AI & the future of logistics

A digital twin is a computer model of a real system, kept up to date with live data. For a supply chain it holds the warehouses, factories, routes, stock and contracts, and how they affect one another. That lets you ask questions you would rather not test in real life. What happens if this port closes for two weeks, if a supplier drops out or if you close a distribution centre? Simulations work through the possible outcomes, and AI helps compare thousands of scenarios to find the measure that best limits the damage from the disruption. The next step is a twin that runs alongside the operation every day, testing for itself what could go wrong and proposing measures before the disruption arrives.

AI combines past sales with open orders, customer promotions, the weather, public holidays and congestion in the ports, and works out how these factors together drive expected demand per item and per week. A distribution centre can then see earlier when a peak is likely and book temporary staff and lorries in good time. A better forecast can help reduce idle stock and costly rush deliveries, and get orders to customers on time more often. The planner adds what the model does not yet know, such as the expected orders from a large new customer. The next step is acting on the forecast automatically: AI agents book lorries and temporary staff within set limits as soon as a peak looks likely.

A control tower brings together the available data on the supply chain: where shipments are, how much stock there is and which suppliers deliver what. Many companies know their direct suppliers well but know little about the suppliers behind them, and that is exactly where the surprises start. So AI also reads beyond the company's own systems: reports of a strike, a flood or a factory fire, ship positions, waiting times in ports. A blocked shipping route affects sailing schedules and distribution centre capacity for weeks; AI calculates that knock-on effect in advance. The next step is a control tower that switches route or supplier by itself when disruption hits, within agreed limits and under human supervision.

An AI agent is given a goal and takes the steps to reach it on its own, within set permissions. Logistics is full of that kind of work: pulling a transport order out of an email, choosing and booking a carrier, checking a consignment note, preparing the data for a customs declaration. Built-in checks decide which bookings the agent may make by itself. At carriers, digital assistants already answer the question of where a shipment is, day and night. Anything out of the ordinary, such as an unusual rate or a query from customs, goes to a person. The next step: AI agents that negotiate price and time slots with the agents of carriers and terminals.

In automated warehouses, mobile robots, storage systems with totes in tall racks, sorting machines and robot arms work side by side. How many orders such a warehouse handles depends on how well those systems work together, including when the mix of orders changes from hour to hour. AI shares the work across the fleet, has robots give way to each other and decides where each product goes: items often ordered together are stored together, and this week's fast movers sit at the front. A large online retailer now has a million robots at work in its warehouses, mobile robots and robot arms combined. The next generation of robots learns from a handful of examples how to pick up a new product, without anyone programming every movement in advance.

Freight platforms match loads with spare capacity in minutes, beyond fixed carrier relationships too. AI does the arithmetic: which load suits a lorry already heading that way, what price is realistic right now, which small consignments can be combined into one full trip. There is a lot to gain: according to Eurostat, lorries registered in the EU drive about a fifth of their kilometres empty. A suitable return load can make an extra trip unnecessary, saving fuel, emissions and money. Soon the agents of shippers and carriers will negotiate on these platforms themselves, within limits set by their owners, and the price for each trip will be settled in seconds.

In a port, a change in arrival time ripples through the use of berths, sheds and towage, and into transport to the hinterland. For a long time every party planned with its own estimate, so a ship waited for a free berth or a lorry queued for hours at the gate. A smart port shares that data and lets AI predict when a ship will actually come in, based on position, weather, tide and congestion. The ship can then slow down on the way instead of steaming at full speed to an anchorage and waiting there. The IMO encourages this just-in-time arrival because it saves fuel and emissions. Terminals plan cranes and shifts around it, and hauliers get a time slot at the gate.

With delivery robots, one remote operator supervises several trips at once. AI lets the robot understand what is going on around it: a kerb, a child crossing, a cyclist passing. Delivery robots have already handed over millions of orders on campuses and in residential areas. Drones fly blood and medicines to clinics that are hours away by road. In busy streets, routes and drop-off points are agreed in advance, and the operator takes over when a robot gets stuck. The next step is delivery at a larger scale: a van as mother ship from which robots and drones cover the last few metres, and lockers and pick-up points that arrange the robot's arrival time with it themselves.

In an AI-first organisation, AI, AI agents and robots take over the work people do not need to do, whether that is sorting parcels or forecasting volumes. People concentrate on the work where their experience, judgement and contact with customers make the difference. Take a customer order. An agent reads it, AI plans the execution and robots do the physical work. The employee picks up the order that does not fit the pattern, solves the customer's problem or has the difficult conversation about price. Around 2040, a routine order could pass through the entire process without human involvement, leaving people to focus mainly on customer contact, exceptions and improving the work.

Automation takes a lot of the number-crunching off the planner's desk, and that changes the job. Instead of planning every trip, the planner checks the system's choices, steps in when there is an exception and improves the rules behind its calculations: the planner orchestrates rather than plans. New roles appear that barely exist today. A fleet supervisor follows ten self-driving lorries on one screen and takes over when one gets stuck. A designer writes the rules within which AI agents may book and negotiate. A trainer teaches robots to recognise and pick up new products. Experienced planners and warehouse staff are the natural candidates for these roles: they know how the supply chain behaves in practice.

In a distribution centre, people and machines increasingly work as one team. The planning system allocates the orders, robots do the driving, lifting and sorting, and people handle the exceptions: assessing possible damage, checking an unusual order, getting a stuck robot moving again. A team like that has its own rules: who may stop a robot, who reports that the system keeps making the same mistake. The shift leader's role changes too: less watching the pace of work, more putting people on the tasks machines cannot do. In highly automated warehouses, the work could shift by 2040 towards training, guiding and repairing robotic processes. People then become the fleet's teachers and its back-up.

A chain that never stops needs leaders who still look ahead. In logistics they take the initiative on AI, robots and AI agents, and sketch a picture of the future so concrete that a shift leader in the distribution centre knows what it means for the team. They make small, tangible changes, involve employees in the plans and choose a direction before everything is certain. That is how they build the supply chain of tomorrow together with their people, and adjusting course along the way is simply part of it.

Change in logistics is about technology, workload and skills, and often above all about trust in a new way of working. Picture a planner with twenty years' experience who knows the drivers and the customers. The system proposes a plan they would never have made. Do they follow it, or put everything back the way they have always done it? That trust grows when the system shows why it chooses something, when the planner may intervene and when it becomes clear afterwards which choice worked out better. The next question of trust is already waiting: will the planner let an AI agent book without seeing the proposal first?

Organisations no longer wait for new recruits to arrive from school with the knowledge they need. They build their own future academy and train their people themselves. Those already on the payroll learn how AI, robots, AI agents and other new technology are changing the work. Newcomers practise straight away with the organisation's own systems and ways of working. Assignments from the company's own processes are the course material, and colleagues learn from one another as well.

New AI applications and shared data bring organisations together with start-ups, scientists, partners and governments from different parts of the world. Together they form an international ecosystem that takes on problems no single link can solve alone, such as congestion, emissions and the shortage of drivers. Elsewhere, parties have worked for years with shared data platforms, often with governments at the table, and with processes that run largely automatically. How well an organisation performs and how fast it renews itself depends more and more on the ecosystem in which it shares data, knowledge and investment.