ny retailers are still lagging behind in integrating applications and hardware into their supply chain operations.  In a Coresight Research October 2023 survey of more than 400 retail decision‐makers across North America, Europe, Australia and New Zealand, only 27 per cent of respondents identified AI and machine learning, including Gen AI, as the advanced technology best suited to help proactively identify unexpected costs.  In an era of unprecedented complexity for the retail industry, influenced by geopolitical disruptions, labor shortages, and sustainability pressures and lest we forget tariffs, Poon argued that integrating Gen AI into a company’s supply chain system will be top of mind for retailers over the next few years. Coresight Research estimates that the global GenAI applications and hardware market will hit a total of USΩ125.0 billion, US$39.6 billion for applications and US$85.4 billion for hardware, by the end of 2025. The research firm predicts that YoY growth for the total market will reach 57 per cent in 2025, marking a major expansion of GenAI use, especially with applications, which are expected to see 106 per cent YoY growth in the same year.  By 2028, Coresight Research anticipates the market to reach US$235.5 billion, with YS$87.9 billion allocated to applications and US$147.6 billion to hardware. What Are the Benefits of Gen AI in the Supply Chain? For retailers still unfamiliar with the term, generative AI (GenAI) refers to the use of large language models (LLMs) to produce human‐like text and images or to simulate scenarios from unstructured data. Gen AI can help retailers generate product descriptions and perform vector search to find vendors that match product requirements.  Additionally, he noted that GenAI technology can be applied to a broader range of AI technologies within the supply chain context, simplifying typical AI/ML processes, which also extend to agentic AI and physical AI.  For example, while some retailers have already integrated traditional AI/ML to analyse data to create insights for decision making, Gen AI takes things up a notch. It enables more accurate demand forecasting by utilising both historical and real-time data within a chatbot dialogue, eliminating the need for data science expertise. With agentic AI, Gen AI enables supply chain agents not only to handle routine tasks but also to perform more cognitively complex duties, such as negotiating with suppliers, generating purchase orders, or rerouting shipments. When Gen AI is properly integrated into a retailer’s operations, along with agentic AI or more traditional AI applications, these combined technologies enable forecasting, scenario simulation, negotiation, routing and physical automation, which previously required multiple systems or manual intervention. Factors to consider before implementing GenAI While GenAI presents notable areas of opportunity to reinvent the retail supply chain, such as cost savings and enhanced customer experience, success is not guaranteed with this technology. Coresight’s Poon noted that companies must determine the strategic fit based on their assessment of the strategic alignment, segment suitability, organisational readiness, risks and obstacles. Here are five steps Poon recommends retailers need to go over before implementing GenAI into supply chain operations: Strategy Poon recommends that Gen AI adoption should be driven by strategic fit, in line with Coresight’s A.S.P.I.R.E. roadmap. A Assess  Retailers must assess strategic alignment, organisational readiness and segment suitability and identify high-impact pain points. S Select partners Choose vendors with domain expertise and responsible AI practices.  P Pilot Identify the business case, estimate potential value, define success metrics and monitor them against existing manual systems. I Integrate and scale  Integrate GenAI outputs into supply chain workflows, then expand successful pilots across categories and geographies.  R Reflect  Set up feedback loops to update models, incorporate new data sources and monitor emerging technologies. Poon also suggested that retailers can leverage insights from early adopters while staying informed about evolving regulations.  E Ethics and governance Retailers must implement data stewardship and compliance checks, in addition to fostering a culture that encourages teams to experiment safely with AI and learn from their failures.  Technologies The right combination of technology, data, and solution partners, under the right circumstances, is the key to launching a successful GenAI Implementation. If done correctly, GenAI can enhance the AI capabilities of technologies within a supply chain context to generate texts and images, sustain conversational AI with supply chain data, enable agentic AI to take actions, and empower physical AI to allow robots to perform human tasks. Operations Identifying the areas that can be automated by GenAI and applying the right technology are key to unlocking the potential of supply chains. Risks and resilience Retailers must determine the level of autonomy to assign to AI systems in their supply chains and mitigate risks before entrusting them with autonomous power. Until that time, human-in-the-loop should be the basis for AI deployment. Ethics and compliance The power offered by GenAI systems, particularly when agentic AI and physical AI systems are trusted to act autonomously, could potentially cause economic or even physical harm if they act incorrectly or due to malicious actions.  The risks of integrating Gen AI in supply chain operations  “Agentic AI connects to the entire system of an organisation, which enables autonomous task completion,” Poon noted in his report. “As such, a cyberattack that compromises an agentic AI system could be extremely harmful to an organisation.” He also warned retailers that gen AI systems and AI agents are susceptible to new types of cyberattacks, such as a “prompt injection”. In standalone LLMs, prompt injection poses a potential hazard that can leak confidential information or customer data, resulting in significant legal consequences for organisations. However, prompt injections do not allow cybercriminals to break into a company’s system because LLMs have no authority to access company systems. Whereas agentic AI systems can be authorised to access a company’s internal systems, presenting a much larger safety hazard.  “Retailers should undergo a step-by-step approach in implementing GenAI transformation. Jumping right into agentic AI and physical AI for full automation could bring unforeseen risks,” Poon warned.  Before moving to full AI automation, retailers need to ensure security and resiliency in the piloting phase, he concluded.