Mohammed Jaffer

Scenario on the Downtime and maintenance Traditional approach Record of the downtime manually and longer analysis. Preventive Maintenance standard scheduling of the machinery based on the OEM recommendation. Both these metrics are outdated misleading due to human based errors, delays, biased decisions Preventive maintenance not done based actual machine conditions. Two approaches AI Predictive maintenance - Maintenance carried out based on the actual machine condition by monitor power, output etc. which provides machine maintenance accuracy and Reduces time and increase equipment life. Automated downtime report - Sensors based reporting on the machine provides real time data, flow or pattern of the downtime and classification of downtime which provides Real time data, No loss of data and identifying the root causes in patterns. The same AI Approach can used quality control data, Regulatory compliance of fortification, generator data & maintenance etc.. Particulars Traditional & outdated AI Quality control data Fixed quality control data Real time product checks - Tracks flour quality fortification Manual measurement Real time flow control- Tracks flour and fortifies with equivalent premix Generator Preventative maintenance Predictive maintenance - Engine oil conditions, temperatures Generator Log sheet data Real time data flow and checks and Predictive failure alarms & stoppages

Flour Bagging Variance - I would share my thoughts on my first DMAIC project back in 2014. Problem statement As a company, we are experiencing a problem with underweight flour packaging in 50 kilograms flour bags. Due to recent customer complaints in underweight in 50 kilograms flour bags. Since the problem has existed for at least two months. It has reduced customer orders which directly affected the profit of the company hence, the Six Sigma project is carried out to reduce the weight variance in flour bagging. When we started analyzing the historical data. The percentage of defects was 49%. It was 10 % underweight and 39% overweight and the problem are due to high standard deviation of 0.150 kilograms. To achieve the 3-sigma limit in flour bagging with the overall defective percentage of 6.5%. 3.25% of defective on underweight and 3.25% of defective on overweight. Which requires standard deviation to be 0.054 kilograms. Since we have achieved the standard deviation 0.050 kilograms in one of the two scales. Our achievable target is 0.050 kilograms. Through Z entitlement our target is 0.050 kilograms. Improvement Method - traditional approach (Daily checks, Weekly checks, Monthly checks) - Calibration - Schedule, levelling, cleaning, load cell voltage checks and zeroing. - Sealing rings - Flour escape containment checks, Compressor pressure and Piston condition - Bag conditions - Bag length, width, weight and type. - Weight checks - 60 bags per truck Traditional improvement Vs Reimagination AI. - DMAIC approach traditionally reactive means problem occurred, solution unknown DMAIC process used. - Facing issue in real time problems issue in high-speed operations. - Customer demands high tolerance requires 50 kg because of batch process example noodle requires 350 kg per batch that's 7 bags. weight variance will have recipe issues in terms output volume and excessive consumption materials other than flour Re-Imagined AI in Flour bagging - Real time weight record and cumulative weight of bags and truck weights can be monitored through sensors and weight checkers and automation software interlinking weighbridge with scale data. - Theft protection. - Adaptive tolerance errors and alerts. - Operator dependency for checks is limited. - Prevention of weight issue and customer complaints due to Proactive action. Should this process be improved further or is it time to reimagine it entirely using AI? Flour milling operations demand of High-speed operations, Minimal tolerance due to increased consumption & Sales. - Traditional methods are good process improved further still will be reactive and still dependent on Manpower. - AI on other hand can produce Real time monitoring, automated alerts on weight control and to maintain compliance on 50 kg bag weights and to increase customer loyalty. This Process - it must be reimagined entirely using AI is necessary increase customer satisfaction, long term success and be competitive in the market.

At a large flour milling company in South Korea, with multiple mills across Busan and Incheon, the leadership team had a strong Lean culture, but they wanted to better connect frontline employee feedback (VoE) to their Business Excellence program. Operators had been mentioning issues related to frequent stoppages at pneumatic conveyors that transported flour from roller mills to packing silos. However, their comments were typically informal, mentioned during shift meetings or captured inconsistently in maintenance logs. The Business Excellence team launched an initiative to formalize Voice of Employee (VoE) capture using a digital suggestion platform combined with AI-powered text analytics. Over six weeks, the AI tool analyzed thousands of free-text comments from operators, technicians, and maintenance staff across five mills. The AI identified that over 40% of employee concerns referenced "air pressure drops," "line chokes," and "product carryover"—an emerging systemic issue. Based on this VoE insight, the LSS team chartered a DMAIC project focused on reducing conveyor downtime. In the Measure phase, they combined manual observations with AI-based real-time monitoring of compressed air systems and ML-driven predictive analytics to analyze historical data on air flow, temperature, and line pressure fluctuations. The Analyze phase revealed two key findings: Incorrect setting of valve timings during shift changes led to inefficient air pulsing and blockages. Micro-leaks in older pneumatic pipes caused unstable pressure levels, particularly during peak shifts. In the Improve phase, they implemented AI-driven predictive maintenance for air compressors and piping and used AI to simulate optimized valve sequencing. Operators were trained in the new standard work, and VoE collection remained active during the rollout to gather real-time feedback. Results: Line stoppages due to pneumatic failures reduced by 72% Energy consumption of air compressors reduced by 18% Operator satisfaction increased, as the project visibly addressed their top concern. The Control phase included an AI-powered dashboard showing air system performance and flagging anomalies early keeping operators and management aligned.

In Flour Industry Why KPI are chosen over KRI's Simplicity and familiarity: KPIs are easier to define, measure, and link to business goals. Tangible outputs: KPIs track results such as production output or yield that leadership can easily interpret. Short-term focus: KPIs reflect current performance, which appeals more to short-term operational targets. Data availability: KPI data is often more readily available, while KRI data may require predictive models or root cause analysis. Limitations of KRI Complexity in identification: Choosing the right KRIs for each process is not always straightforward. Lag in actionability: KRIs may highlight potential issues but don’t always suggest immediate solutions. Overreliance on historical data: Some KRIs are based on trends, which may not predict all types of risk. May require cultural shift: Teams focused only on output may ignore warnings until they materialize. Process Area KPI Example KRI Example Milling Efficiency Extraction Rate (%) Wheat Moisture Variation Maintenance Downtime (hrs/month) Number of Delayed Preventive Maintenance Tasks Quality Assurance Batch Rejection Rate (%) Gluten Index Fluctuation HSE LTIFR (Lost Time Injury Frequency Rate) Number of Dust Explosivity Alarms Supply Chain On-Time Delivery Rate (%) Vendor Lead Time Deviations Energy Management Power Consumption per Ton Generator Efficiency Drop Table shows only the possible KRI for the Milling efficiency there are multiple factors which enhance milling efficiency like Flow segration, Roll settings from millers, type of wheat itself (Harder the better milling performance) etc..

As an industry leader in Nigeria’s flour milling sector and a certified Lean Six Sigma Black Belt, I constantly assess how best to respond to both persistent inefficiencies and sudden operational crises. A recent case illustrates this well: we encountered a sand-bite issue in our flour, traced to excessive mud content in the raw wheat. This posed a serious quality and safety concern, prompting us to halt production and completely re-engineer specific milling and cleaning sections to eliminate sand contamination. This reengineering effort was disruptive and led to a temporary drop in yield. A classic BPR scenario, where a fundamental process redesign is necessary despite short-term losses. Once the new configuration was in place, we switched back to Lean Six Sigma methodology to restore performance. Through root cause analysis, data tracking, and process optimization, we incrementally brought the extraction rate back to optimal levels while maintaining improved flour quality. This integration of BPR and LSS allowed us to achieve both immediate remediation and long-term efficiency demonstrating that the two approaches are not only compatible but complementary. BPR helped us redesign the system to meet new standards, and LSS ensured its continuous optimization. In Nigeria’s complex flour industry, the ability to lead with flexibility knowing when to radically transform and when to systematically improve is essential for operational excellence and sustainable growth.

Mercury Buhler 2025: A User-Friendly AI for Wheat Milling Operations ( On Board live working sample) Intuitive, Conversational Tone Mercury Buhler 2025 uses clear and friendly language that feels like talking to a helpful coworker. Instead of dry, robotic alerts like “Temperature sensor error,” it would say, “Hey, I noticed the roller mill temperature is higher than usual. Let’s take a look to avoid any damage.” This approach reduces stress and encourages quick, confident action from operators. Fast and Context-Aware Responses The AI delivers near-instant responses, prioritizing the most relevant information based on what the operator is currently working on. For example, if an operator is checking the flour’s moisture levels, Mercury Buhler 2025 quickly provides moisture trends, recent adjustments, and helpful suggestions without the need for extra commands, making the workflow smooth and efficient. Multi-Modal Feedback Mercury Buhler 2025 combines voice notifications with visual alerts on control panels. While operators monitor machines, a gentle chime accompanied by a popup message might say, “Moisture level in Bin 206 is 15% target is 14%. Consider reducing for 10 more minutes or to increase the storage time by 4 hours or blend with low moisture wheat.” This multi-sensory approach allows operators to stay informed without losing focus on their tasks. Proactive Error Handling with Guidance When something goes wrong, Mercury Buhler 2025 doesn’t just report the problem. it offers clear, step-by-step guidance. For instance, if vibration sensors detect unusual activity, the AI would alert, “Vibration levels on the sifter are high. Please check for loose bolts or material blockages. Need a checklist?” This hands-on help reduces downtime and supports quick problem-solving. Personalized User Experience The AI learns each operator’s preferences and experience level. For new operators, Mercury Buhler 2025 might remind them, “Don’t forget your daily safety inspection checklist!” Meanwhile, for experienced operators, it could suggest, “Try adjusting roller speed by 5 RPM for better throughput.” This personalized touch helps keep everyone engaged and productive. Training and Continuous Improvement Mercury Buhler 2025 also supports ongoing learning by offering quick tutorials and updates on demand. If an operator seems unsure about a process, the AI might ask, “Would you like a quick guide on optimizing flour blend ratios?” This feature encourages continuous skill development without interrupting work. These design choices reduce cognitive load by giving operators exactly what they need, when they need it, in a clear and friendly manner. Quick, actionable insights help maintain smooth operations, while the AI’s supportive tone builds trust and confidence. Designing Mercury Buhler 2025 this way ensures technology works with people, making it a reliable and helpful teammate on the plant floor. Buhler Mercury system.pdf

Again, I like to use flour Mill as an example - Hallucination in Flour Milling Operations A factory manager uses an AI-powered assistant to optimize wheat blending ratios for specific flour quality targets (example) for pasta vs. noodles). The LLM, based on prior conversations or generic knowledge, suggests: "Use 70% hard red winter wheat and 30% soft red wheat to meet noodle flour specifications in Nigeria." However, this blending recommendation is hallucinated — it doesn’t match the actual wheat properties available in the silos or the target ash and protein content needed. Following this advice could lead to: •Off-spec flour impacting product quality •Rejection from QC or downstream customers •Increased rework or wastage •Loss of trust in AI system recommendations Steps to Prevent or Correct Hallucination 1. Prompt Engineering Use prompts that anchor the AI to local, real-time data: Based on current silo stock wheat types and lab-tested specs, suggest a blend for noodle flour with 12.5% protein and 0.65% ash. 2. Flow Logic Safeguards •Require the LLM to reference lab test results or ERP data before answering. •Set rules to stop AI from generating recommendations without access to validated data. 3. System Design Improvements •Integrate the LLM with a wheat spec database or lab interface (RAG-based system). •Add warnings or uncertainty tags when confidence is low or input data is missing. •Enable a review mode, where a miller can verify and adjust the recommendation before implementing. By embedding AI into the real-time process flow with verified data and human checks, flour milling businesses can leverage LLMs effectively while minimizing operational risk from hallucinations.

Simple example on Compliance Monitoring Report: Vitamin A Date: May 15, 2025 Prepared by:Compliance Team Scenario A production manager at XYZ group drafted an email to the warehouse/ store team: To issue 50kg of vitamin A premix from Supplier X (expired premix) for Batch - 45. Skip lab testing to meet the delivery deadline. AI Detection & Intervention 1. Email Scan (Outlook Integration): - Flagged phrases: “expired premix” and “skip lab testing.” - AI Action: Blocked email send and triggered a private reply to the manager: - Compliance Alert: Supplier X’s certification expired on 8/2024 (Orion Rule §3.7). Lab testing mandatory for Batch - 45 2. Excel Log Update: - Manager entered “Batch - 45: 50kg premix into the Excel tracker. - AI Action: Highlighted the cell in red and added a comment: - Error: Premix supplier not approved. Use Supplier Y (certified until 12/2025) Required dosage: 2kg/100kg flour = 600 ppm. 3. Orion Sync: - AI cross-referenced Supplier Y’s certification in Orion’s database. - Auto-generated a purchase requisition for compliant premix. Resolution- - Procurement:Switched to Supplier Y (approved vendor). - Testing: Lab confirmed Batch - 45’s vitamin A level: 610 ppm (compliant). - Audit Trail: - Updated Excel log: “Batch 45: 2kg premix/100kg flour → 610 ppm. - Orion marked status as “Compliant” and archived email alerts. Outcome- - Risk Avoided: $7,500 fine for using expired-certification suppliers. - Time Saved: 8 hours (avoided rework from a non-compliant batch). - Compliance: Met 600 ppm mandate with documented proof for regulators.

9. Imagine a process where the AI agent must balance two or more objectives — for example, minimizing response time vs. maximizing customer satisfaction, or sticking to process rules vs. delighting a VIP client. Describe one such situation and explain how you would guide the AI to handle the trade-off. What logic, rules, or signals would help it make the right call? Situation Case Flight EK 380 has a case scenario where AI should decide to opt one of two passengers to fly. One is the Platinum tier(Highest of all tier) passenger who exactly has arrived one hour prior the flight and Other is passenger who is a blue tier (Lowest of all tier) member who travels in economy class once a year but customer is seriously ill he should travel immediately with doctors recommendation.Unfortunately economy class is completely filled an hour before due to festive season. AI is asked to tactical situation. Conflict - Stick with process vs delighting a vip client. Signals - Key factors Platinum tier passenger To be upgraded immediately to business class most of the time of his travel. Loyalty/ Revennue - Generators $10,000 a year and always fly the same airline Present booking - Economy Blue tier passenger Passenger who serious ill required medical attention Loyalty/ Revennue - Fly the airline at least once a year and $1000 a year Present booking - Economy Operational Parameters - Flight starts 60 min decision to be taken the soonest - Security check and reaching the boarding gate takes 45 mins - Next flight 8 hours later - Compensation to Blue tier $300 dollars and next flight booking - Compensation to Platinum tier $1000 dollars and unhappy customer. AI Rules - Emergencies should be given priority over VIP clients. - Cost of compensation to the airlines. - Loyalty status of the customer AI Decision - Upgrade blue tier passenger with medical emergencies to priortize life risk to upkeep airline PR & advert status which can bring more new businesses to airline. -Platinum tier passenger given a ticket for next flight with free business class upgrade without detecting any points. And offer free lounge access with spa services for the long wait/ lounge access with $200 additional voucher. - AI decision summary points - Platinum tier passenger still satisfied with additional rewards given by the airline and awaits for next flight. Airline makes sure platinum tier passenger is satisfied and loyalty program member is maintained throughout the year. - Blue tier passenger understand the airline compassion and policy towards life risk issues than loyalty or revenue. -both decision are understood by the customer and upheld.