Transport and logistics consist of a mix of work that varies widely in nature. Planning and routing, order processing, document handling for cross-border transport, track-and-trace communication with customers, maintenance scheduling of equipment, and the administrative processing of trips and invoices together take up a large share of office hours. Alongside that is the driving and operational work itself: loading, unloading, driving, physical inspection of cargo. These two worlds do not respond the same way to what AI can take over. Document processing and planning are digital and repeatable; the physical work often is not, or only becomes so after substantial investment in sensors and vehicle technology.
The circumstances that drive the outcome are not the same everywhere. A company that mostly drives domestically on fixed routes has a different planning complexity than a company with varying international loads and customs formalities. A fleet with sensor data from vehicles gives AI something to train on; a fleet without that data does not. Regulations on driving and rest times, customs, and hazardous materials are, moreover, something AI applications must conform to, which slows rollout in places where those rules are strictly enforced.
In planning and route optimization, it is already visible in many places that software does the majority of the calculation work, with a planner assessing the outcome and adjusting it based on knowledge the system does not have: a driver who has just returned from leave, a customer who prefers not to be visited early in the morning. That is the middle of the three patterns running through the entire sector: AI carries out the majority of the work, a human approves or rejects it with reason. Document checking for international shipments is shifting at a comparable pace, with systems flagging deviations and an employee handling the exceptions.
Customer communication about shipment status is already largely automated in a number of companies: a system that reports deviations without a human drafting every message. Maintenance scheduling of equipment follows a different pattern: predictive maintenance based on sensor data can indicate when a vehicle needs attention, but the assessment of whether that maintenance must happen now or can wait for a scheduled stop remains human work as long as the consequences of a wrong estimate are significant. And driving the vehicle itself remains human work in the vast majority of the sector, even where autonomous technology is being tested.
The difference between companies that already organize this way and companies that do not is rarely a matter of ambition. It lies in what data is available, in how much routine is in the process versus how many exceptions, and in how liability for errors is arranged. A planning department with years of trip data can train a system that a department without that history cannot build. That is not a choice one makes, it is a starting position one has or does not have.
A strategy for transport and logistics often rests on an assumption about where the cost advantages and the bottlenecks lie: for instance, that planning is scarce and expensive, or that document handling at border crossings has a fixed turnaround time that competitors likewise cannot shorten. If AI increases that planning capacity or shortens that turnaround time for part of the sector, the playing field shifts without any announcement preceding it. The plan remains standing on paper, while the assumption behind it is no longer true.
This is precisely the pattern also at play in other sectors, each with its own pace and its own trigger: how professional services deal with shifting advisory tasks and administrative processing, how retail handles automated inventory management and customer contact, and what it means for a strategy when an assumption lapses without announcement. The underlying mechanism is always the same: AI is not a separate project with a start date, it is a change in the cost structure and speed of certain tasks that at some point pulls the assumption out from under it.
This is not a statement about what an individual company should do with its workforce. Decisions about roles and FTE numbers are up to the employer, and where such choices are tied to a reorganization, their own statutory requirements apply, separate from what is described here. What this page is about is the question of which work is actually transferable, partly transferable with human oversight, or remains human work, and what that means for the assumptions on which a strategy is built.
The question of which work in your company can actually be taken over by AI cannot be answered at the sector level. That requires looking at your own process, your own data, and your own exceptions; that is what the work scan from FTE TO AI maps out task by task. A first step, broader than just work distribution, is applying the method for mapping a horizon of changes yourself to your own strategy. Those who prefer to start directly with the assumptions themselves can take the free assumption check: a short round in which you name the key assumptions behind your strategy and see, for each one, when it was last confirmed. The full pressure test, with sector data, a self-plot for the management team, and an ongoing assumption list, is under construction.
Stel uw vraag. Vaak zit de echte vraag een laag dieper — daar mag ik naar vragen.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.