At this year’s Automotive Logistics and Supply Chain Europe event in Bonn, one theme cut through the noise: the European automotive industry isn’t standing still – it’s recalibrating to the new reality.
Despite persistent narratives about disruption and decline, what I saw — both in conversations and on the Blue Yonder–sponsored panel “Leveraging Data to Build More Visible, Resilient, and AI-Driven Supply Chains” – was something far more constructive. Not denial. Not panic. Execution.
Across OEMs, suppliers, logistics and technology partners, the conversation has matured. AI is no longer a distant promise or a side experiment. It’s here. It’s being deployed. And increasingly, it’s being embedded into real processes.
The real question is no longer “Where can we use AI?” (solution looking for a problem). It’s rather “Where does it actually improve decisions? Both in speed and in quality?”
Moving beyond visibility
For years, supply chain visibility was the primary objective. And for good reason — many organizations were operating in the dark. Today, that’s changed. Most companies now have some form of:
- control tower
- event tracking
- freight monitoring
- supplier visibility
But here’s the uncomfortable truth. Visibility is no longer the bottleneck. The bottleneck is what happens next. Can you take that visibility and turn it into fast, consistent, cross-functional decisions?
In many cases, the answer is still no. Decisions remain manual, delayed, inconsistent across functions aka siloed, and dependent on individual experience rather than system logic. This is where the real gap sits. Not in data quality and access, but in decision capability.
This leads to a simple but important reframing. Most companies don’t have a data problem. They have a decision architecture challenge.
The missing element: decision architecture
A lot of investment has gone into data platforms, dashboards, and integrations. That’s necessary – but not sufficient.
Data alone doesn’t create value. Value is created when high-quality data feeds decision engines in an orchestrated fashion.
In supply chain, those engines already exist:
- forecasting models
- inventory optimization
- supply planning optimization
- routing and transport optimization
- network simulation
- production scheduling and sequencing
- pick wave optimization & task management
- etc…
The issue is that in many organizations, these capabilities are fragmented across systems, inconsistently applied and disconnected from real-time signals.
What’s missing is a coherent decision architecture that connects:
data → decision engines → workflows → execution
Without that, attempts to deploy AI end up as a layer of insight generation sitting on top of the data – i.e. “next to the process”, instead of being embedded into how decisions are made. And that’s where many AI initiatives stall. AI is not one thing!
Another theme that came through clearly in the discussion: AI is often misunderstood. It’s talked about as if it were a single capability. It isn’t. In supply chain, AI is better understood as a stack of different capabilities, each playing a specific role.
At the foundation, you have predictive AI (Machine Learning): Used for forecasting, ETA prediction, and risk detection.
It’s good at answering: What is likely to happen?
You also have optimization and decision engines: Used for planning—inventory, production, routing.
They answer: What are the best feasible decisions given constraints?
And increasingly, Generative AI and agentic workflows: Used for interaction, explanation, and orchestration.
They help answer: What should we do next—and how do we execute it?
This is where the idea of an “AI layer” often comes in. But in practice, that layer is not replacing planning or optimization.
It’s acting as an orchestration layer—connecting signals, triggering decisions, and coordinating workflows across functions. So rather than thinking of AI in Supply Chain as one tool, it’s more accurate to think of it as:
prediction + decision engines + insight generation + orchestration working together
What generative AI actually adds
There’s a lot of hype around generative AI, so it’s worth grounding its role. Generative AI is not a planning engine. It doesn’t optimize complex, constrained systems. It doesn’t replace forecasting models.
What it does do very well is:
- synthesize large amounts of information
- surface relevant context across functions
- explain outcomes in a way humans can understand
- support decision-making in ambiguous situations
- orchestrate across stakeholders
- act on behalf of users (within well-defined guardrails and business rules)
In simple terms:
- Machine Learning extrapolates and predicts
- Generative AI reasons and explains
What they have in common is just as important. Both depend entirely on data quality and context. Both need to be grounded in real operational systems to create value.
Data: from IT topic to operational discipline
Another shift that came through strongly is how organizations think about data. For a long time, data was treated as an IT responsibility. That model doesn’t hold anymore. Because the quality of decisions depends directly on the quality of data - and the people who understand that data best are not in IT.
For supply chain they sit in: procurement, manufacturing, logistics, and planning.
The organizations making real progress are those that move toward business ownership of data. This is less about governance frameworks and more about accountability at the source.
This accountability drives the next important shift: Instead of fixing the output why not invest the effort in fixing the input and rely on the output.
In many companies, planning still works like this: run a plan, identify issues, and manually adjust.
That approach doesn’t scale. The future model is different. You don’t tweak the output. You improve the inputs to the decision engines. That’s what enables consistency and automation.
Control towers: from visibility to action
The concept of the supply chain control tower has evolved significantly. Early versions focused on aggregating data and creating visibility. Today, that’s not enough.
A control tower only creates value if it:
- provides contextualized insights
- connects directly to decision engines
- triggers actions within workflows
Otherwise, it becomes just another dashboard. And this is where many organizations still struggle – not because of technology limitations, but because of operational integration. The challenge is embedding insights into daily processes and workflows including users plus cross-functional decisions and partner collaboration. This requires alignment across systems, teams, and decision logic. It’s harder than building a dashboard – but it’s where the value sits.
ERP, data platforms, and the reality of architecture
Another topic that came up: cloud transformation and ERP modernization. These are important initiatives. But they’re often misunderstood in the context of AI. ERP systems are designed to record transactions. They are not designed to compute complex decisions or support AI at scale.
Even modernized ERPs largely retain that transactional nature. This leads to a key architectural shift: AI and advanced decision-making require a different data foundation. This is where data platforms and data clouds come into play.
They allow organizations to:
- integrate data across functions
- create a consistent, real-time view of the supply chain
- make that data accessible to decision engines
- make that data accessible to AI for insight generation and contextualization
This raises an interesting question for the future:
What if supply chain applications were natively built on a shared data platform, instead of duplicating data across systems and reconciling different versions of truth? You would have one consistent state, directly accessible to users, decision engines, and AI. This is where things get interesting.
A shift in mindset
Perhaps the most important takeaway from the panel wasn’t technical. It was about mindset.
There’s still a widely held belief of “Let’s get the basics right first—then we can adopt AI.” That sounds reasonable, yet it’s also wrong. AI is not the reward for maturity. It’s one of the fastest ways to achieve it.
AI helps organizations:
- identify data issues
- standardize processes
- improve decision consistency
- scale best practices
The companies moving ahead are not waiting for perfection, rather they are picking specific decisions, applying AI where it adds value, and improving step by step, use case by use case. That’s how momentum builds.
The road ahead
If you step back, a clear pattern emerges.
The next phase of supply chain transformation is not about more dashboards, more data collection, and more isolated AI pilots. It’s about building coherent decision systems.
That means:
- treating the supply chain as a unified decision environment
- connecting data to decision engines
- embedding AI into end-to-end workflows
- aligning ownership across functions
- and executing pragmatically, one use case at a time
In today’s environment, resilience isn’t built on bold claims. It’s built on the discipline to turn insight into action – consistently and at scale.
Curious to learn more? Visit us here.
