Giridarasanmugaraja

In the Analyze phase, to find out what's really breaking the process instead of fixing the noise from the process, 1. List down the entire process from start to end; it will help visualize the data in each section where it goes wrong in the process. In the given example, the process mapping is as follows: Sales Forecast Procurement Production/Order from supplier Items in Transit Items in Warehouse Stock 2. Root cause analysis techniques must be implemented, such as the 5 Whys technique and fishbone diagram, to find out the cause of the problem. By doing the 5 Whys technique, root cause analysis as, Problem statement: Sales are dropping because of stockouts. 1. Why? Not enough products on the shelf 2. Why? There is no item stock in the warehouse and other distribution centers. 3. Why? The procurement team did not make enough orders for this item. 4. Why? The order by the procurement team is based on the sales forecast, which is inaccurate. 5. Why? The forecast model by the team did not consider the seasonal promotions and order demand. Here the problem is not the delayed delivery; it is because of poor sales forecast planning not considering the seasonal days it follows the previous sales historical data. How to prevent the mistake of chasing the wrong cause? 1. Look into the data; check the actual numbers, not by guess. 2. Involve all the stakeholders in the process and discuss and analyze with them before concluding the solution to the problem. 3. Focus on the process, not the people.

In the given example scenario, tracking wait times in a busy hospital. In this case the problem statement to be defined, let's say, is "Customer dissatisfaction because of long wait time during the visit." The values that are easy to measure in the process, such as 1. Number of patients seen per hour states that total volume of patients 2. Time from check-in to discharge states total time of the patient in the hospital includes wait time, treatment time, test time, etc. To pick the meaningful number in the process, let's break down the total wait time of patients when they enter the hospital. 1. Time of arrival at reception 2. Patient Vital Test 3. Waiting room (to consult specialized doctor) 4. Pre-Consultation Time 5. Medical Test Time (if required) 6. Post-Consultation Time 7. Pharmacy Wait Time 8. Discharge Time By collecting the data in this format, it will show where the wait time gets increased in the routine cycle. It will be used to analyze and improve the process by reducing the possible wait times at each section. Keep monitoring the KPI and KRI in the process. KPIs such as average wait time to see a doctor, average wait time for a particular test and results, and number of patients who leave without being seen. KRI such as number of nurses/doctors on duty, lab turnaround times, and patient volume by hour. Tricks to catch bad data: 1. Ensure the person who is collecting the data is well trained and has a complete understanding of the process so none of the data will be missed during collection. A pilot test run of the complete process is to be done to validate it before implementing it in the live environment. 2. Use data validation logic at the time of data processing; for example, a patient's discharge time cannot be before their arrival time, and a patient's wait time can't be negative. 3. Randomly collect data and check its accuracy. 4. Analyze the data in the chart format to check the pattern that does not fit the process. (Usually the wait time of the patient to see the doctor is 20 to 40 minutes, but some of the data shows 200 minutes.) 5. Automate the data recording process and validate it from time to time. If any data entry is manual, train the staff to record accurately.

Scenario: The beverage company faces frequent stoppages, which lead to production loss and variation in the bottle quantity (overfilled or underfilled), which leads to increased production waste and customer complaints. To eliminate this problem and customer complaints, the operation improvements team approached line operators and maintenance technicians and discussed the problems and listed the causes, such as 1. The operator frequently identified that jams were caused by worn guide rails and sensor misalignment in the packaging line. 2. By listing out the machine health check parameters, they found that many bearings in the conveyor section are noisy and vibrating, which may lead to breakdown, and some of the filler nozzles are also clogged because the nozzles get dribbling and vibrating after a certain cycle of usage, which causes the sticky residue in the filling nozzle. 3. Maintenance technicians suggest increasing the lubrication cycles for the specific area where the wear and tear parts are high and changing the filler nozzle before it starts dribbling and servicing it in a timely manner. From the suggestions from the operators and technicians, they decided to implement the action points in the production line as 1. Major problems are caused because of the minor issues in the production line, so the operator gets well trained in machine maintenance, such as routine machine checklists, cleaning, and lubrication, and any minor alignment can be done by the line operators. This will reduce the downtime, and by training them, it can be fixed without maintenance team interference for the minor issues that they face during running production. 2. Preventive maintenance and replacement will be the solution to address this issue by developing a schedule for proactive replacement for high-wear parts like conveyor bearings and filler nozzles before they fail, based on operator/technician recommendations. 3. Increase the frequency of regular sensor calibration and replace the faulty ones, which will eliminate the wastage and improve the productivity. From this scenario, business excellence leverages the voice of the employee (operators and technicians) by getting feedback from them about the process issues, solutions to rectify the issues, and improvements to be done in the process to avoid the challenges in future operations. Challenges hinder gathering actionable employee insights; 1. Fear of speaking up—in some cases, they don't trust their boss or feel sometimes their honesty could lead to negative consequences. 2. Continuous collection of data or feedback from employees, but not using it. Strategies to overcome these challenges: 1. Build Trust: Use anonymous surveys or suggestion boxes so employees feel safe. When action is taken based on employee suggestion or input, reveal it to the employee and thank them; it will build employee confidence. 2. Regular Feedbacks on Action points: Once you get the feedback, assign someone to review and look into it and update the employee on a regular interval basis on what is being done with their feedback.

KPI: Key Performance Indicators, which indicate how the process is performing or performed. KRI: Key Risk Indicators, which indicate what will be the potential threat that could affect the process performance. For example, consider a machine that works 24/7 in a manufacturing industry. When the machine works continuously, it will get overheating issues, and a breakdown may happen. In this machine, we can set an alert or warning signal that if the oil temperature reaches above 70 degrees, then the maintenance team will attend to the machine and do the preventive maintenance instead of machine breakdown during the production time. It gives the planned downtime for preventive maintenance. In this case, KPI will give “total machine downtime in a month” where the downtime occurred because of the machine breakdown during the production. KRI is the “warning signal hydraulic oil temperature is high.” It will help to identify the machine breakdown that could happen in the process and prevent it from achieving the production performance. If we check the KPI, it would be late to prevent the machine breakdown. In another case, in a food manufacturing company, they rely on the packaging material to pack the products after the production. If the packaging material is not available, it can’t be packed and dispatched. If the lead time of the packaging material is one week, then we should set a KRI for inventory. “Reorder the material when the packaging materials serve for 10 days of production.” It will alert the purchase team, and they will follow up with existing vendors or new vendors and get the materials in time and avoid the production halt. In this case, KPI is the “number of days the production line was idle due to lack of packaging materials.” It shows only the machine stoppage because of the unavailability of packaging material. KRI is "Packaging material inventory days remaining in the warehouse.” The organization prioritizes the KPI over the KRI because It focuses on the results or success. It can be easily measured and explained to other stakeholders. It is an indicator that gives the actual performed results, e.g., production data per month and sales performance per month. KRI can be used to manage the process. In short, KPI is the performance outcome of the process where KRI is used in the process to identify the potential risk and allows you to prevent, plan and achieve the KPI.

BPR vs Lean Six Sigma When to use BPR, Complete transformation is required, it involves complete rethinking, remodelling, and redesigning the core processes starting from scratch. The incremental improvement in the ongoing process is ineffective and requires the change in the process. Example: A food manufacturing company wants to eliminate the paper-based report completely and implement digital real-time reports and a monitoring system using AI and automation. When to use Lean Six Sigma, It focuses on the improvement of the existing processes by eliminating the waste, variations, and defects. It is more about the better improvement in the ongoing processes. The current process is performing okay (85% effective) but needs improvements to make it more efficient (90 to 95% effective). Example: A food manufacturing company is trying to reduce waste and defects in a production line by implementing Six Sigma techniques. Both of them complement each other, because once the new process is created and implemented by business process reengineering, then Lean Six Sigma will play a role in the continuous improvement of the process, and it will eliminate emerging wastes, variations, and defects. In some cases the existing process cannot be improved by applying Lean Six Sigma techniques because of its ineffectiveness; then BPR will be initiated and implemented by using the historical data analysis from the Lean Six Sigma in the process.

I will take as an example of the procurement report (weekly/monthly). Why I can say that report is efficient because of the following reasons mentioned below: It is an auto-generated report that pulls the data from the system and sends an automatic email to the team (weekly). The report comprises the vendor details, item description, category or section, cost (material, freight & duties), order status, and timeline to reach the factory, etc. This allows team to track the purchase list and plan the work (maintenance) at the factory. The report is generated monthly in a standardized format with consistent charts and tables, which makes it easy to visualize at a glance. The report will give real-time visibility to the report because it is integrated with Power BI, so anyone can access and view the report. From a business excellence perspective, this seems ineffective because it focuses on reporting data rather than driving strategic actions and value optimization. Lack of Actionable Insights: While the shows you that what was spent and where, it rarely explains why the purchase occurred or what opportunities exist for optimization. For example, it might show a high spend in the "machine category or particular machine section," but it won't show if that purchase was strategic, redundant, or if there were better alternative options. Hiding Inefficiencies and Poor Practices: A high expense shown for "emergency breakdown machine parts that we bought and also items shipped by air cargo (air freight cost higher than sea freight)" in the report, but the report doesn't give an idea why these purchases occur and what the manufacturing loss will be if these items do not arrive on time to the stakeholders (e.g., poor planning, lead time, vendor issues, etc.). Supplier Performance Valuation: The report focuses on items bought in different "categories and section" and how much the "supplier contributes in overall purchases", not on supplier performance or contract compliance. It won't highlight if a low-cost supplier is consistently delivering the material late or with quality issues, which leads to manufacturing downtime & production loss. Focus on Cost, Not on Value: The report highlights the expenses, it doesn't show what value was received for that purchase. Higher expenses in a certain category might be justified to the stakeholders if they lead to significant improvements in productivity and quality.

Scenario Garden AI agent, which helps a gardener in a smart garden to manage the plants, ensuring the growth, preventing common issues of over/under watering, and pest control. Tone of responses: The AI agent should use a clear, gentle tone but not robotic and condescending responses. For example Watering the Plants: To water the plant (using a soil moisture sensor), instead of "Water now," the AI agent would say, "It is time for watering the plants, the soil moisture shows they are ready for a drink. Gentle watering now make them happy." Infected Plants: Plants affected by aphids (using an aphid sensor) instead of "Plants are infected," the AI agent would say, "It looks like your plant has a few aphids, don't worry & it is quite common, and I can guide you with a solution." Response time: The AI agent should give immediate alerts for urgent needs, regular updates, and instant advice on plants. For example If the soil sensor detects the dryness, it will give immediate notification as "Tomato plant needs immediate watering now to prevent wilting." Feedback style: The AI agent will give the primary feedback on the plant's health (using the smart sensor data), watering trends, and reminders. For example, In the application it shows the watering trends for all the plants, based on the plant type and watering frequency to maintain the healthy growth AI agent will remind us, "The plant loves to be fed every four days, the next feeding is due tomorrow." Handling errors: The AI agent detects the problem, diagnoses the cause, and shares the solutions then it will explain why the problem occurs. Sometimes AI may misidentify the problem, then the user will correct it, and the AI knowledge base will get updated. For example, AI agent gives the message "plant dying" because of leaf discoloration instead, it can say, "It looks like the plant leaves are turning yellow, it indicates a lack of nutrients. I recommend the organic fertilizer, which is rich in nitrogen. Can I show you how to use it?" If AI misunderstands the situation of leaf discoloration, the user gives feedback as "No, it is not a problem; this is normal discoloration from the sun," then AI will reply as "My apologies, thanks for the clarification, I will update my understanding for this plant type." which is polite and takes the user feedback to update the knowledge base. Personalization: AI agents should be stored with specific care profiles for each plant type, e.g., flowers, vegetable plants, herbs, etc. They will advise the user based on their gardening experience beginners or experts. It will also allow users to keep a personalized journal of their plant progress. For example, an AI agent will explain how to prune for the gardeners (beginners), but for the experienced gardeners, it will simply suggest, "Prune tomato plants next week," and it will give advice for specific plants. Users can add the notes and photos to a digital plant journal. Additional feature that I like to add Transparency and Explainability Users need to understand why the AI agent is sharing the recommendation. This will build trust and help the gardener learn and develop their own skills. For example, The AI agent says, "Consider moving the plant to a brighter spot" (using the light sensor). Then the user asks, "Why so?" Then the AI should respond, "Its leaves are stretching, which is a sign of not getting enough sunlight to grow." User Control and Flexibility Ultimately, a gardener is an expert for their own plant and garden maintenance & care. An AI agent is a helper; a user must always be able to easily dismiss the suggestions and allow manual log actions and input their own observations that contradict the AI agent.

AI Agent: Customer Support Agent Scenario: A customer inquiry about the feature of the product that he/she wants to buy: the user wants to confirm whether this product has this feature or not. The e-commerce platform provides a customer service agent which will clarify the customer's query regarding the features of the product. Product: Mixer Grinder Customer question: Does the mixer grinder model A210 have a self-cleaning feature? Hallucination: The AI agent replies, "Yes, the mixer grinder model A210 is equipped with the self-cleaning feature after the grinding by pressing the self-cleaning button in the mixer grinder." Actually this model doesn’t have self-cleaning feature and there is no button provided in the mixer grinder. Consequences: The customer might buy the product expecting that it have that feature of self-cleaning in the model. When the customer receives the product, finds the missing feature in the model, and is disappointed with the product, it leads to the return of the product and a negative review. The customer will lose trust in the company's AI agent and also the brand. To make it better, Prompt: When a customer asks about the product feature of any model, always refer to the official product specification mentioned in the product database or website. Do not invent the features, if a requested feature is not explicitly listed in the product data or specifications, respond to the customer as "The enquired feature is not listed in the specification for this model." Flow Logic: Knowledge Base restriction The AI agent should look up in the database, especially for the queries related to product specifications where it is listed as model, feature list, and specifications. Product verification If the feature is found in the database, then the AI agent should reply, "Yes, the feature is available." If not, "This model doesn't have this feature." It should not create any false positive response to the customer when the feature is not available in the database. "Was this information helpful about this Mixer Grinder A210?" This kind of small follow-up message will allow customers to correct if they suspect an error during conversation. Control The AI agent should be integrated with and primarily restricted to the official product information database. There should be limitations on the LLM's ability to generate information based on the specific products; the output should be derived from the verified product data in the knowledge base.

Scenario AI Agent: Restaurant Online Order Taking Agent Current practice In a famous busy restaurant, usually the customer calls the restaurant for an online order, then the responsible person will take an order, and he/she will write it down and pass the order. In some cases they might mishear what the customer's order is exactly or forget to place the order in the kitchen. Also if a foreigner knows about this famous restaurant, when they want to make an order, the person can't interact with them properly because of different languages. If they are able to communicate in their own language, they might be more comfortable understanding the dish and other details. Future AI Agent When you call the restaurant, the AI agent will answer the call without any urgency in tone. It will clearly recognize your voice and repeat your order back to you for confirmation when you say it. It automatically records the item in the system, which will never miss passing the order to the kitchen section, and it can also suggest some special dish of the day or special dish of the restaurant or any other items that you may like based on the past history of orders. AI Agent will be trained with all different languages, and then it will be easy for any customers to interact instantly without any translation app. It will make people more comfortable, and AI agents can explain the history or authenticity of the dishes to customers, and they can understand it easily, and they will try new dishes instead of their routine ones. Risk AI agents should be trained in all languages to communicate with people in both written and spoken ways. It will be challenging to train the AI agent in all languages because spoken slang and accents will be different depending on the different regions of people; it may produce awkward or incorrect translations, leading to misunderstandings. AI agents are trained to understand a wide variety of accents, speech patterns, local slang, etc., to avoid mistakes in understanding and translation.

AI Agent - Schedules appointment for a service. Scenario: A customer trying to book an appointment past one month, the system shows no availability for next two weeks. AI agent detect that he/she is a regular customer book for the service and also cant able to get the appointment in all the past bookings. Conflict: According to the rules, Scheduling agent strictly shows the available slots and inform the customer that there are no slots available. In this case customer can feel frustrated, especially if they have an urgent need. Approach: Scheduling agent should identify the High value or Loyal customers and prioritize interactions with them. Agent can be pre programmed to allocate the slots by checking for cancellation slots or any other special slots (limited duration slot) which may help for the loyal customer. In case of no slots are available, AI agent can add the loyal customer to the priority waiting list and inform them that you are in priority waiting list, if slot opens up they will get prioritized. Limitation: Scheduling agent should not able to overbook the appointment or scheduling conflicts for a loyal customers. Scheduling agent should not guarantee for the appointment, it can only offer to check for possibilities.