The logistics sector is under pressure: volumes are growing and consumers expect faster delivery. At the same time, organisations are trying to work more efficiently with fewer staff, and see AI as an easy way to speed things up. But on the shop floor, the biggest friction often lies in recurring problems, such as lost parcels, bottlenecks, returns fraud, cyber threats and security incidents. Quick wins therefore do not necessarily start with ‘more AI’, but with a smarter approach: realistic expectations and tackling specific incidents and measurable processes.
Visibility is often the missing link here: only when you can see what is actually happening on the shop floor can you make targeted improvements and demonstrate the impact. Epko van Nisselrooij, Security Advisor at Axis Communications, therefore outlines five misconceptions that, in practice, lead to unnecessary costs and delays.
1) “AI offers new possibilities”
AI offers many possibilities, but its success depends entirely on a clear focus. Many organisations still rely on outdated processes and manual steps that cost time and money every day. There are often quicker ways to realise benefits here than through a large, complex AI implementation. So start with a single, specific problem that has a demonstrable impact. Think, for example, of an operator who spends hours tracking down a parcel, or unnecessary queuing times at the gate. Above all, ask: what does this cost us per week or month, and how can it be improved? Once the problem is clear, you can make a much more targeted choice about which solution – AI or otherwise – is genuinely needed.
2) “More data automatically means better decisions”
More scanners, sensors and systems do not automatically provide better insight. Data is not the same as insight, as you often still lack the context. Sensors can detect a bottleneck, but not the cause. Scanners track parcels until they ‘disappear’, and then the guesswork begins. Even when cameras are in place, systems are often not linked, meaning staff have to compare and search for data manually.
So focus on data with context. Where possible, combine data records with visual context. Start small with one location, one process and one measurement point, and build on that. For example, if you see that a conveyor belt has come to a standstill, without context you won’t know whether a parcel has fallen off the belt, there has been an incorrect input or there is a blockage.
3) “AI must be 100% error-free”
Many organisations believe that AI is only useful if it is 100 per cent accurate in every situation. That sounds logical, but in a warehouse, you often don’t need that level of perfection to make progress. The aim is usually not to prove every single incident beyond doubt, but to identify patterns: where do things go wrong more often, in which locations, and at what times?
That is why insight into patterns is often valuable in itself. For example, if you notice that near-misses are increasing at a particular junction, or that speeding is mainly occurring amongst a specific shift or in a particular area, you’ll know where to intervene with a process adjustment, a change to routing or extra supervision on the shop floor. Combine such signals with simple checks, such as spot checks or brief verification by a member of staff. This way, you move from assumptions to evidence-based improvements, even if the measurement isn’t always perfect.
4) “You always need ‘the latest AI’”
When considering an AI implementation, organisations often think they need the latest versions or the most advanced tools, otherwise it will be outdated. But many logistics improvements do not require complex or ‘trendy’ AI at all. Many quick wins can be found in proven, practical technology that eliminates repetitive tasks and delivers rapid results. Think, for example, of number plate recognition at the gate or congestion detection on conveyor belts. You should therefore choose technology that suits your needs. Overly complex solutions make business cases more difficult and slow down implementation.
5) “Departments will start working together automatically”
Organisations often think: “If everyone sees the value, collaboration will happen naturally.” In practice, things often work differently: suppliers are still frequently approached by just one department. Security receives requests such as “can you retrieve CCTV footage?”, but has limited budget and capacity. As a result, teams continue to work in silos and an investment quickly comes to be seen as a cost rather than a joint business improvement.
So, from the outset, make it a joint endeavour and establish ownership. Who manages the process and the data, and who takes charge of follow-up and improvement measures? This is particularly important because logistics environments increasingly rely on integrated IT and operational systems (IT/OT): without clear agreements on ownership and governance, value remains untapped.
Start by asking the right question
AI can deliver value, but it is not a magic solution. That’s why improvement starts with visibility: knowing what’s really happening on the shop floor, what it costs and where things go wrong. And it is precisely uncertainty – moments when you have to guess, discussions keep coming up or the cause of losses or incidents isn’t clear – that often points directly to an opportunity for improvement.
So start by measuring a single pain point, such as wasted time, incidents, or disputes and claims. Add context where you currently have to guess, make targeted improvements and measure again. So don’t start by asking, “Which AI should we use?”, but rather: “Where are we losing time or money every week, and why can’t we explain it properly?” Once you have a clear picture of this, it’s easier to make choices about improvements, and solutions – whether you use AI or not – are much easier to justify.
This blog was written by Epko van Nisselrooij, Security Advisor at Axis Communications.
