Optimization, GenAI and quantum : From optimization models to resilient supply chain decisions

Boardroom conversations about supply chains are changing.

The question is no longer, “What is the optimal plan?”

It is, “How does this plan perform when reality diverges from the forecast?”

In part 1 of this blog series, Optimization, GenAI and quantum: Making complex supply chain decisions faster and more accessible, we explored how optimization must evolve from a black-box model into a transparent and accessible decision-making capability. Under uncertainty, that evolution becomes even more critical.

Tariff shocks, geopolitical tension, climate disruption and demand volatility are reshaping the risk landscape. In this environment, leaders are no longer simply choosing the “best” plans. They are choosing risk positions.

• How much downside exposure is acceptable?
• How much buffer should we fund?
• How do we make our networks more resilient?

This is fundamentally a discussion about uncertainty and risk, which requires resilience to be engineered into the decision system.

Addressing supply chain uncertainty with stochastic optimization

Traditional optimization answers a clear question: What is the best plan given what we know? That approach works when variability is limited. Unfortunately, complex supply chain questions are not always that clear-cut.

Today, uncertainty is persistent and systemic. Costs shift by the load, lead times fluctuate, and capacity can tighten without warning. Robust optimization approaches for varied parameters address this reality directly. Instead of optimizing for one forecast, they evaluate performance across many plausible futures. They allow leaders to compare:

• Average-case plan versus variation-inherent plan
• Single-point service level versus distribution of service outcomes
• Nominal cost versus risk-adjusted cost
• Efficiency gains versus tail-risk exposure

Stochastic optimization makes these trade-offs explicit. It translates uncertainty into measurable decision policies, risk metrics and trigger thresholds.

Designing credible scenarios with GenAI to manage supply chain risk

The quality of those trade-offs depends entirely on the quality of the scenarios behind them. Weak scenario design can lead to two outcomes:

• Optimizing for a future that never materializes
• Adding buffers everywhere and eroding performance

Effective operations research modeling[PK8.1] requires disciplined scenario generation grounded in data, expert judgment and transparent assumptions.

This is where generative AI (GenAI) becomes valuable—not as a substitute for rigor, but as an accelerator of structured thinking.

GenAI helps translate risk discussions into structured scenario inputs that can be debated and calibrated. For example, it can help teams quickly develop:

• A catalog of plausible disruption scenarios
• Defined adjustments such as longer lead times, higher costs or reduced capacity
• Structured multi-stage scenarios that reflect how events might unfold over time.

Crucially, these outputs remain working drafts. Assumptions must still be reviewed, challenged and supported by data. GenAI speeds up the process, but it does not replace evidence or human judgment.

Using GenAI to make optimization decisions accessible to leaders

Well-designed optimization models[PK9.1] deliver value when decision-makers can clearly interpret and act on their insights. As scenarios multiply and trade-offs grow more complex, model outputs can become difficult to explain. Leaders need clarity, not technical detail. They need to understand how a policy shift affects service exposure, cost variability and operational stability.

GenAI can help bridge that gap.

Instead of waiting for a revised analysis, leaders can ask structured questions in business language:

• What happens if tariff exposure increases by 5%?
• How does service risk change if we cap premium freight?
• What is the downside exposure if we reduce safety stock in Region A?

The model translates these questions into clear decision narratives: what changed, what constraints became binding, where penalties were incurred, and where risk increased. This transparency builds trust. And trust determines whether optimization informs decisions or remains an analytical exercise.

Taking a hybrid architecture approach that scales

As introduced in our part 1 blog, scalable decision systems are rarely built on a single technology. They are hybrid by design :

• AI and machine learning forecast key drivers such as demand, lead times and disruption probability.
• GenAI translates unstructured inputs into structured model updates and explains outcomes in business terms.
• Optimization produces auditable, feasible decisions aligned to objectives.
• Simulation or digital twins test policies against operational variability.
• Quantum-based optimization can accelerate exploration of complex decision spaces, such as solving high-dimensional scheduling, routing and portfolio trade-off problems.

Each component has a distinct role. Together, they create a governed decision ecosystem rather than a standalone model.

An example: Fleet optimization under tariffs, time and cross-border complexity

Consider a shipper aiming to maximize full-truckload (FTL) utilization across a cross-border network with tight delivery windows and shifting tariffs. Thousands of orders, trailers, drivers and constraints can produce millions of feasible assignments and routing combinations.

Optimization can assign orders to trucks and routes to maximize FTL utilization while respecting time windows, equipment constraints and carrier contracts. However, as tariffs alter landed costs, border crossing times fluctuate, and carriers adjust capacity, the challenge scales rapidly.

In high-complexity environments like this, quantum computing, particularly hybrid quantum-classical approaches, can accelerate specific combinatorial problems, such as assigning loads to trucks or routing through heavily constrained networks. Quantum techniques can explore large solution spaces more efficiently, while classical optimization refines and validates the most promising options.

The business objective is not simply lower cost. It is about arriving at strong, defensible decisions faster when planning timelines are tight and volatility is high.

At the same time, leadership intent must translate quickly into execution. Leaders express policy in plain language:

• “Prioritize FTL consolidation on Corridor A.”
• “Cap tariff exposure at $X per load.”
• “Avoid border crossing B for two weeks.”
• “Reduce fleet size by X trucks/trailers.”

GenAI translates these directives into structured model updates (decisions and constraints), creating scenarios that planners can review and approve, significantly reducing the time between policy decisions and operational adjustments.

GenAI also explains the outcomes in business terms: what changed, which constraints became binding, how tariff exposure shifted, and what re-plan triggers should be activated.

Fleet planning becomes a continuously adaptive system, not a periodic optimization exercise.

Building risk-aware, resilient supply chain decision systems

We know that optimization must operate as a governed capability rather than a one-time initiative. In volatile markets, that capability must extend beyond operational efficiency to active risk oversight at the enterprise level.

Today, supply chain decisions are less about selecting a single optimal plan and more about defining an intentional risk posture that balances cost, resilience and exposure with clarity and discipline.

The combination of stochastic optimization, GenAI and quantum computing enables this shift. Together, they support decision systems that quantify uncertainty, clarify trade-offs and adapt at scale.

Connect with our experts to learn how this approach can help support your supply chain strategy.