Why guessing the future is a losing game — and what to do instead

In today’s supply chains, the old playbook of forecasting future demand based on past trends is quickly becoming obsolete. Shocks and disruptions — whether pandemics, geopolitical upheavals, or sudden shifts in consumer behavior — strike faster and harder than ever before. Companies that rely solely on traditional forecasting methods risk costly delays, stockouts, or excess inventory.

But there’s a smarter way: scenario planning powered by artificial intelligence and large language models (LLMs). Instead of betting everything on one predicted outcome, this approach prepares your business for multiple plausible futures, helping you turn uncertainty into a strategic advantage.

What is scenario planning and why does it matter?

Scenario planning isn’t about predicting the future, it’s about preparing for a range of possible futures. Inspired by the insights of John Kay and Mervyn King in Radical Uncertainty (2020), this method encourages managers to use qualitative reasoning and narratives rather than relying solely on quantitative forecasts. The key question becomes:
“What is the range of plausible outcomes, and how can we remain successful across that range?”

By identifying critical uncertainties and exploring divergent but plausible scenarios, businesses can test strategies, build contingency plans, and increase resilience without needing to assign probabilities to every outcome.

How technology makes scenario planning smarter and more practical

Thanks to advancements in Optimal Machine Learning (OML) and digital twin technology, supply chains can now be modeled with remarkable detail and realism. Digital twins create high-fidelity, dynamic simulations of your supply chain network: tracking inventory, capacity, economics, and forward-looking factors like forecasts and promotions.

Cloud-based infrastructures ensure that these simulations are scalable and cost-effective. The Inventory and Supply Chain Management tool developed by AD3 and delivered in partnership with LTPlabs, is a prime example. It allows planners to experiment with strategies such as adding buffer stock, switching suppliers, changing transportation modes, or expanding capacity. Every scenario is measured by critical KPIs, including revenue, profit, costs, and service levels.

The game-changer: Large Language Models (LLMs)

Large Language Models add a conversational layer to scenario planning that transforms how planners interact with complex data. With LLMs, you can:

• Configure and tweak scenarios through natural language commands

• Translate qualitative narratives into structured inputs for simulations

• Validate assumptions against historical and real-time data

• Interpret simulation results and get answers to both qualitative and quantitative questions

LLMs such as Generative AI make scenario planning faster, more accessible, and easier to explain to decision-makers across the organization. As Menache et al. (2025) highlight, this means planners spend less time wrestling with software and more time making strategic decisions.

Why This Matters for Your Business

What happens if your supply chain is caught off guard by the next disruption?
Can your current planning approach handle an uncertain, fast-changing future?

Companies stuck in traditional forecasting risk missing critical warning signs and losing agility. On the other hand, scenario planning powered by AI helps your business:

• Reduce costly supply chain disruptions

• Make faster, more confident decisions

• Increase operational flexibility and resilience

• Gain a competitive edge in volatile markets

In short, this isn’t just about surviving uncertainty, it’s about turning uncertainty into your biggest strategic advantage.

Getting Started: Key Questions for Practitioners

To successfully adopt this approach, supply chain leaders need to reflect on:

• What are the most critical features needed in scenario planning?

• Which strategies should be tested under uncertainty?

• What metrics will provide the clearest insights for decision-making?

Equipped with comprehensive data infrastructure, accurate digital twins, advanced optimization, and AI-powered interfaces, planners can adopt a “real option” mindset, investing in flexible, adaptive capabilities that keep your supply chain resilient no matter what the future holds. Stop waiting for certainty to arrive. Prepare for every future and turn uncertainty into your advantage.

‍

Application of LLMs to Support Supply Chain Scenario Planning

Authored by M. Cohen and V. Deshpande and presented at Supply Chain Thought Leaders conference, on June 17, 2025, this document explores the integration of large language models (LLMs) into supply chain scenario planning to manage radical uncertainty. Traditional forecasting methods are inadequate in today’s unpredictable environment, leading many companies to delay decisions. Scenario planning, instead of aiming to predict specific outcomes, provides a framework for evaluating decisions across a range of plausible futures. As highlighted by John Kay and Mervyn King in Radical Uncertainty (2020), managers should use qualitative reasoning and narratives to guide their thinking. The central question guiding this approach is: “What is the range of plausible outcomes and how can we remain successful across that range?”

Scenario planning involves identifying critical uncertainties and driving forces to construct divergent but plausible scenarios. These scenarios do not require probability assignments but are used to evaluate specific strategies and develop both pre-emptive and reactive plans. Gad Allon (2025) emphasizes this qualitative, narrative-driven approach as key to dealing with complexity.

Optimal Machine Learning (OML) supports this process by enabling agile and resilient supply chains through the use of extensive, detailed, data inputs and high-fidelity digital twins. These twins link data, decisions, and outcomes within a cloud-based infrastructure that supports scalability and cost-efficiency. As detailed by Agrawal, Cohen, and Deshpandes, in the Harvard Business Review (March–April 2024), such infrastructure can provide the backbone of data-driven supply chain management.

The AD3 Scenario Planning Tool brings this to life by modeling the supply chain in terms of its network structure, inventory and capacity status, economics, and forward-looking information like forecasts and promotions. Scenario planning within AD3 involves modifying this model to represent future business possibilities, including changes to the network, product flows, and supply chain economics. Strategies evaluated include holding buffer inventory, adding supply sources, using alternate transport, shifting production, and building new capacity. The effectiveness of each strategy is measured using metrics like revenue, profit, costs, fill rate, and service coverage for each scenario outcome.

LLMs enhance scenario planning by offering a conversational interface to configure scenarios, translating narratives into quantifiable inputs, and validating them against existing data. They also assist in interpreting simulation results and can answer both qualitative and quantitative queries about supply chain behavior. Menache et al. (2025) highlight how generative AI simplifies access to planning systems, supports decision explanation, monitors compliance, and accelerates scenario development.

Key takeaways from this work include the importance of building comprehensive transactional data infrastructure and creating accurate digital twins of supply chains. Advanced modeling, state-of- the-art optimization solvers, and machine learning help define optimal decisions, while LLMs make scenario planning more interactive and accessible. Planners are encouraged to adopt a “real option” mindset that promotes flexibility and resilience by investing in adaptive capabilities.

To implement this approach effectively, practitioners must ask: What are the most important features for scenario planning? Which strategies should be evaluated? And what metrics should scenario tools provide?