRCE Algorithm: The focus is significantly placed on the PRCE algorithm, aiming to identify and address bias within it.
• Roadmap for Managing Bias: The paper offers a detailed plan for mitigating bias in ERP software customization, highlighting its importance not just for academic research but for practical application as well.
• Recommendations for Further Research: Suggestions are made for further studies to enhance the efficiency and fairness of these algorithms.
• Guidance for Practitioners: Practical advice is provided for ERP project managers on how to effectively manage bias and ensure equitable outcomes in customization efforts.
• General Applicability: Acknowledges that biases in AI-enabled ERP customization tools are part of a broader trend of bias in AI deployments.

Some Fun Research Notes to Ponder for Artificial Intelligence-Enabled ERP

• How could additional empirical evidence support the author’s work? Would further empirical testing or case studies further support the assessment of the real-world effectiveness and accuracy of these algorithms in predicting ERP customization levels?
• The theoretical concepts and insights into the algorithms are a needed part of research. Can you think of any concrete implementation examples or validation studies? That is, what are some real-world applications and testing of the algorithms that would be necessary to evaluate the practical utility and effectiveness of the framework, models, and concepts?
• As Nestell & Associates has shared and discussed consistently, each and every organization is different. Should practitioners be very cautious with any assumption of uniformity? To what extent will there be a uniform approach to ERP customization across different industries and organizations? ERP requirements and customization needs can vary significantly based on industry, organizational size, geographic location, and other factors. Will we ever have proposed algorithms that may capture this diversity adequately?
• How will we know the extent to which the organization’s capital (time, money, and effort) compares customization requirements analysis from manual versus computerized tools versus AI-enabled computerized tools? Is there an effective way to compare the ROI of the options? A comparative analysis would provide a better understanding of the advantages and limitations of the proposed approaches.
• From the author’s great work, how could ERP practitioners provide guidelines for implementing these “non-biased” strategies in practice?

The Benefit of the Article’s Insights to Practitioners

Awareness of algorithmic bias: The article highlights the presence of algorithmic bias in enterprise resource planning (ERP) customization algorithms. Organizations need to be aware of this bias to avoid making decisions based on inaccurate or unfair predictions of customization requirements. Again, it is all about the organizational capital (time, money, and effort) required of success and organizational performance!

Risk mitigation: By understanding the potential sources of bias in ERP customization algorithms, organizations can take proactive steps to mitigate these risks. This includes careful selection of input data, validation of algorithm outputs, and ongoing monitoring for bias.

Improving ERP implementation outcomes: By addressing algorithmic bias in ERP customization algorithms, organizations can improve the accuracy of predictions regarding customization needs. This can lead to more successful ERP implementations with better alignment between software capabilities and organizational requirements.

Enhancing decision-making: Accurate estimation of ERP customization requirements is crucial for making informed decisions about software selection, implementation strategies, and resource allocation. By considering the insights from this work, organizations can enhance their decision-making processes in ERP projects.

Continuous improvement: The article emphasizes the importance of ongoing validation and adjustment of ERP customization algorithms to address changing requirements and minimize bias. Organizations should adopt a mindset of continuous improvement in their ERP implementation practices to adapt to evolving needs and technologies.

Final Thought: Towards a Bias-Aware Future in ERP Customization

In embracing the future of AI-enabled ERP systems, the journey toward recognizing, addressing, and mitigating bias is both critical and continuous. The insights and recommendations presented offer a roadmap for navigating this complex landscape, urging a proactive stance towards bias that encompasses rigorous research, strategic implementation, and unwavering commitment to improvement. As the technology evolves, so too must our approaches to ensuring that ERP systems serve not just the operational needs but also the ethical standards of fairness and inclusivity. By fostering a culture of awareness and adaptation, organizations can unlock the full potential of AI-enabled ERP, paving the way for more equitable, efficient, and effective business processes.

You Might Also Like These Related Articles:

The Future of ERP: Intrinsically Linked to the Evolution of Artificial Intelligence

Intelligent ERP Systems: Artificial Intelligence and Its Impact on ERP Organizational Change

What do ERP Critical Success Factors and Artificial Neural Networks have in Common?

What is ERP Organizational Change Research? And, Why Should You Care?

Machine Learning in ERP: Envisioning the Future of Business Performance

Artificial Intelligence Research: What Do 85 Peer-reviewed Articles Say about AI in Information Systems?

“Why Machines Will Never Rule the World”-Exploring the Future of Artificial Intelligence with Jobst Landgrebe

Related Podcast Episodes from “The ERP Organizational Change Journal” Podcast:

Episode 85: Harnessing AI for ERP Innovation: Balancing Growth, Ethics, and Governance

Episode 89: AI’s Role in ERP and Organizational Evolution

Episode 38: How Artificial Intelligence is Transforming the ERP Systems