Q 812.  

In many processes, AI and humans make different kinds of mistakes — AI may miss context, while humans may overlook data-driven insights. The real power comes when both learn from each other’s decisions over time.

Think of a process in your domain where humans and AI work together (or could). What specific outcomes or interactions should AI learn from humans, and what should humans learn from AI to continuously improve the process?
 

⚠️ Note: Any answer that is generic or does not connect with a specific, relevant process will not be approved.

🏆 The best answer will be selected on the basis of:

• Relevance and clarity of the chosen process

• Depth of insight into mutual learning between AI and humans

• Practicality of implementing this learning loop

 

Note for website visitors -

• This platform hosts two weekly questions, one on Monday and the other on Thursday. 
• All previous questions can be found here: https://www.benchmarksixsigma.com/forum/lean-six-sigma-business-excellence-questions/. 
• To participate in the current question, please visit the forum homepage at https://www.benchmarksixsigma.com/forum/.
• The question will be open until Monday or Thursday at 5 PM Indian Standard Time, depending on the launch day. 
• Responses will not be visible until they are reviewed, and only non-plagiarised answers with less than 5-10% plagiarism will be considered for winner selection. 
• If you are unsure about plagiarism, please check your answer using a plagiarism checker tool such as https://smallseotools.com/plagiarism-checker/ before submitting. 
• All correct answers shall be published, and the top-rated answer will be displayed first. The author will receive an honourable mention in our Business Excellence dictionary at https://www.benchmarksixsigma.com/forum/business-excellence-dictionary-glossary/ along with the related term.

• Some people seem to be using AI platforms to find forum answers. This is a risky approach as AI responses are error-prone because our questions are application-oriented (they are never straightforward). Have a look at this funny example - https://www.benchmarksixsigma.com/forum/topic/39458-using-ai-to-respond-to-forum-questions/

• We also use an AI content detector at https://quillbot.com/ai-content-detector. Only answers with less than 45-50% AI-generated content will be considered for winner selection.

In our department Quality Assurance in manufacturing one of the most important processes what we call as Complaint to Correction (C2C).
It starts when a customer complaint is received and ends when the problem is fully corrected in production process. Here both AI and humans can learn a lot from each other.

Where AI should Learn from Humans

AI can read data but it does not understand customer expectations or product function requirement the way a quality engineer does.
That is where it needs to learn from people.

1. Understanding the “Why” behind decisions
   When a QA engineer decides whether a defect part can be reworked / DN raised or must be scrapped, they look at things AI can not ex:- whether the defect is on a sealing surface. How strict that customer or specification is or if it’s a approved deviation is one-time conditionally accepted.
   If we record those reasons clearly, AI can start recognizing similar situations and give better suggestions next time.
2. Learning Root Cause Patterns
   Engineers often find the real reason behind a failure maybe a worn tool, wrong setup, missed Tool Life Monitoring or handling issue.
   When AI studies those final root cause reports, it learns to connect machine data (setup change / Tool Life Monitoring) or process changes with real defects.
3. Customer Sensitivity
   AI should also learn, how different customers react and what different customer specification permit for each defect. Some accept small cosmetic defects, others reject even minor ones. By learning from our past decisions, AI can predict customer reactions more accurately.

 

Where Humans Should Learn From AI

While AI cannot see every detail on the shop floor, it can process massive amounts of data that we might miss.

1. Spotting Early Warnings
   AI can analyze NCR’s and say “This same scratch issue happened right after a tool change last month.” That helps quality engineers act early before the same issue reaches the customer again.
2. Keeping Us on Track
   AI can monitor complaint timelines 48 hour containment, 4 - day RCA for L1 complaint issues  and remind or escalate when deadlines are at risk. This informs the team to be more disciplined and prevent any delays effecting the KPI.
3. Showing Useful Trends
   AI can highlight patterns like “80% of similar cases were successfully reworked.”
   That helps new engineers make quicker, more confident decisions using real data.

 

The Learning Loop

Here is how this partnership works in practice:

• AI helps by blocking the shipment in ERP when NCR is registered for a particular WO, checking for recurring issues and reminding teams about complaint timelines (Containments action, RCA, CA & PA).
• Humans step in to inspect the parts, measure the cosmetic defect (length, Width & Depth) decide the real root cause and choose the right corrective actions based on drawings and customer needs.
• The decision and reasoning are saved in the system. AI then learns from that for future cases.

Next time, AI gives smarter recommendations and engineers make faster, data driven decisions.
Over time, both get better.

 

The Result

This shared learning loop reduces delays, prevents repeat complaints and keeps us compliant with customer KPI’s. RCA cycle time & PPM.
AI brings the speed and data based decisions, humans bring experience based judgment and customer understanding.
When both AI & QA learn from each other, the result is a stronger, more proactive quality processes and not just reacting to problems, but preventing them before they happen.

 

Process Chosen: Underwriting Process for Life Insurance Policy

 

Globally there has been a debate over right AI model applicable for the underwriting process for Life Insurance Policy.

An Actuarial AI model used by Life Insurers is required to perform risk assessment and flag high risk profiles based on various factors like:

1.     Pre-requisite medical check-ups conducted at Insurance Company appointed labs which includes blood samples, X Rays etc.

2.     Evaluation of previous medical history of the applicant which includes being admitted to a hospital for a list of ailments.

3.     In addition, the declaration done by a person in the application like lifestyle preferences like if he / she consumes alcohol, tobacco etc. If he or she exercises or not etc.

Human underwriters are then required to review and analyze the indicators flagged by the Actuarial AI model. The final decision is made by human underwriters specifically in the borderline or complex cases, by applying qualitative judgments before approving, modifying, or rejecting policies.

 

What AI Should Learn from Humans:

1. Compensating & Changing Circumstances: There are cases wherein an underwriter approves an applicant with high-risk factors where they see a controlled chronic illness for years (Eg. An applicant has heart related issue but has been stable & controlled for 5 or more years), radical change in adoption of good lifestyle preferences for recent years (Eg. An applicant used to consume alcohol and smoke but has left it for more than 5 years or more). These compensating & changing circumstantial factors is what AI can learn instead of automatically flagging them as high risk.
2. Handwriting & Documentation: Humans can interpret ambiguous medical reports and seek clarity if not fully understood. Handwriting of doctors, paramedics, nurses notes can be interpreted by humans easily and they can seek clarity if not fully understood. An Actuarial AI model will run its full course of action and end up misclassifying an ambiguous lab report or poor handwriting. Training an Actuarial AI model on these insights can improve reliability of AI model.
3. Ethical Factors: An Actuarial AI model can learn incorporating ethical factors like fairness, anti-discrimination against a certain gender, race, religion based on past learnings of the model, or approving a policy for a high-risk applicant with additional conditions.

What Humans can learn from AI:

1. Hidden Risk Patterns: AI can identify subtle correlations between lifestyle, lab results, and long-term mortality risk that may not be obvious to human underwriters.
2. Consistency and Benchmarking: Reducing subjective variability in policy decisions and to standardize risk scoring across underwriters can be achieved by Humans using AI recommendations
3. Predictive Efficiency: AI can suggest optimal premium adjustments or policy terms based on large-scale actuarial analysis, helping humans make faster, data-driven decisions.

Continuous Learning Loop Implementation:

• Every underwriting decision (AI recommendation vs. human override) is logged for model retraining.
• Monthly calibration meetings review high-impact discrepancies, annotating reasons for approval, modification, or rejection.
• Underwriters receive dashboards comparing their decisions with AI risk scores and rationales, enabling reflection and skill growth.

Outcome:
Over time, AI develops a more nuanced understanding of applicant risk beyond pure data patterns, while human underwriters enhance their data-driven decision-making and consistency. This co-learning loop reduces underwriting errors, improves risk-adjusted profitability, and accelerates policy processing with greater fairness and transparency.

 

I am trying to give a shot to this question with the help of a scenario where we support Energy & Utilities Industry client. The Client operates on SAP technology as its core system and trying to visualize how humans and AI can collaborate.

 

Scenario – Predictive Maintenance of Power Grid Equipment

Systems monitor equipment (like transformers, circuit breakers, etc.) using sensor data, historical failure patterns & environmental conditions to predict potential failures. Field Engineers then validate, check, or override these predictions based on field experience and subject knowledge

 

AI learning from Humans – Engineers might ignore a failure prediction because they know the equipment was recently serviced or the sensor is faulty. AI should learn to incorporate such overrides and flag similar patterns in future predictions. Humans may delay maintenance due to local festivals, weather conditions, or workforce availability. AI can learn to factor in these constraints when scheduling or prioritizing tasks. When humans diagnose root causes post-failure, AI should include this feedback to refine its models (e.g., distinguishing between thermal overload vs. insulation breakdown). These show scenarios from Contextual overrides, environmental, social factors and Failure attribution.

Predictive Analytics via SAP Predictive Maintenance and Service 

IoT Integration using SAP Asset Intelligence Network 

Machine Learning models embedded in SAP BTP (Business Technology Platform)

 

Humans learning from AI - AI can detect subtle correlations (e.g., humidity spikes + voltage fluctuations = higher failure risk) that humans may not identify easily. Engineers can use these insights to revise inspection protocols or preventive measures. AI can suggest which assets need urgent attention based on risk scores, helping humans prioritize better. This improves operational efficiency and reduces downtime. AI provides 24*7 monitoring, alerting humans to anomalies they wouldn’t catch during periodic inspections. These show areas for implied patterns in data, optimal HC allocation, triaging mechanisms. Field engineers and planners interact through SAP PM (Plant Maintenance), feeding in real-world insights, overrides, and contextual decisions.

 

For effective deployment of AI with Human framework, we should work on an effective governance model –

Feedback Capture: Every human override or comment is logged in SAP

Model Retraining: AI models in SAP BTP are updated using this feedback

Insight Delivery: AI-generated insights are pushed to SAP Fiori apps or mobile dashboards

Human Validation: Engineers validate or reject insights, creating a continual improvement cycle

 

This model will help reap benefits in long term through reduced equipment failure rates, optimized maintenance costs, improved grid reliability, enhanced and trusted workforce productivity.

Symbiotic Intelligence Loop: Where Ai and Human examiners help each other to correct their blind spots

In healthcare BPO setup, we see AI and human examiners make different but complementary mistakes, for e.g. in pend resolution and claim adjudication processes. I think future lies in creating Symbiotic Intelligence Loop, where continuous learning occurs from each other

Mistakes that AI makes and where human examiners can help it improve

·        AI may flag a claim as “Provider mismatch” because it may not know that new provider ID was added last week. Human examiner will look at the context, correct the classification and add a note. AI will learn this pattern and stop flagging false pends from next time ( false positive)

·        AI may auto deny claims that looks same as past denials, even if the policy rules have changed. Human examiner will override such cases and feed updated rule in AI. AI will realign with the new payer logic ( false positive)

·        Free text notes including comments, one-word instructions, or human annotations (e.g.: “confirmed via email “are many times interpreted incorrect by AI. Human examiner gives structured feedback. AI learns to tag these patterns as valid justifications  ( false negative)

·        AI struggles interpreting exception scenarios (for e.g.: duplicate claims linked to two different provider NPI). Human examiner in this case will provide valuable training inputs which will help AI to broaden its exception logic ( false negative)

 

Thus, with human examiners  as contextual partners, AI can become more adaptive, context aware and trustworthy

 

Mistakes that Human examiners make where AI can help improve

·        Human examiner may miss trends across hundreds of claims that get processed every day for single facility. AI can help detect data patterns, do better analysis- create dashboards and correlation charts helping human examiners arrive at exact root cause

·        Different human examiners may pend different claims differently. AI can help standardize decision rules, flag outliers for review making the process consistent

·        Human examiners have bandwidth issues and limited attention span which causes dip in accuracy in case of sudden volume spikes. AI can pre screen the claims, auto process straightforward logics so that examiners can only focus on claims steps needing true human judgement

·        Human examiners rely on periodic audits to identify process gaps for improvement. AI can give real-time feedback and predictive alerts helping examiners to correct issues proactively

 

This way, with AI as analytical partner, examiners can start making faster, data driven, more accurate decisions

 

In essence, this “Symbiotic Intelligence Loop” is not about humans training machines or machine replacing humans but both evolving together to create a faster, adaptive and smarter claims ecosystem

Winner of Q 812 – Shashi Prakash 

Shashi’s Life Insurance Underwriting entry nailed the brief: a specific, high-impact process with a two-way learning loop that’s actually implementable. AI flags risk; human underwriters apply nuanced judgment (compensating circumstances, ambiguous docs, fairness). Overrides are logged, reviewed monthly, and fed back into the model; underwriters get dashboards to benchmark and learn from AI. This is exactly how mutual learning should work.

 

🥈 Runner-up – Rohan Modak

A strong Healthcare Claims Adjudication model showing concrete mutual corrections (policy changes, free-text notes, exception handling) and how AI standardizes and triages in real time.

 

✅ Also Approved

 

• Adil Khan – Manufacturing QA (Complaint-to-Correction): excellent, shop-floor learning loop tying NCR, RCA, and customer sensitivity to smarter future recommendations.
• Lakshmi Iyer – Power Grid Predictive Maintenance (SAP): clear governance (feedback capture → BTP retraining → Fiori insights) and pragmatic human overrides.