Q 813.  

How Should AI Decide When Two Good Goals Conflict?

 

In real-world processes, AI systems often face trade-offs — speed vs. accuracy, cost vs. quality, efficiency vs. empathy.

An AI solution may need to choose between two desirable outcomes, and how it makes that choice determines whether it builds trust or causes friction.

 

Think of one process in your domain where such a conflict could arise.

How should AI decide which goal to prioritize — and what role should human oversight play in guiding that balance?

 

⚠️ 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 realism of the chosen conflict
• Depth of reasoning behind AI’s decision logic
• Practicality of balancing human and AI judgment

 

 

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.

Domine Chosen - Quality Assurance in Manufacturing

 

How Should AI Decide When Two Good Goals Conflict?

In manufacturing, AI often works in environments where both objectives are right — but they clash.
The Quality Department wants zero defects and compliance.
Production wants continuous flow and on-time delivery.
Supplier Quality wants stable cost and lead time.
Each goal is valid — but AI must learn how to prioritize them intelligently, not blindly.

---

Process 1: Production Throughput vs. Quality Compliance

This is the most common conflict in any factory.
Production says, “The parts look fine — don’t stop the line.”
Quality says, “It’s a critical surface — we can’t take that risk.”
AI sits in the middle, monitoring data from in-process inspections, sensors and rejection trends (as it has access to ERP parts produced so far, Shipped so far & rejection details).

 

How AI should decide:
AI should never “choose sides.” Instead, it should base its decision on risk severity and customer impact:

• If the deviation affects a functional (sealing surface) or safety-critical feature (CTQ / KC), AI must side with Quality — hold the batch, flag the risk and recommend containment (stock purge / Hold shipment).
• If the issue is cosmetic or within statistical control limits, AI can side with production — allow continuation but tag the lot for recheck or deviation (DN) approval.

AI’s role is to translate emotion into evidence. It provides data-backed probabilities like:

“Risk of customer rejection (particular part FPY / Rejection rate): rework delay: 6 hours; potential cost impact: €2,000 approx.”

 

Human oversight:
Quality and Production jointly review AI’s assessment.
If the risk is real, Quality leads containment (block shipment, stock purge). If not, Production proceeds under controlled deviation (DN).
The final authority stays human, because accountability and customer trust can’t be delegated to an algorithm. One bad batch delivered & PPM KPI goes to Toss.

---

 

Process 2: Supplier Quality — Delivery Lead Time vs Cost vs Consistency

AI may detect that a particular supplier’s rejection rate has increased and recommend re-sourcing to improve quality metrics.
But humans know the bigger picture, In few cases sources are customer defined (E.g. Raw Material from fixed mills / Special process processing at fixed suppliers). You do not much room to change source or switching suppliers. Also keep in mind it may take 6 – 8 weeks of lead time and higher cost to find a new supplier. which might hurt delivery commitments.

How AI should decide:
It should weigh all three factors Quality trend, response speed and commercial impact rather than looking only at defect percentage.
If supplier is improving and meeting agreed timelines, AI should suggest continued collaboration with close monitoring on performance.
If they are repeatedly failing to meet deliveries or rejection rate increases, AI can recommend escalation to HOD.

Human oversight:
Supplier Quality Engineers (SQE) validate this by performing on-site QMS audits, GEMBA Walks, noticing the communication tone (actions speak better than words) and trust history (honoring the commitments) AI cannot read these.
This ensures decisions taken balances both lead time, cost and quality, not one at the expense of other.

---

 

Process 3: Customer Complaint Handling — Speed vs Depth

When a new customer complaint is received, AI instantly block related parts in ERP and tracks response time. That helps in stopping the bleeding (close the tap) and meet SLA targets like 48-hour containment.
But too much speed risks quality of investigation (addressing the symptoms instead of root cause).

How AI should decide:
For high-risk or repeated complaints, AI should recommend deeper root cause analysis with human validation (AI has access to internal rejection data & RCA). AI can also check if a particular Work center Is causing more issues. In our ERP we have a grouping of all the parts (bellows / Purge Hood etc) so if we accepted complaint on one part of this group. AI also shows similar parts in the group. So human can check if same actions need to be horizontally deployed in the other parts from the family.

 

Human oversight:
Engineers review AI recommendations, if it fits the customer expectation and past complaint history. Example same complaint could have been received in the other part from same family. This keeps speed and depth in balance.

Horizontal deployment is very important aspect during customer complaint handling which several times is overlooked by QAE. If we deploy the same corrective actions in similar family parts then repeated type of customer complaints will come down drastically.  

---

The Human Role in the Balance

AI should prioritize using risk - based logic, but it must always flag uncertainty for human review.
Humans set the boundaries as per specification requirement (Drawings / standards), interpret customer expectations and carry responsibility for outcomes.

AI brings objectivity humans bring judgment.
Together, they turn conflicts between speed and accuracy, cost and quality, production and compliance into balanced, data-informed decisions that protect both the customer and the company’s integrity.

 

Process chosen: I have personally worked with a couple of companies dealing with U.S. debt collection — credit cards, health bills, and loans. This is a highly regulated industry, as these companies must comply with Federal Laws, State Laws, the Fair Debt Collection Practices Act (FDCPA), TCPA and the Health Insurance Portability and Accountability Act (HIPAA). Violations can result in severe financial penalties of up to $500,000 for class-action lawsuits, plus actual damages and attorney fees. Attorney General offices and the Better Business Bureau (BBB) typically handle initial escalations from complainants before the matter moves to legal remedies.

Conflict: Maximizing collection efficiency (contacts per hour, reduced AHT, automated dialing) vs. preserving legal & ethical compliance and consumer dignity (FDCPA/TCPA requirements, state rules, empathy for vulnerable consumers).

Why this conflict is realistic

Offshore operations are measured on throughput and recovery. At the same time, U.S. federal law (FDCPA) and the TCPA impose strict rules on when, how and with what consent collectors may call; many states E.g., Illinois: Limits calls to seven per week per debt, Michigan: Also applies a “seven calls per week” rule, New York: Restricts collectors to seven calls in a seven-day period etc mandate an additional layer of requirements and disclosures. An overly aggressive AI dialer that optimizes contacts/AHT can therefore trigger statutory violations, damaging reputation and incurring statutory penalties.

How AI should decide — practical decision logic

1. Hard legal constraints (non-negotiable rules).
   • Do not place autodialed or prerecorded calls without the appropriate TCPA consent for that number/type. If consent cannot be verified, do not autodial. (TCPA / FCC rules).
   • Respect FDCPA safe-harbor times (assume convenient time i.e. 8:00 a.m.–9:00 p.m. consumer local time unless affirmative info otherwise) and always include required initial-contact disclosures (the “Mini-Miranda”) where applicable.
   • Enforce consumer-specific flags: DNC, cease-and-desist, attorney representation, bankruptcy, time-zone mismatch — these must block calls immediately.
2. Soft operational objectives (optimize within constraints).
   • A constrained optimizer maximizes recovery KPIs (contact rate, promise-to-pay, AHT) subject to the legal constraints. The optimizer receives a dynamic risk score per call (combining consent certainty, debt age, consumer vulnerability flags, state of residence, prior complaints). Calls with low legal certainty get lower priority or are routed for manual contact.
3. Risk-tiered runtime behavior.
   • High legal risk or sensitive signals (no express consent for autodialing; detected emotional distress; regulated debt type): route to manual agent with a compliance-approved empathy script and supervisor on alert.
   • Medium risk: attempt manual, human-initiated call (no previewed robocall), or send written validation notice first.
   • Low risk + verified consent: allow optimized autodial + limited automated agent assist (script suggestions, real-time compliance prompts).
4. Conservative default & “hard-stop.”
   • Whenever the model’s confidence in a legally material fact (consent, timezone, identity) is below a threshold, the system defaults to the conservative behavior: no autodial, route to human, and log an audit event. This avoids costly TCPA/FDCPA exposure.
5. Explainability & audit trail.
   • Every decision (why a number was dialed, why AI suppressed a call, which script was shown) is logged with immutable timestamps, consent evidence, and model scores — essential for internal QA and external defense.

Human oversight — who does what, and how often

• Compliance Council (daily): reviews automated alerts (attempts blocked for legal reasons, flagged voicemails, consumer complaints), approves emergency rule changes, and signs off on changes to dialer aggressiveness.
• Legal team (weekly / on change): keeps rule engine synced with federal/state updates and court decisions that change TCPA/FDCPA interpretation.
• QA / Supervisors (real-time daily): sample calls, approve agent overrides, and convert representative overrides into retraining labels so the AI learns human judgment patterns.
• Model Governance Board (monthly): reviews performance vs. safety KPIs and decides whether to relax or tighten soft objectives.

Measurable governance & feedback loops

• Track Compliance Exception Rate (#calls blocked for legal risk per 1,000 attempts), Legal Incidents (regulatory complaints / suits), AHT, contact rate, and consumer complaint rate. Use a dual-threshold alerting system: immediate stop for spikes in complaints; slower policy review for drift in other KPIs.
• Simulate policy changes in a sandbox (play back historical call logs) and run a legal-scenario stress test before production rollout. Where legal precedent is shifting (e.g., TCPA interpretation), increase conservatism until legal clarity is restored.

Final Conclusion

Treat compliance and consumer dignity as a non-negotiable constraint and efficiency as the objective to optimize within that safe set. AI should be the consistent enforcer of the rules and the suggestion engine for efficiency — humans remain the arbiters of judgment, upgrades, and legal accountability. When in doubt, do not dial; escalate.

Taking an example from the Publishing sector. The publishers publish research papers based on defined criteria. The common conflict is Speed vs. Quality of Publishing.

 

Process

When a manuscript is submitted to a Journal, AI tools are used to perform an initial assessment — checking relevance, ethics, plagiarism, and even estimating novelty or impact. The objective is to determine the appropriate speed at which the manuscript should be sent to peer reviewers or rejected outright based on the assessment.

 

Conflict

- Speed: Authors expect fast decisions. Hence, Journals want to reduce turnaround time to increase author satisfaction.

- Quality: A deeper AI analysis (e.g., semantic similarity, citations, novelty) takes longer but improves decision accuracy in false rejections or sending the manuscripts to the peer reviewers.

 

AI algorithm

1. Based on the Journal's reputation

- For high-impact journals, it should prioritize quality by using deep Natural Language Processing and citation network analysis to assess novelty and relevance.

- For low-impact journals, it relies on lightweight checks and prioritize speed.

 

2. Analysis confidence level score of the output

AI assigns a confidence level to its triage decision based on the data for which it is trained.

- If the confidence level is high, the decision is taken automatically by AI

- If the confidence level is low, it will escalate to the editorial staff.

 

Role of human oversight

- Editors need to define the weightage of speed and accuracy (e.g., 75% accuracy, 25% speed) based on journal type (high impact/medium impact/low impact).

- They must periodically audit AI decisions to ensure calibration with editorial standards. This will lead to refinement in the AI algorithms.

- For manuscripts in emerging technologies or interdisciplinary areas, human editors must train AI to avoid false rejections.

 

When Two Good Goals conflict, AI should decide by using a framework that prioritizes ethical principles, stakeholders value, and quantifiable impact.

Let us take Insurance as a domain as an example and understand how AI should decide if there are conflict between two good goals.

A common conflict is between the good goal of Maximising company profitability by minimizing payouts and prevent fraud. The good goal of providing accessible and fair coverage by approving more claims quickly.

1.       Goal definition and Metric identification

The AI must first clearly define the TWO conflicting goals & their underlying values  

Goal 1: Minimize Risk / Fraud

Metric : Reduction in false claim

Goal 2 : Maximize Customer satisfaction

Metric : Speed to process the claim

2.       Establish Priority and Constraint threshold

Ai needs to set Constraint and Values based on business strategy, regulatory requirements  and ethical guidelines

Constraint: In case of Insurance domain AI should always comply with anti-fraud regulations and never unfairly reject a legitimate claim. This sets a goal of “ Minimize Risk/ Fraud”

Values : Assign different values to each goal to reflect business priorities

3.       Risk based triage and Segmentation

Ai should use Predictive Model to anticipate and segment cases and apply different decision strategies

Low-Risk Claims: The AI should prioritize speed. These claims are automatically approved and paid out quickly to maximize customer satisfaction (e.g., a simple car insurance claim with police reports and clear evidence).

 

High-Risk Claims: The AI should prioritize investigation. These claims are flagged for deeper scrutiny (e.g., claims filed immediately after policy purchase, or those with highly unusual circumstances). This maximizes the fraud reduction goal but incurs a temporary, necessary delay in customer experience.

 

Medium-Risk Claims: The AI can apply an adaptive strategy. It might initiate a brief, automated verification process (e.g., a quick check of public records or cross-referencing against fraud databases). If the check is clear, it proceeds to fast payout. If a minor red flag is raised, it might involve a human for a rapid, focused review, balancing both goals.

4.       Dynamic trade-off Analysis

For cases where a direct conflict must be resolved, the AI should use  its established weights and metrics to calculate the potential "cost" of favoring one goal over the other.

Scenario: A claim has an estimated 5% probability of being fraudulent (Risk Cost: 5% of the claim amount) and an estimated investigation time of 5 days (Experience Cost: 5 extra days of delay).

Decision Calculation: The AI weighs the expected financial loss from potential fraud against the quantifiable cost of poor customer experience (e.g., a measured drop in CSAT or increased churn risk due to delay).

If the weighted cost of the delay/poor experience is higher than the weighted cost of the potential fraud, the AI favors payout speed (approves the claim).

If the weighted cost of potential fraud is significantly higher, the AI favors investigation (delays the payout).

This structured approach ensures that decisions are transparent, consistent, and aligned with the insurer's regulatory obligations, ethical standards, and overall business strategy.

Human oversight plays a critical and indispensable role in guiding the balance between conflicting goals for an AI, particularly in the sensitive domain like insurance .The AI's decision framework (prioritizing risk v/s customer satisfaction) is only effective and ethical when overseen by humans who provide the necessary context, ethical judgment, and strategic calibration. Human oversight transforms the AI from a purely predictive tool into a responsible and accountable decision-support system. The AI handles the high-volume, standard decisions, while humans manage the strategic direction, ethical integrity, and complex exceptions.

In the aerospace domain, particularly within airport services and operations powered by SAP technologies, a frequent conflict arises between operational efficiency and passenger experience. Taking this as a scenario, trying to elaborate my take on how both these good goals create a conflict and how AI could be leveraged here.

For example, SAP-based AI systems managing gate assignments or baggage handling may face a trade-off between minimizing turnaround time (efficiency) and ensuring personalized service (experience). An AI might recommend rerouting ground staff or reallocating gates to optimize aircraft flow, but this could result in longer walking distances for passengers or reduced accessibility services.

To navigate this, AI should be designed to:

• Analyse real-time operational metrics (e.g., aircraft delays, gate availability, staff capacity)
• Incorporate passenger-centric data (e.g., special assistance requests, connecting flight urgency)
• Simulate outcomes to forecast the impact of prioritizing one goal over another with the help of statistics and already existing datasets

The decision logic should be dynamic — for instance, during peak hours, efficiency may take precedence, while during off-peak times, passenger comfort could be prioritized.

Human oversight plays a pivotal role in setting these contextual thresholds. Airport operations managers, supported by SAP dashboards & monitoring mechanism, must define business rules and override AI decisions when exceptions arise for e.g. VIP movements, emergency landings, or regulatory constraints

Metrics AI Should Monitor

To make informed decisions, AI systems integrated with SAP (e.g., SAP S/4HANA, SAP EWM, SAP Fiori) should continuously monitor:

1. Operational Efficiency Metrics
   • Aircraft turnaround time
   • Gate utilization rates
   • Staff availability and workload
   • Baggage transfer time
   • Real-time flight schedules and delays
2. Passenger Experience Metrics
   • Walking distance to gates
   • Wait times at check-in, security, and boarding
   • Special assistance requests (e.g., elderly, disabled passengers)
   • Connection times for transfer passengers
   • Passenger satisfaction scores (from surveys, feedback apps, random on the go feedback on airport waiting areas to get the Gemba validation)
3. Contextual and External Factors
   • Weather conditions
   • Regulatory compliance requirements
   • Emergency Landings or VIP movements
   • Real-time disruptions (e.g., system outages, strikes)

Important Factor: How Should AI Decide Which Goal to Prioritize?

AI should not rely on static rules but instead use context-aware decision logic.

• Dynamic Thresholds: AI can adjust priorities based on time of day, passenger volume, or disruption severity. For example, during peak hours, prioritize efficiency; during off-peak, enhance comfort
• Weighted Scoring Models: Assign weights to each metric based on business rules. For instance, if a flight is delayed but carries many connecting passengers, AI may prioritize passenger experience over gate optimization
• Predictive Analytics: Use historical data to forecast the impact of decisions. If rerouting baggage saves ~5 - 10 minutes but increases mishandling risk by ~20 – 30 %, AI can flag it for human review
• Scenario Simulation: AI can simulate multiple outcomes and recommend the one with the least operational risk and highest passenger satisfaction

Important Parameter: Role of Human Oversight

Human oversight is essential to ensure AI decisions align with strategic, ethical, and regulatory considerations:

• Define Guardrails: Operations managers must set business rules and thresholds that AI cannot override such as accessibility standards or regulatory mandates
• Override Mechanism: SAP dashboards should allow supervisors to override AI decisions in real time, especially during emergencies or exceptions
• Continuous Feedback Loop: Human feedback should be fed back into the AI model to improve future decision-making. For example, if AI’s gate assignment led to passenger complaints, that data should refine future logic
• Ethical Judgment: AI may not fully grasp nuances like cultural sensitivities or reputational risks. Human judgment ensures decisions are empathetic and brand aligned

Efficiency vs Experience - AI should act as a decision support system, not a decision maker. By combining real-time SAP data, predictive intelligence, and human oversight, Airport Ops can strike the right balance between efficiency and experience — building trust with passengers while maintaining operational excellence

 

AI with Human Compass- Balancing the Speed and Quality through Context aware intelligence

In the world of healthcare BPO – process improvement and automation, one reoccurring conflict is between speed and quality in claims adjudication. Both goals are desirable for the business- faster turnaround improves cost efficiency and client satisfaction while quality of claim processing protects trust and compliance. Yet when we use AI systems for claims auto adjudication, these goals ca easily pull in opposite direction

For e.g.: AI might adjudicate 80% if the claims in few seconds, but rushing through complex notes (involving multiple providers or new payer edits) could result in wrong denials. Conversely, being too cautious and routing all the claims for manual intervention negates the very efficiency AI was meant to deliver

That’s why we need to implement below principle in AI design: AI with human compass- A framework where AI doesn’t make binary choices but instead used context aware and logic which goals to prioritize. Below is how it will work:

·        For low risk, rule-based claims, AI will prioritize speed. driving costs and time savings

·        For high-risk ambiguous claims, AI will escalate for human intervention, ensuring compliance and accuracy

This layered decision logic is not static, but risk adaptive. AI continuously learns from past interventions updating its confidence scores and thresholds.

AI should be guided by human values and judgment frameworks, not just data driven optimization. AI can be trained to recognize “risk zones” where human oversight is essential, just like copilot alerting the pilot when conditions are uncertain.

Human oversight plays a guiding role in this ecosystem, not as a supervisor but a moral and contextual governor. Human examiners will define trade-off zones, review exceptions, and retrain the AI where Payer logic and policy nuances arise

Thus, AI with Human Partnership creates a partnership where:

·        AI drives scalability

·        Human examiners anchor judgement, empathy, trust

This intelligent, self-improving balance between performance and prudence- An AI that doesn’t just process faster but decides smarter

Winner of Q 813 – Shashi Prakash (U.S. Debt Collection)

Shashi delivered a standout design for balancing collection efficiency vs. legal/ethical compliance. The AI optimizes only within hard constraints (FDCPA/TCPA/HIPAA), uses risk-tiered runtime behavior, defaults conservatively when consent/time-zone confidence is low, and maintains a defensible audit trail. Human oversight is crystal clear (Compliance/Legal/QA/Governance cadence) with sandbox testing and dual-threshold alerts. This is best-in-class, production-ready thinking.

 

Runner-up – Adil Khan (Manufacturing QA)

Excellent, shop-floor logic across throughput vs. quality, supplier delivery/cost vs. consistency, and complaint speed vs. depth. CTQ-first decisions, real containment, and SQE audits/Gemba make it highly practical.

 

Also Approved

• Rohan Modak — Healthcare Claims: Risk-adaptive triage (auto vs. human), evolving thresholds, human-defined trade-off zones.
• Manik Sood — Publishing: Journal-tier policies, confidence-based escalation, audit & calibration for speed vs. quality.
• Indrani Ghosh — Insurance Claims: Clear fraud-vs-CX framework with triage and human ethical guardrails.