Delnaz Irani

ISO 31000 has set international standards for managing risks, it was developed in 2009 by International Organization for Standardization. ISO 31000 is tailor-made for any organization seeking clear guidance on risk management. FMEA is also a risk management process that helps the organization to identify all the possible risks and come up with a mitigation plan. So, integrating ISO 31000 and FMEA would provide a structured and comprehensive approach to risk management process. ISO 31000 FMEA ISO 31000 provides a broad, principle-based framework to identify risks, do the assessments and provide mitigation plan FMEA is a detailed, bottom-up approach which systematically identifies potential failure modes within a system. FMEA focuses on the occurrence, severity and detection of the failures. ISO 31000 encourages enterprise-wide risk assessment. This will ensure that risks from different sources such as technical, operational, security, compliance etc. are considered. FMEA will quantify the risks using a method called Risk Priority Number (RPN). This method helps us to prioritize mitigation efforts based on numerical values. Higher the RPN number (more than 120), means the risks are more. ISO 31000 will emphasize on a continuous risk management process. This will include communication, monitoring, and improvement. FMEA will focus on failure prediction and prevention. This will make it an effective tool for designing robust systems and processes. ISO 31000 identifies risks beyond technical failures, including regulatory, financial, and reputational risks. FMEA offers a structured way to analyze potential failures in software platforms, ensuring that critical risks are mitigated at an early stage. To explain the synergies between the two methodologies, I have considered an example from Software and Platform industry – Content platforms such as Amazon, Netflix, Spotify etc. use AI-based recommended engine, because it personalizes the user experience by analyzing the historical data of the customers, their preferences, and behavioral patterns to suggest relevant products, movies, or songs. However, there can be failures in the recommendation engine which would lead to poor user experience, revenue loss, and reputational damage. Let us see how by applying ISO 31000 we can identify and assess these risks. 1. First risk can be fairness risk or Algorithmic Bias risk - for an AI based recommended engine its observed that the model may over-recommend certain content, which can lead to a lack of diversity and user dissatisfaction. 2. Another risk can be performance degradation. As and when the users evolve, and their preferences change the AI model will degrade over time. 3. There can be an operational risk involved wherein if the response time is slow then it will lead to a poor user experience and increased churn rate. 4. If the data privacy laws such as GDPR, CCPA are not followed then it will lead to legal penalties for the organization. 5. Hackers can manipulate the recommended AI results by injecting fake information. By integrating FMEA here we can start analyzing and prioritizing these risks. Failure mode Effect Severity (S) Occurrence (O) Detection (D) RPN = S*O*D Mitigation plan fairness risk or Algorithmic Bias risk Users may move to another platform 8 6 7 336 Audits, ML based algorithms Performance degradation Outdated information will be provided 7 8 6 336 Real-time monitoring, automated retraining Operational risk Increase in the churn rate 9 5 5 225 Optimize algorithms Data privacy Legal penalties, loss of user trust 10 3 5 150 Data anonymization, consent management Cyber security Data manipulation or bot attacks 9 4 6 216 Fraud detection, anomaly detection Based on the ISO 31000 framework and FMEA analysis, mitigation strategies will get prioritized. Algorithm bias risk and Performance degradation have the highest RPN hence they are critical to get addressed. However, the severity of Data Privacy and Cyber security is highest, which makes them equally critical if they occur. By aligning ISO 31000’s continuous risk assessment cycle with FMEA’s structured failure analysis, the recommendation engine’s risks would get regularly monitored, reassessed, and mitigated. Hence this hybrid approach will ensure higher accuracy, fairness, security and compliance, improved user experience, engagement, and trust.

Lean is a process of continuous improvement techniques and activities used in manufacturing or service. “Digital Lean” is an advanced approach towards process improvement by combining the principles of Lean and Digital transformation technologies. The traditional Lean process focuses on eliminating waste, improving efficiency and optimizing workflows whereas Digital Lean will integrate all Digital tools such as Artificial Intelligence, Robotic process automation, Internet of Things and advanced analytics of Lean to enhance the process. If the organizations would leverage digital technologies, then they can gain real time insights, improved decision-making, more agile and efficient operational environment. Key concepts of Digital Lean would be identifying value streams, creating flow, utilizing a pull system, and eliminating waste through data-driven insights. · Value stream - Map out the entire process to eliminate non-value-adding activities. · Creating flow – Ensure smooth and uninterrupted workflow. · Pull system - Produce only what is needed and reduce the inventory waste. · Eliminating waste – Streamline the processes and identify the waste which can be eliminated through data driven insights. How Digital tools can enhance Lean when it is applied to these concepts – · RPA Automation – When automation is deployed it reduces the manual intervention and errors conducted by human. · Real-Time Data (IoT, Sensors, Digital Twins) – this will provide instant feedback and give real time update on the performance for process optimization. · Advanced Analytics & AI – AI will enable predictive maintenance and decision-making. · AI-driven Process Mining – AI driven process mining identifies any bottlenecks and inefficiencies dynamically. Some of the examples of Digital Lean implementations – Bosch and Siemens, the two leading manufacturers in home appliances and industrial equipment faced challenges in minimizing machine downtime and optimizing production workflows. Their manual inspections would lead to delays in detecting the defects and maintenance needs. These organizations felt the need to deploy IoT sensors and Digital twins. This helped them to continuously monitor the performance and detect any signs of wear and tear. The machines were now able to generate real time data which was used to create digital twins to simulate operations and predict failures if any. AI algorithms were implemented to analyze sensor data and predict if a machine would require maintenance. This would prevent any unexpected breakdowns. RPA was deployed for the quality checks to ensure seamless workflow and reduce human errors. The above digital tools resulted in - ü 15-20% increase in machine uptime, leading to higher production efficiency. ü 25-30% reduction in maintenance cost due to predictive analytics. ü 12-15% improvement in quality control by reducing defective products and rework. Another example is of Meta (Facebook) a renowned social media platform. Meta receives billions of user-generated contents which gets posted daily across Facebook, Instagram, and WhatsApp. The content moderation team needs to manually review the content and tag the violations, which was time-consuming and prone to errors. Meta developed AI models like DeepText and RoBERTa to automatically analyze the text, images, and videos uploaded on their site to detect policy violations such as bullying, hate speech, misinformation, violence, child exploitation etc. This had reduced the need for human content moderators to manually review the content every time it was flagged. Reviewers would now focus more on the complex cases. Incase if AI incorrectly flags the content, then the human reviewers would provide feedback back to AI. This ensures that the system is continuously learning and improves the accuracy of identifying violating content. By deploying automation, Meta was able to reduce manual moderation workload by almost 70 - 80%. Almost 90% of hate speech and harmful content was now removed automatically before the users would report it. Hence Digital Lean can help multiple industries such as manufacturing, healthcare, finance, logistics, or telecom etc. to achieve remarkable gains in productivity, fraudulent activities, cost savings, and process efficiency.

"Boundary spanning" refers to reaching out to different organizations, industries, departments etc. to share their knowledge, collaborate with each other and bring out innovative ideas. It can involve individuals or groups who act as “boundary spanner” connecting isolated information, ideas or expertise. This concept is important in today’s complex and interconnected business environments, where innovation, adaptability, and collaboration are critical for success. When individuals from different geographies and departments work on a project, they all come with their own experience and expertise, as well as perspectives. In such scenarios boundary spanning can lead to novel solutions and breakthrough innovations. It gives a platform to the team to access new tools, technologies, and expertise that may not exist with an individual. For example, deploying a knowledge repository tool which can be used by all the sites globally covering around 10K people. The app development team, Operations team and Business transformation team, tech team all come together and brainstorm on different ideas which would also facilitate open communication and builds trusts. Boundary spanning would allow the organizations to respond to challenges more effectively by leveraging diverse resources and knowledge. During covid many organizations had to adapt to the sudden change and broke the traditional way of working to ensure that the businesses survive and thrive. The collaboration between Apple and Nike is a good example of boundary spanning across industries of technology and sportswear. In 2006 the partnership was made by introducing Nike+iPod Sports Kit. This kit included a sensor so that the runners could track their performance data such as distance, speed, and calories burned. This data was then synced with the iPod, allowing users to monitor their fitness progress and listen to music simultaneously. In 2010 they moved up a level by introducing Nike+ Running App which was compatible with iPhones and iPads. The app allowed users to track runs, analyze data, and share progress on social media. In 2016 the partnership came up with another level - Nike+ Apple Watch Series 2. Specialized version of the Apple Watch - Apple Watch Nike+ with enhanced features such as custom watch face, personalized coaching, progress tracking etc. This partnership worked and had a huge success because of the collaboration between the two industry – technology and sportswear. Both these companies had a shared vision and aim to enhance the fitness experience by integrating technology into people’s lifestyles. They were focused on enhancing customer experience by introducing user-friendly products and apps that encouraged fitness and self-improvement. Hence, we can conclude by saying that “Boundary spanning” is a critical organizational capability in today’s interconnected world. By breaking down silos, it enables innovation, improves collaboration, and enhances adaptability.

Blockchain and LSS can be integrated as they both share a common goal of improving processes by enhancing efficiency, reliability and transparency. Let’s look at how blockchain and LSS can enhance the initiatives – Blockchain provides a decentralized ledger which ensures that all the transactions and the data received is recorded as well as accessible to all the relevant stakeholders. In a supply chain management domain blockchain will track the goods on real time basis allowing LSS practitioners to remove any barriers and ensure smooth processes. Blockchain facilitates a single source of truth, which helps in reducing conflicts between stakeholders and that enhances the collaboration during process optimization. If there is a multi-vendor project, then blockchain will synchronize the data and help LSS team to identify any areas of improvement. Data recorded in blockchain cannot be altered, which makes it more reliable. This helps for accurate process analysis and improvement. Blockchain can provide reliable data for statistical analysis, improving the accuracy of identifying defects or inefficiencies. Since the data recorded in Blockchain cannot be changed or deleted it simplifies audit process. If a healthcare organization uses blockchain then they can track the patient data to identify deviations from standard processes and maintain compliance with regulations. Blockchain can automate various manual processes. This aligns with the Lean principle of eliminating waste and improving efficiency. Challenges with blockchain initiatives – Small and medium size organizations might find it difficult to deploy blockchain as the initial set-up and maintenance is at a higher cost. If we are dealing with large volumes of data, then the blockchain networks can face issue with speed. The data protection laws such as GDPR might affect usage of blockchain as it needs to store the data and then use it for analysis. At time the organizations may face resistance from employees and stakeholders if they are unfamiliar with blockchain. This could hinder its adoption. Therefore, it’s important to provide training on both blockchain and LSS principles to ensure seamless adoption. Lot of expertise and in-depth knowledge is required to align blockchain technology with LSS processes and systems. Some of the successful examples who use both blockchain and LSS - a) Walmart uses blockchain to trace the journey of food items from farm to shelf. Combining it with LSS principles it helps the platform to improve the cycle time, enhance their customer experience and reduce errors. b) Pfizer pharma company uses blockchain to trace the authenticity of drugs across the supply chain. While blockchain provides transparency, LSS principles helps the organization to identify inefficiencies. c) De Beers is a diamond company which uses blockchain to track the diamonds from mine to retailors. They have integrated blockchain with LSS to reduce fraud and streamline their processes. We can conclude by saying that to implement blockchain with LSS successfully, it is important that the organization outlines how blockchain will address specific LSS objectives, such as reducing process waste, improving data quality, or enhancing transparency. By doing this the organizations can use LSS tools like Value Stream Mapping (VSM) or SIPOC diagrams to identify bottlenecks and inefficiencies and then use blockchain to add value.

The Black Box Paradox in AI arises from the trade-off between the high performance of complex AI models and the lack of transparency in how they arrive at their decisions. Businesses can balance leveraging these models while ensuring accountability and transparency through a combination of strategies, such as explainable AI (XAI), robust governance, regulatory compliance, and stakeholder engagement. Different strategies to address the Black Box Paradox: Adopting Explainable AI (XAI): Businesses should invest in tools and techniques that make AI models interpretable without compromising their performance. XAI approaches, such as SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model-agnostic Explanations), can highlight which features influenced an AI decision. This would help stakeholders to understand the rationale behind decisions. for example: Financial institutions like JPMorgan Chase use XAI to ensure transparency in credit scoring and fraud detection models. This helps regulators and customers understand why certain applications are approved or flagged. Robust AI Governance: Establishing clear governance frameworks for AI development and deployment is essential. This includes defining accountability, setting ethical AI principles, and regularly auditing models to detect biases and errors. for example: Google developed its AI Principles to ensure ethical development and accountability. The company evaluates its AI systems against these principles to minimize unintended consequences. Regulatory Compliance and Standards: Businesses must adhere to evolving regulations and standards, such as the EU’s AI Act, which requires high-risk AI systems to meet transparency and documentation requirements. Ensuring compliance builds trust with stakeholders and mitigates legal risks. for example: Healthcare providers deploying AI for diagnostics, such as IBM Watson Health, comply with stringent regulations like HIPAA to ensure the safety and transparency of AI recommendations. Human-in-the-Loop (HITL) Systems: Integrating human oversight in critical decision-making processes allows businesses to validate and interpret AI outputs, ensuring transparency and mitigating risks. for example: In autonomous driving, Tesla and Waymo incorporate HITL approaches during development and testing phases to monitor and validate AI decisions. Stakeholder Engagement and Communication: Educating stakeholders, including customers, regulators, and internal teams, about how AI models function fosters trust. Clear communication about model capabilities and limitations is vital. for example: E-commerce platforms like Amazon explain recommendation systems to users through statements such as "Customers who bought this also bought..." to build transparency and trust. Balancing the benefits of complex AI models with the need for transparency requires a multi-faceted approach. By integrating explainable AI, robust governance, regulatory compliance, human oversight, and transparent communication, businesses can navigate the Black Box Paradox effectively.

Polanyi’s Paradox states the challenge that AI would face in replicating tacit knowledge which means the knowledge that is deeply rooted in human intuition, experience, and context, which we often cannot fully articulate. Understanding this paradox provides a framework for identifying areas where human skills remain indispensable and for shaping strategies to integrate human and AI capabilities effectively. Below are the areas where human skills would remain indispensable - 1. Creativity and Innovation – While AI can generate ideas and help in innovation a lot of tasks like artistic expression, and lateral thinking, are difficult for AI to emulate. For Example: Designing new products, writing compelling stories, or creating innovative marketing strategies. Emotional Intelligence (EQ) - AI lacks the ability to genuinely empathize with individuals or understand human emotions. It can only provide language or the material however the main element of the human touch would be missing. For Example: Counseling, conflict resolution, and leadership roles requiring interpersonal skills. Complex Problem-Solving - Humans can approach problems with flexibility and make sense of ambiguous or incomplete data which is difficult with AI. For Example: Strategic decision-making, scientific discovery, or navigating ethical dilemmas. Ethics and Judgment – Human intervention will be required for decisions involving values, fairness, and societal norms that impact human oversight and moral reasoning. For Example: Legal judgments, policy-making, and medical ethics. Tacit Knowledge-Intensive Roles - Tasks involving implicit understanding, situational awareness, or physical dexterity remain challenging for AI. For Example: Skilled trades, caregiving, or cultural context-based work. Different strategies that individuals and organizations can adopt to thrive alongside AI rather than fear job displacement would be 1. To invest in continuous learning and upskilling themselves. 2. Organizations can offer training programs to help employees adapt to technological shifts. 3. We can leverage AI to do data analysis while humans can provide strategic insights and do the decision-making. 4. Organizations should encourage employees to build expertise in areas that AI struggles to replicate. 5. Leaders and organizations should promote a culture where employees view AI as a tool to enhance productivity and explore new opportunities rather than treat it as a threat. 6. Organizations should prioritize transparency and ethical considerations in deploying AI to build trust and ensure equitable outcomes.

Ensemble methods are machine learning techniques that combine multiple models to create a single, robust predictive model. Some ensemble methods focus on reducing bias by iteratively correcting errors made by previous models. Ensemble methods aggregate insights from diverse models, capturing more aspects of the data patterns. They provide stable predictions even in noisy data, improving robustness. Advantages of ensemble techniques in S&P domain - 1. It helps in enhancing the user behavior prediction. For example, the video streaming platform like Netflix or Hotstar can use ensemble technique such as Random Forest combined with gradient boosting to predict which users are likely to churn. It analyzes the user behavior such as their viewing history, time spent and skipped content. Basis on the results, the company can then come up with retention campaigns to ensure that the clients do not leave the platform. 2. Ensemble techniques will improve detection of any irregularities. For example, if there is an unexpected surge in unusual login attempts the ensemble model will raise a flag that there is a potential cyber security threat. This will help the platform to take preventive measures to address such risks. 3. Ensemble models like bagging and gradient boosting helps the platform to predict server load based on user activity trends. For example, during peak season the model will forecasts increased resource requirements and send a trigger to the cloud server. This ensures optimal performance by reducing operational costs. 4. Ensemble techniques will help in improving bug predictions and fix the issues. The model predicts which modules are likely to have bugs which will help the developers to prioritize their debugging efforts, leading to improve the product quality. Here are some of the limitations of ensemble techniques in business decision-making · Ensemble methods are computationally expensive which means that it will require significant resources for training and deployment. For example, training a large ensemble like a Random Forest on high-dimensional e-commerce data may need advanced hardware and significant time. · Some ensemble techniques, like Boosting, are prone to overfitting if not properly tuned. For example, overfitting in a model predicting stock prices might lead to unrealistic investment decisions.

When we analyze datasets using an LLM like ChatGPT, below are the potential pitfalls or errors that could occur 1. LLM can misinterpret the dataset if the excel file is not structured correctly, for example headers are missing or the format is not the same. 2. Duplicate datasets might not be identified automatically leading to skewed results. 3. If LLM is not able to identify the outliers, then it could result to incorrect results. 4. If LLM has not completely understood the context of the data or the specific research question, then it will give us inappropriate analysis or irrelevant conclusions. 5. At times LLM's explanations might not be detailed enough for users to fully understand the analysis. 6. LLM might not function appropriately if the dataset is too large and requires computations.

Any kind of a change is not initially accepted by the team. A Leader `s role is to minimize the resistance of the change by showcasing the impact that it will bring in for the team and the organization as a whole. When people understand the big "Why" i.e. Why the change is happening and how is it going to benefit everyone, the resistance will reduce to a large extent. The Leader needs to take practical steps before announcing the change, such as taking stakeholder feedback, be transparent and clarify roles and responsibilities to avoid any confusions or hindrances in the future, make them aware of the vision. Therefore, the focus of a Project leader should be on reducing or addressing the resistance.

Blackbelt training was enthralling and fruitful which cleared lot of my concepts. I am more confident to implement the learning in my work area and yield good results. Special thanks to the trainer Parag Mehta, who's training technique made a complex topic also simple. Thank you team Benchmark and Parag Sir. Regards Delnaz Irani Lead - Tata consultancy services limited

Application of our knowledge would matter irrespective of the Industry or size of the organization. You can be the game changer and take your organization to another level which they would have not thought of