Reflections from Gartner Supply Chain Symposium Barcelona 2026

The Gartner Supply Chain Symposium in Barcelona confirmed that AI is now central to the supply chain conversation. But it also showed how early the market still is. There was plenty of discussion about AI, embedded AI, AI-native tools, data accuracy, and automation, but much less clarity on what new AI-enabled supply chain processes will actually look like.

AI was everywhere at the event, but my impression was that many attendees still had a limited practical understanding of what AI is capable of and how it should be applied. The conversation often remained broad: Where should we use AI? How do we add AI into the supply chain? How do we prepare our data? What will AI do for planning? These are understandable questions, but they also show that most companies are still at the exploration stage. The real value of AI will not come from adding a general AI layer across the business. It will come from focused models solving specific planning and execution problems.

Many companies are still spending heavily on modernising their non-AI IT backbone, especially with programmes such as SAP S/4HANA. That matters because these projects consume budget, management attention, and organisational bandwidth. Companies may talk about AI, but if most of their capacity is absorbed by ERP transformation, there is little room left for truly new thinking. This is one reason real change still feels some way off for most businesses.

Some of the strongest presentations came from larger companies developing their own AI and ML tools. These examples were encouraging because they showed that machine learning can work well when applied to specific problems with enough data, internal expertise, and focus. But they also raised a question: how transferable are these solutions? Many appeared highly specific to the company that built them. They worked because they were tailored to that organisation's data, operating model, constraints, and processes. That proves the effectiveness of ML techniques, but it does not automatically create a general market solution.

One useful point from Gartner was that AI does not always need to be perfect. In many cases, it needs to be generally correct and directionally useful. That is an important mindset shift. Human planners often believe they are better decision-makers than a machine. But when planner decisions are measured through research, humans are often found to be wrong around 50% of the time on the first attempt. That is a surprisingly low bar. If AI can produce decisions that are consistently good enough, faster, and with less manual effort, it does not need to be perfect to create value. It only needs to improve on the existing human-heavy process.

Gartner's distinction between embedded AI and AI-native tools is useful. Embedded AI is likely to be the first step: focused AI and ML introduced into existing systems and processes to improve specific decisions. AI-native tools are further away, but they should eventually reshape the process itself. In planning, that probably means merging old fragmented processes into a more streamlined, responsive model. The process cannot change beyond recognition, because supply chain still interacts with the real world: customers, suppliers, factories, warehouses, transport, lead times, and physical assets. But it can become less manual, less fragmented, and more mechanised.

The traditional planning model was demand plan into MRP into supply plan. In practice, this usually meant spreadsheets, manual overrides, planner judgement, and constant intervention to make the plan fit reality. That model was never elegant. It was labour-intensive and heavily dependent on the skill of individual planners. It also fed the monthly S&OP process, which many businesses have struggled to execute properly for years. In a faster, more automated world, monthly S&OP in its traditional form looks increasingly like a relic. A strategic or annual view of capacity and load may still be needed, but day-to-day and week-to-week planning should become far more continuous and system-led.

The larger software vendors still appear to be pitching broad solutions to AI-ify the business, often centred on demand management, forecasting, or generic platform-level AI. That does not feel like the real breakthrough. The real opportunity is in specific supply-side decisions: what to order, which orders to change, how to plan around production constraints, how to plan around transport constraints, what shipments to build, what machines to load, which supplier to choose. These are concrete decision problems. This is where AI and ML can create measurable value.

Gartner's messaging around AI in supply chain still felt vague in places. Data accuracy was repeatedly cited as a major barrier to AI benefits. That is true, but incomplete. Poor data will limit AI. But the bigger issue is often that companies have not defined the problem properly. AI needs a target. It needs a specific decision to improve, relevant data, a clear success measure, and a feedback mechanism. Without that, even clean data will not deliver much. The problem is not just data accuracy. It is unfocused ambition.

Having attended these events over many years, I am aware of how slowly the supply chain world moves and how detached much supply chain design can become from operational reality. I saw nothing in Barcelona that challenged the O8 Organic Planner proposition in terms of reach and automation of ordering. If anything, the market looks ripe for disruption. Most of the conversation is still too broad, too demand-focused, too ERP-constrained, or too dependent on elite companies building niche internal tools for themselves. O8's opportunity is different: focused AI and ML for specific planning decisions, particularly around mechanised supply ordering and execution.

Barcelona showed a market that knows AI matters, but is still unsure how to turn it into practical supply chain change. The companies making real progress are applying machine learning to focused problems with enough data and expertise. The rest of the market is still talking, preparing, or modernising its old IT backbone. The future will not be one giant AI platform making everything smarter. It will be focused models solving defined problems, linked together into a more responsive and mechanised planning model. That is where the real value will come from.