#autonomousplanning

[deepvuinc]

Demand Planning: Moving Beyond Basic ML Models with Shock Resilient AI Decisioning Agents

In the rapidly evolving landscape of business analytics, demand forecasting stands as a critical pillar for manufacturers  across FMCG  industries. Accurate predictions of customer demand enable businesses to optimize inventory management, production schedules, and resource allocation, ultimately driving efficiency and profitability. However, as many organizations have experienced, maintaining the accuracy of demand forecasting models can be a persistent challenge, especially with basic machine learning (ML) approaches that lack resilience to shocks and unforeseen events.

In recent years, the shortcomings of conventional ML models in handling sudden disruptions and maintaining accuracy over time have become increasingly apparent. Factors such as commodity spikes, new trade constraints and complexities, market fluctuations, consumer spend (Personal Consumption Expenditures), and unexpected events like supply chain disruptions or global geopolitical crises can quickly render traditional forecasting models obsolete, leading to costly errors and missed opportunities.

DeepVu is at the forefront of addressing these challenges with our VuDecide product, with its innovative Shock Resilient Decisioning Agents. These AI Agents represent a paradigm shift in demand planning, leveraging advanced AI techniques to adapt dynamically to changing conditions, recommending actions that optimize directly for your KPIs maintaining# accuracy in the face of uncertainty.

So, what can businesses do when their demand forecasting models start fading in accuracy? Here are some key considerations:

Evaluate Current Model Performance: Before making any changes, it's essential to assess the performance of existing forecasting models. Identify where and why inaccuracies are occurring, whether due to data quality issues, model limitations, or external factors.

Explore Advanced AI Solutions: Basic ML models may struggle to cope with the complexity and volatility of real-world demand dynamics. Exploring advanced AI solutions like DeepVu's Shock Resilient Decisioning Agents can provide a more robust framework for demand forecasting, capable of adapting in real-time to changing conditions and outlier events.

Integrate External Data Sources: Enhance the accuracy and robustness of forecasting models by incorporating a diverse range of external data sources. We call this VuGraph an expansive and continuously growing supply chain knowledge graph. This  includes macroeconomic indicators (interest rates, treasury yields, unemployment rates, wages etc),  commodity prices, production volumes, PPI (producer price index), weather data, or industry reports, providing valuable context and insights for more informed decision making.

Implement Continuous Learning: Static models quickly become outdated in today's dynamic business environment. Implementing mechanisms for continuous learning and model refinement ensures that forecasting algorithms remain adaptive and responsive to evolving patterns and trends.

Invest in Resilience and Flexibility: Recognize the importance of resilience and flexibility in demand forecasting. By investing in technologies like DeepVu's Shock Resilient Decisioning Agents, businesses can build a more agile and responsive forecasting infrastructure capable of withstanding shocks and disruptions.

Monitor and Iterate: Continuous monitoring and iterative refinement are essential for maintaining the relevance and effectiveness of demand forecasting models over time. Regularly evaluate model performance, solicit feedback from end-users, and iterate on improvements to ensure ongoing alignment with business objectives.

In conclusion, the challenge of maintaining accuracy in demand forecasting/planning models is a pervasive issue faced by many enterprises today. While traditional ML approaches may struggle to adapt to changing conditions and unforeseen events, innovative solutions like DeepVu's Shock Resilient AI Decisioning Agents offer a promising path forward. By embracing advanced AI techniques, integrating diverse data sources from a rich industry specific knowledge graph, and fostering a culture of continuous learning and collaboration, businesses can enhance the resilience and effectiveness of their demand planning efforts, driving better decision-making and sustainable growth in an increasingly dynamic marketplace.

[deepvuinc]

Embracing the Evolution: AI Planning Agents Assisting Human Planners

In the landscape of supply chain resilience and planning, a fundamental shift is underway—one that challenges traditional forecasting-centric approaches and embraces the transformative power of AI Decisioning Agents. Building upon the insights shared in my previous article, "Resilient AI Planning vs Human Planning," let's dive deeper into this unique relationship between AI decisioning agents and human planners in shaping the future of shock-resilient supply chain optimization.

Reimagining Planning Paradigms

In the realm of supply chain planning, the mantra "demand planning is the backbone of our supply chain efficacy" echoes throughout boardrooms and planning sessions. Yet, as the complexities of global markets and dynamic environments continue to evolve, the efficacy of traditional forecasting models, that have inconsistent accuracy and are unresponsive to shocks, is being called into question.

Enter AI decisioning agents—a paradigm-shifting process that augments the capabilities of human planners by leveraging the power of AI, reinforcement learning with human feedback, digital twin simulations and knowledge graphs.

Unlike conventional forecasting models limited by human cognitive constraints and lacking scenario planning capabilities, AI decisioning agents possess the capacity to process vast datasets, simulate many diverse scenarios, and optimize decisions for business KPIs in real-time. No more focusing on model performance in terms of MAPE metric, but rather on business outcomes (OTIF, freight costs, inventory holding costs, labor costs etc.

Autonomous Resilient Planning in Action

At the forefront of this evolution is the concept of autonomous resilient planning—a framework characterized by the orchestration of multiple AI decisioning agents trained on digital twins that simulate normal and shock scenarios. These agents, informed by a rich tapestry of external signals and a comprehensive knowledge graph, empower human planners with AI-driven insights and recommendations.

Within this framework, human planners serve as orchestrators, wielding AI superpowers to navigate the intricacies of supply chain dynamics and risks. They define the shocks scenarios, current or contemplated, that matter to the business, and let the AI system map them fully to the dynamics of the value chain. Armed with real-time KPI impact assessments and scenario analyses, human planners can make informed decisions to approve/deploy, selecting the most applicable AI agent to address specific contexts and challenges.

Embracing Complexity, Embracing Innovation

The integration of AI decisioning agents into supply chain planning heralds a new era of resilience and adaptability—one where complexity is embraced as an opportunity for innovation. From demand forecasting and production planning to inventory management and order fulfillment, every facet of the supply chain ecosystem stands to benefit from the agility and foresight afforded by AI-driven decisioning.

Moreover, the continuous learning loop inherent in autonomous resilient planning ensures that the system evolves and adapts in tandem with dynamic market conditions and emerging risk events. By enriching datasets with a diverse array of external signals (the supply chain Knowledge Graph), organizations can capture the pulse of the world and proactively mitigate risks before they escalate into disruptions.

Join the Conversation

As we chart the course towards AI-driven autonomous resilient planning, it's essential to foster dialogue and collaboration within the supply chain community. Have you encountered challenges or success stories in deploying AI decision models within your organization? Are there specific use cases or scenarios you're curious about exploring?

I invite you to share your insights, questions, and experiences in the comments below or on our page DeepVu. Together, let's harness the transformative potential of AI decisioning agents and human expertise to forge a more resilient and agile future for supply chain management.

The convergence of AI-driven innovation and human diligence and ingenuity holds the key to unlocking unprecedented levels of resilience, efficiency, and sustainability across global supply chains. Let's embrace the evolution and reimagine the possibilities of supply chain planning in this exciting digital age.

In that light, and given the timeliness of this topic, we’re launching a new “Resilient Planning using AI” podcast in March, and we invite all senior planning executives to comment or  DM me directly on Linkedin or email ceo (at) deepvu (dot)co with their interest to participate.

It’ll be only 19min, with 9 questions that will be shared with everyone selected to interview well ahead of your scheduled date.

Thank you for joining us on this journey of exploration,  learning,  and growth.

[deepvuinc]

Expanding the Spectrum of Supply Chain Shocks and Resilient Planning

The landscape of supply chain vulnerabilities extends beyond the macro-scale shocks that capture global attention. While major events like pandemics and trade wars seize headlines, minor yet impactful local shocks often go unnoticed. Consider the effects of a Mississippi River drought or a local worker strike – seemingly minor incidents, yet capable of disrupting regional supply chains, and costing manufacturers and distributors millions of dollars.

Moreover, shock scenarios need not be restricted to sweeping global occurrences. They can manifest within a company's realm, stemming from last minute unforeseen alterations in customer orders or the bullwhip effect in demand fluctuations. These micro-shocks might seem insignificant on a broader scale but hold the potential to wreak havoc within a specific supply chain, and often cause material impact on bottom line margins.

Furthermore, shocks aren't confined geographically. A disruption in a remote country like Thailand or Vietnam can reverberate across borders, indirectly affecting your supply chain. The interconnectedness of global trade means that shocks in distant regions can transmit ripple effects, impacting businesses far removed from the epicenter.

Navigating the Complex Web of Supply Chain Shocks

Contemplating and simulating these diverse shocks become paramount in resilient supply chain planning. While macroeconomic trends and geopolitical conflicts remain pivotal, the agility to simulate and have AI models that help planners respond to minor, localized disruptions holds equal significance.

Consider a scenario where a tier-2 supplier in a remote location faces unforeseen production challenges due to a regional issue. Even though this seems distant from your operations, it can lead to delays or shortages down the line for your tier-1 suppliers, and your Bill of Materials could change dramatically. Such instances emphasize the need for a comprehensive approach to shock simulation and preparedness.

Fostering Adaptive AI Decisioning Strategies

Supply chains built solely on forecasts are vulnerable to these myriad shocks. The essence of resilience lies in adaptability and readiness. Hence, simulating diverse shock scenarios isn't just about prediction; it's about cultivating adaptable strategies that can withstand shocks, big or small.

By envisioning scenarios that encompass local, regional, and even company-specific disruptions, supply chain planners gain agility in decision-making. This adaptability ensures continuity and minimizes the impact of unforeseen events, irrespective of their scale or origin.

Conclusion

At DeepVu, our commitment to resilient AI planning extends beyond predicting disruptions. We empower supply chain planners to navigate a spectrum of shock scenarios, enabling them to fortify their strategies against the unpredictable. Together, let’s transcend the conventional boundaries of supply chain planning and create resilient AI planning systems that thrive amidst all uncertainties.

Join us in redefining supply chain resilience. #AdaptiveSupplyChainPlanning #resilientAutonomousPlanning

[deepvuinc]

Resilient Demand Planning with AI Decisioning Agents

Navigating today's supply chain planning landscape is more challenging than ever, riddled with unforeseen external factors like climate changes, geopolitical tensions, and labor-related disruptions.

Despite these shocks, manufacturers must maintain accurate demand planning to optimize production capacity, efficiently allocate inventory to DCs, and meet customer service expectations and OTIF delivery targets.

So what's a planner to do?

Enter autonomous resilient planning!

We can take a customer's demand forecast as an input, which maybe manual i.e. simply aggregated from human sales reps estimating which orders are coming by when, or it maybe a simple traditional ML based forecast. Then we add our own external signals data from our VuGraph knowledge graph and perform the shock simulations in our digital twins along with the shock's impact on your own supply chain dynamics. Consequently, the AI Agents are trained to recommend the production quantity levels that optimize for the customers business KPIs for both the nornal scenarion, as well as the shock scenarios.

It's important to note: while AI agents offer recommendations, the human planner retains the ultimate decision-making authority. Informed by the agent's insights and the recommended actions, the planner can make informed choices, carefully weighing their impact on business KPIs.

In summary, the integration of AI-driven decisioning agents into the demand planning process empowers companies to proactively anticipate and respond to disruptions, ensuring a more resilient and adaptive supply chain.

For further insights into how AI can optimize your demand planning, explore DeepVu's website and discover how these innovative strategies can boost your enterprise's resilience, margins, and sustainability. Feel free to email us directly at [email protected] or ceo<at>deepvu<dot>co

[deepvuinc]

Perpetual shocks: can anyone predict the next supply chain shock?

Perhaps you have Yoda on your planning team – he’s got centuries under his belt, he’s seen every conceivable shock in the multiverse and he’s basically a 99.99% accurate crystal ball so perhaps he can predict the next shock. But the vast majority of us mere mortal planners cannot, and kidding aside, Yoda isn’t recruitable!

The C-suite execs acknowledge it’s not realistic to predict a pandemic, an EU war or a continuously morphing trade war, however, they still expect their planners to be ready to mitigate these shocks and have a “resilient” plan that optimizes the business KPIs. The supply chain planners are stressed, the C-suite execs are stressed and the suppliers are stressed!

So what’s the solution?

Clearly, the solution is to actually envision and map out all potential shocks, rank them in order of likelihood, and simulate each one of the top priority ones in your supply chain digital twin. An AI Decision Model would subsequently recommend actions for each one of these scenarios, along with their impact on the business KPIs, and your planners can grade such actions and select which ones are most appropriate per geography, market segment, set of suppliers, set of SKUs etc.

Scenarios, Scenarios, Scenarios

If this brings to mind what military leaders do in their war game simulations, you’d be right on point. Even though, luckily, this is operational planning not an armed conflict,  it’s very much analogous to fighting against an invisible enemy such as a pandemic - causing factory shutdowns and container pile up at seaports, or climate-catastrophe-causing droughts impacting semiconductor fabs, or a crippling components/materials shortage.

Okay so you get it, the solution is a decision model trained on a multi-environment digital twin that can simulate numerous shock scenarios. That seems logical, we can get behind that.

But it’s still missing a key component.

So let's say the overall shock is, to take a very timely one, the war in eastern Europe. It’s clearly imposing global shortages of wheat.  As you can see in the price chart below, Wheat was a $618 $/Bu back in July 2021! On May 31st, it was $1103 $/Bu, and today it's at $774 $/Bu, which is understandable since commodities markets tend to price-in near term forecasted shortages.

Src: Trading Economics

That key component is the supply chain Knowledge Graph those external signals that capture the shocks in near-real-time as they manifest. Such signals such as the primary, secondary and tertiary commodities that impact a given industry vertical. They also include all the macro-economic signals that influence and interact with the industry specific signals such as interests rates, unemployment, CPI and PPI etc.

The more expansive and cross-industry a knowledge graph is, the more impactful its signals and predictive weights.

Role of Governments and Resilient Planning

The most obvious example of this is the latest baby formula shortages due to the closure of one of Abbott’s Similac factories in the US due to contamination. Clearly no one in the FDA simulated that scenario, and everyone was scrambling to find production capacity with other manufacturers to cover the shortage to feed American babies. Ultimately, it took the POTUS invoking the Defense Production Act to instruct the other manufacturers to increase production capacity to meet the formula demand. Luckily this intervention was possible this time, but clearly one can envision scenarios where it may not be, and it could unfortunately lead to loss of life!

That’s the importance of putting scenario-based resilient planning and risk mitigation decisioning at planners’ fingertips so they can see all recommended mitigation actions and their impact on the KPIs, and choose the best decision for the scenario at hand!

Conclusions

Supply chains are truly a matter of life or death and national security at this point, whether they affect infant formula, food commodity shortages, vaccines, silicon components, battery materials, or otherwise. 

It will take public private partnerships and the best resilient planning solutions to address this pressing problem effectively and at scale. It is a top priority for the nation this decade and the decades to come. It's our focus at DeepVu and we're available and eager to partner.