Nobody at IoT Tech Expo Asked Whether AI Would Change the World
I recently attended the IoT conference in Amsterdam, and I can honestly say that nobody I spoke to asked whether AI would change their world. They know it already has. The conversation has moved beyond tools.
People talked about process, governance, orchestration, and management, and, across it all, where the human stays in the loop. Discussions zeroed in on what stops a project or slows it down and what has to be true before you let it run.
There were five threads in particular worth writing down and sharing.
The Operator Who Put It In Writing
The person I most want to draw on is Alex Duca, who leads automation and operational technology at A.P. Møller Maersk. I have known Alex since my Navis days delivering terminal automation across APMT facilities on three continents. We spent years on opposite sides of the same debate about where a system should decide and where a human should stay in the loop.
What struck me this time was not a tool or a platform. Alex had put out a manifesto to his leaders, his peers, and his team. He would do his own work with an AI-first approach, and people should expect his responses to be both AI-supported and AI-assisted. That is not a budget line or a pilot. It is a mandate, and it carries weight precisely because it came from the top and applied to him first.
Behind it sits a human and machine symbiosis that scales from a single ship to the whole fleet. Autonomy handles the routine, humans hold the critical calls, and the routing is designed to reduce the number of interventions a human has to make at all. Chips and sensors give visibility where there used to be none. The human decides where it counts.
The Incident Is Only The First Ring
The second thread came from a major European infrastructure and civil engineering group working across several cities on critical systems, including transport and disaster response.
Spotting an incident is the easy part. The hard part is everything after it. How do you respond with the facilities you actually have nearby? Then come the concentric rings. The knock-on effect to travel and transport. The pressure on ambulances and emergency services. The impact on energy and the other systems a single event can take down with it. The ambition is not to hold all that information. It is to give command centres clear enough visuals to absorb the richness of the data, turn it into intelligence, and make decisions on action.
Trust Is The Thing Being Built
People want systems that can share intelligence and act on decisions without manual intervention at every step. My background is in integration, and I understand what people are reaching for here. We even had a name for it: swivel chair integration. Someone finishes a task on one system, turns to the next, keys the result in by hand, and the chain continues.
Today's version is more subtle. Systems surface IoT analytics and intelligence; a human reads it, makes a call, and swivel chairs it onward. Some steps are automated, some are AI-driven, but the boundaries between the decisions are still manual and uncodified. If you want genuinely autonomous systems, that is the gap you have to close.
The best analogy I heard all week came from the web itself. Services trust each other through SSL and certificates, underwritten by a certificate authority. Do the same for decisions. Let a decision travel through a business flow, across different data, different models and different owners, as a trusted and decoupled block, attested by a human, that the next step can use without having to understand or even see the underlying data. The shift is from passing data forward to passing trust forward.
And it is not one giant model. It is many small models composing into one, each running its part and handing a trusted decision to the next. It is the swivel chair chain, finally codified, with trust built into every handoff instead of bolted on at the end.
Measure, Then Monetise The Middle
Walk the exhibition floor and the dominant story is hardware. Beautifully built kit with more sensors than you can imagine, and vendors positioning on the quality of the hardware and the data it collects. All of it is interesting. But hardware is a low-margin, high-throughput business, and the ones who stood out were not selling boxes. They were selling the broader solution, an ecosystem of sensors whose value is that you are never reliant on a single data point. Many points, combined, give you intelligence about a room, a road, a factory floor.
That middle layer is where intelligence gets shared and, increasingly, monetised. The pattern I kept hearing was consistent. Combine public data with live IoT on what physical infrastructure already exists and how well it is being used, and you can plan where to place new assets, judge whether an investment is worth it, size the long-term return, and move IoT from a readout of now into a prediction of what comes next. The hardware is the commodity. The intelligence on top is the product.
The What-If Engine
That last thread is still taking shape, and it started in Amsterdam. The city has spent decades engineering defences, making rising water a very real challenge rather than an abstract threat. The conversation naturally turned to where similar protections might be needed elsewhere.
Someone from Google brought up an interesting example: the Kalpasar Project in northwest Gujarat, India. The proposed dam across the Gulf of Khambhat would create one of the world’s largest coastal freshwater reservoirs, providing water for agriculture and growing cities. A road built across the dam would also eliminate the roughly 240-kilometre journey around the gulf, shortening transport routes and reducing the carbon footprint of moving goods across the region.
The conversation wasn’t about building the dam. It was about the method. What data, from Europe and the rest of the world, would let a government and the businesses around it forecast the impact of a civil project at that scale before a shovel goes in the ground? Fresh water is the obvious benefit. But shortened routes open new commercial corridors, and that ripples across segments and industries no single department owns. It is a what-if engine, and like the trust problem above, it runs on many small models working as one, each testing its own slice of the scenario.
Where It All Converged
Every one of these threads lands in the same place, and it is where the floor had already arrived without being prompted. The thing slowing AI down is not the technology. It is the absence of governance, orchestration and management around the decisions.
The surprise, especially in Europe where GDPR and the AI Act get written off as the brake, was that the organisations doing serious work did not treat governance as the bottleneck. They welcomed it. Governance gave them clarity, and with clarity came speed. When you know the foundation is solid, you can build high, build strong, and build fast.
That is the lesson I took home from Amsterdam. Clarity is what lets you move quickly. It is also the problem Axonis was built for. Governed, federated decision intelligence, with the human in the loop and trust travelling through the flow. Nobody needed convincing that AI matters. They needed to know what it stands on.