Supply Chain | Innovation | Next-Gen Visibility | Collaboration | AI & Optimization | Strategy
OpenAI is about to release the first AI models that feature creative thinking. In essence, the new models o-3 and o-4 Mini are able to come up with their own ideas. The technology might soon come up with new ideas on how to attack problems such as designing or discovering new types of materials or drugs, for example. Let's take a look at how this plays out in supply chain management. OpenAI had shifted towards reasoning-based models last September already when it became clear that the evolution of traditional models was slowing. Reasoning models perform better the more time they spend on processing answers, and they excel in problems with solutions that can be verified objectively, such as mathematical theorems. The two new models are small and cost-efficient, designed to deliver strong reasoning capabilities. The o-3 model was especially designed for complex tasks that includes decision-making in ambiguous or complex scenarios. The model generates a detailed, step-by-step internal analysis through reasoning tokens before producing its answer. Interestingly, OpenAI believes that they can eventually charge $20,000 per month for these capabilities. This is roughly the fully loaded salary of a senior researcher. How does this affect supply chains? More immediate, there are a lot of real-life situations where an operator (or machine) may need to adjust on the fly such as changing routes, consolidating freight, or switching capacity. "Brainstorming" may come in handy when planning complex networks. Models produce counterintuitive results more flexibly and much faster than simulations. The latter are always constrained by a handful of variables, while reasoning models have a lot of latitude in what they consider relevant to decision making. In terms of demand planning, a model such as o-3 would significantly change the game. For example, it can break down soft signals, say a lot of TikTok mentions, into a causal chain and can then make accurate predictions. It can also work with dozens of parallel inventory models for a given set of products or materials to optimize them and adjacent processes from transportation to manufacturing extremely well. o-3 might audit supplier contracts in the future using safety policies, flagging clauses that violate compliance such as forced labor risks and suggest alternative sources in seconds rather than hours. Models already digest news in dozens of languages, can include IoT sensor data, and port congestion patterns to predict delays 14 days in advance with a high accuracy. Given how strong reasoning-based models are in coding tasks, it is conceivable that they may eventually generate much of the supply chain software we use today. In the end, we are still a long way away from these scenarios. However, it is most sensible to think about these things now. #supplychain #truckl #innovation