Leveraging Large Language Models to Turn Complexity into Opportunity

This blog was co-written by Sai Buddhavarapu, VP, Product Management, and Angela Troccoli, Senior Product Marketing Director.

Operating in today’s disruptive and complex global environment can be challenging for any supply chain.  Changes in demand, unexpected supply shortages, employee absences, or extreme weather events have always required immediate action. Still, supply chains could rely on the next period of normalcy to get things back on track. With significant disruptions like port delays reducing imports, truckers facing border rules, winter storms slamming logistics networks, workforce shortages, and volatile customer demand swings all becoming the norm – not an exception – resilience and speed of decision-making have never been more critical.

Our customers are faced with the need to adapt to dynamic market behavior constantly. But traversing today’s extreme market complexity is a formidable challenge – and not for all the reasons you may think. Supply chain solutions span vast networks with diverse data sources and are riddled with siloes, redundancies, and inefficiencies. It has become a daily struggle to navigate the massive amounts of data across various systems and teams to make real-time, context-aware plan adjustments. The past few years of extreme ebbs and flows have placed unprecedented pressure on planners, resulting in a record-high voluntary attrition rate among supply chain professionals.

While managing and interpreting vast amounts of data can be daunting even for the most skilled personnel, increasing labor constraints are causing executives across industries to look for a new approach to managing their supply chain problems.

An Intelligent Solution to Our Customers’ Knowledge-Gap Problem

When generative AI entered the scene, this sparked the immediate question, “Could we create super intelligent supply chain assistants to help with our customers’ knowledge-gap problem?” The inevitable answer was, “Yes, but where do we start?”

There are limitations and concerns with generative AI technology, like the hallucination effect, which causes the AI to deliver an incorrect response due to its inherent creativity and the concern about data and information security.  In addition, negative perceptions in the industry caused by past failed attempts, poor user experiences, and overly prescriptive results have raised the bar on customer expectations. We have to work within the confines of these existing challenges while increasing planner productivity and upgrading the user experience.  Blue Yonder’s approach ensures that data security and accuracy of responses are built right into the architectural foundations of the solution.

Today’s user experience across enterprise applications is built to accommodate all solution use cases, forcing Users through a predetermined set of screens built for a prescribed journey to reach their desired answer or end state. This makes the user experience rigid, inefficient, time-consuming, and even more odious when you throw vast amounts of data into the mix. It’s no wonder supply chain professionals are getting fed up. Today’s environment requires supply chains to be resilient and flexible, with the ability to initiate a business process from anywhere and synchronize decision-making from end to end. As we move toward this new world, the user experience must evolve to suit.  

Our objective was to turn market complexities into growth opportunities by:

  • Streamlining data analysis and scenario planning,
  • Predicting downstream strategies,
  • Automating tedious tasks, and
  • Unlocking capacity so planners could spend more time on strategy, what-if scenarios, and executing outside the norms.

Imagine if you could cut the lead time for your supply chain in half, predicting market trends with razor-sharp accuracy months in advance. How might this level of precision and speed transform your market competitiveness, inventory management, and overall profitability? How would this improve a day in the life for supply chain professionals?

Blue Yonder Orchestrator is Delivering the Next Generation Supply Chain Decision Intelligence

With today’s applications, a planner must perform hundreds of permutations and combinations to create perfect plans. Still, this goal remains out of reach due to the sheer number of anomalies and volume of data. Blue Yonder is bringing together the natural language capabilities of LLMs and the depth of Blue Yonder’s supply chain solvers to address this challenge, delivering dynamic decision-making and orchestration. Integrated within the Luminate® Platform, Blue Yonder Orchestrator allows users to query in natural language, then pulls data from all relevant and permissible sources, determines correlations, and serves up in an easy-to-consume format all relevant data along with insights and recommendations to guide the next steps and decision making – all without the hassle of wrangling data or toggling between multiple software applications. The Blue Yonder Orchestrator executes the decisions using the vast array of supply chain planning and execution microservices.

Leveraging Generative AI to Transform from Predictable to Prescriptive

A generative user experience is not one-sized for every problem. It adapts depending on the situation and context, bringing together information and the necessary constructs to solve the problem. Large language models (LLMs) must also answer questions correctly – considering available data, user roles and permissions, and context within the supply chain application. And to avoid the hallucination effect, guardrails must also be established. There are thousands of questions a supply chain user can ask and millions of ways to ask them.

To meet these requirements, Blue Yonder applied prompt architecting, a blend of prompt engineering that focuses on composing, priming, refining, and testing prompts within a single conversation and consideration for the overall architecture of LLM operations ensuring accuracy and data security. The models are exposed to proprietary algorithms, information on how Blue Yonder’s supply chain software works across all use cases, and Blue Yonder intellectual property, including 400+ patents, then connected to Blue Yonder’s applications and universe of data. The result: high-quality, super intelligent, all-knowing supply chain navigator available to enhance the user experience, augment intuition and accelerate decision-making.

Let’s consider a common scenario. Today, a planner reviews their sales and margin plan by month. Based on prior experience, they anticipate a spike in demand in Month 8, but to be sure, they want to know the sales actuals, by quarter, for the prior calendar year and therefore open another application to dive into it. They then want to analyze seasonality trends and open another tab. They then switch to internal BI systems to understand what products drove the performance.  Then they get a phone call from a colleague requiring immediate attention. And they have to return to this task later in the afternoon. Sound familiar?  

Now let’s apply Blue Yonder Orchestrator to this exact scenario. The planner is reviewing their sales and margin plan. They open their chat and enter the prompt: “Show me the sales actuals by quarter for 2023 and 2022.” In the same window, they are provided a table summarizing the figures. They are also provided insights, including high, low, and average sales volumes. Planners can delve even lower to understand which products and regions mainly contributed to the performance. Blue Yonder Orchestrator notes that sales tend to be higher in Q3 and Q4 and recommends that the planner take advantage of peak season, planning inventory and promotional activities accordingly. The Blue Yonder Orchestrator has scoured all available data sources to which the user has access, presented a summary of the pertinent information along with recommendations, and done it in seconds. Rather than toggling back and forth between applications, the information is at the planner’s fingertips.

And on the horizon, imagine the planner entering the prompt: “This summer will be longer than usual. I want to increase men’s short sleeve shirts and shorts sales by 10% for Months 8 and 9. Please make the change and regenerate the plan.” And the Blue Yonder Orchestrator, having been granted this agency, completes the task. What once would take hours now takes mere seconds.

Blue Yonder is Investing Today to Deliver Customers Benefit Tomorrow  

AI allows for quicker processing and analysis of large volumes of data to help accelerate the decision-making process and, as AI gains broader understanding and adoption, we expect to see more and more companies applying it in these and similar use cases. By utilizing generative AI technologies, we will transform the user experience with guided decision making and process automation, taking our solutions to the next level so that our customers can be better prepared to plan and pivot supply chain operations as quickly as the market evolves.

While optimizing the user experience is a great place to start, it’s just the beginning. Generative AI technology is still in its infancy, and there are many unsolved problems it can tackle – once the tech is ready. Any off-the-shelf technology will require further tuning to deliver meaningful results, and this refinement requires both investment and skilled engineers.

As the technology continues to advance, we predict new use cases will emerge that merit such an investment, and Blue Yonder also looks forward to being at the forefront of this next chapter.

Want to learn more? Watch our Blue Yonder Live session on Aug. 17, 2023, at 1 p.m. ET.