Friday 5 | Chris Burchett
25 Billion Decisions a Day: Blue Yonder’s Blueprint for Agentic Supply Chains
Breaking the Bottleneck is a weekly manufacturing technology newsletter with perspectives, interviews, news, funding announcements, manufacturing market maps, 2025 predictions, and more!
💥 If you are building, operating, or investing in manufacturing, supply chain, or robots, please reach out - aditya@machinafactory.org.com I’d love to chat!
🏭 If you were forwarded this and found it interesting, please sign up!
25 Billion Decisions a Day: Blue Yonder’s Blueprint for Agentic Supply Chains 🎙️💬
"As early adopters lean in and bring agentic teammates on board, they’ll make their human counterparts faster, more efficient, and more accurate as they respond to day-to-day needs. Over the next 12-24 months, more tasks will be delegated to agents. Eventually, agents will be used more autonomously, increasing the speed of decision-making and resilience of supply chains."
From building AI-driven supply chain tools at i2 to co-founding Credant to leading security software at Dell and now shaping generative AI at Blue Yonder, which moments most influenced how you think about connecting digital intelligence to real-world operations?
i2 Technologies was one of my early experiences with AI, and though the technology was limited in its ability to solve real-world problems, it offered a glimpse of the potential that it held. Later at Credant, I led AI-based data classification efforts and product enhancements; this was a pivotal moment in my career, when I saw firsthand how AI can deliver real value. One of the most significant learnings from my time there was how powerful it is to build AI products directly with customers in the loop at every step to ensure that the solutions are delivering measurable impacts once deployed.
Today, my work at Blue Yonder has reinforced the approach of leveraging AI to drive rapid decision-making in line with business objectives and constraints. By the time I joined the company in 2019, AI algorithms and infrastructure were much more mature than in my earlier roles. So, a key part of building our platform was enabling AI development at scale. We wanted to remove cognitive load for data scientists building machine learning (ML) based cloud services and help accelerate the speed of ML-based innovation while ensuring high reliability. Now, Blue Yonder drives over 25 billion AI-based predictions every day, in turn, delivering real-time recommendations unified across operational silos that optimize planning, evaluate risks, and improve agility for our customers.
You have said the next leap is when large language models and AI agents make decisions and take actions in the physical supply chain. What technical and organizational hurdles must be overcome before we trust an agent to reroute inventory or reschedule production in real-time?
As with all technological advancements, the most significant hurdle that organizations will have to overcome is cultural transformation. Building a level of comfort among leadership and staff working with agents first as assistants, and eventually leading to more autonomous decision-making in real-time. Today, leading organizations leverage AI to assist their knowledge workers where it is most reliable, with guardrails and human oversight to drive confidence in AI performance. It’s also important for AI providers to proactively instill confidence in the agents through enhanced data security and quality assurance, as well as regular algorithm training maintenance that reinforces accurate, relevant responses and insights. We’re still a way out from completely autonomous supply chains, but that scenario is the direction that we’re heading. To build trust among organizations in leveraging agents for real-time decision making, it will be essential to offer them the same level of comfort and reliability.
Blue Yonder is rolling out domain-focused agents, such as Warehouse Ops and Network Ops. Can you walk us through a recent challenge your team faced in ensuring accuracy and preventing hallucination, and how you balanced machine speed with human oversight?
At Blue Yonder, we always work with our customers when building agents. This keeps us focused on providing tangible value. The challenge my team often faces isn’t ensuring accuracy in the agent’s outputs, but getting the customer to trust those outputs. When developing the agent, we ensure that it is grounded in proper, contextualized data and effective prompts to ensure accuracy and prevent hallucinations.
Our customers are experts on their own businesses, and they sometimes have a hard time believing that AI could develop that same level of expertise. For example, a customer who works in warehouse management may initially have difficulty trusting AI’s recommendations for inventory placement, especially if the recommendations differ from what was done previously. To address this, we explain the reasons behind the AI’s recommendations and why they may work better than the previous placement within the warehouse.
Human oversight is also critical for ensuring the accuracy of machine speed decisioning. By incorporating real-time feedback loops, such as leveraging a chat with the agent, the user can give the agent feedback on its decision, even something as simple as a thumbs-up or thumbs-down. The agent will remember this scoring and learn from it, enabling improvement over time. By receiving feedback on the real-world impacts of their responses, agents can get smarter and make better predictions and recommendations.
Your platform is built on Snowflake and Azure, and you often speak of the power of three. What have you learned about structuring partnerships so that data, cloud infrastructure, and AI models truly amplify one another rather than creating new silos?
The key insight that I have learned is the power of the right partnerships to co-innovate something truly differentiated in the market. Snowflake and Microsoft innovate with us directly on a unified architecture that is seamlessly built across all dimensions. This is our Blue Yonder Platform, and it brings together the data, cloud infrastructure, and AI models, ensuring that these three elements operate as one cohesive system. This achievement has been a massive effort by all parties, and involved changing roadmaps, building new capabilities, enabling a unified architecture, and driving supply chain innovation – all changes that customers can see and benefit from. Because we build together as the power of three, we can offer customers unprecedented scale, security, reliability, and game-changing functionality that gives them an edge over the competition and speeds time to value. This is how we have achieved the 25 billion AI-based predictions per day for our customers and how we are innovating new features in Gen AI and beyond.
Which capabilities will differentiate supply chain winners from laggards over the next five years, and what steps should leaders take now to prepare their data, talent, and processes for autonomous planning and execution?
Generative AI and agents are still in the early stages of development. As early adopters lean in and bring agentic teammates on board, they’ll make their human counterparts faster, more efficient, and more accurate as they respond to day-to-day needs. Over the next 12 to 24 months, more tasks will be delegated to agents. Eventually, agents will be used more autonomously, increasing the speed of decision-making and resilience of supply chains.
For those who aren’t applying AI to the supply chain, connecting disparate data silos will be a critical first step. This will enable unified decision-making across each supply chain function, from planning to logistics, allowing organizations to see the full impact of a change or disruption. Next, businesses should consider a supply chain network, which offers end-to-end visibility and collaboration to upstream suppliers and carriers. Networks integrate data sources from multiple tiers to provide a complete picture of how a disruption in one part of the supply chain will ripple through the rest of it. AI can then be layered on top of this platform to offer predictions, minimize the impacts of disruptions, and take supply chain management to the next level.
To contact Chris, reach out to him on LinkedIn here. He’s always happy to help and share valuable insights.