109 are leveraging the enterprise-grade security of the Azure OpenAI service which guarantees data is not shared outside of the organisation or with OpenAI.” One commonly cited drawback to generative AI are ‘hallucinations’, where the AI delivers misleading or incorrect information as fact. “Addressing hallucinations is a natural part of working with large language models (LLMs) and to tackle these we use a form of templates called playbooks, which guide precise responses,” explains Buddhavarapu. “We are also combining retrieval augmented generation and LLM fine-tuning using Blue Yonder-specific data, documents and model constructs.” So how might generative AI provide added support for the supply chain professional in the future? Buddhavarapu says: “Granting autonomy to generative AI is the logical next step. Time-sensitive problem-solving could be an excellent first use case for granting agency to generative AI for recurring or predictable events.” For example, if a logistics planner is troubleshooting a delayed shipment, they can turn to a generative AI assistant that can rapidly evaluate shipment options and discover a viable alternative based on current market conditions. “The generative AI conducts all the upfront work, and all that’s required from the planner is a final sign-off,” says Buddhavarapu. “Given the magnitude of disruptions in today’s supply chains, this technology makes an ideal tool to address these challenges at scale.” INDUSTRIALS & MANUFACTURING
RkJQdWJsaXNoZXIy NzQ1NTk=