Tuesday, 8 September 2026
Artificial intelligence is beginning to change logistics at a deeper level than simply improving route planning or automating administrative work. Across freight brokerage, warehouses, fulfilment centres and international supply chains, companies are increasingly attempting to build intelligence directly into the systems that decide how goods are priced, routed, stored, monitored and delivered. At Ai4 2026 in Las Vegas, executives working across logistics, data science and emerging computing technologies described an industry where the principal constraint is increasingly not the capability of AI models themselves, but the fragmented information and legacy systems surrounding them.
The discussion brought together Karthikeyan Ilangovan, Vice President of Data Analytics and AI/ML at MODE Global; Pouya Dianat, Chief Product Officer at Quantum Computing Inc.; and Arjun Srinivasan, Senior Vice President of AI and Data Science at ShipStation Global. The panel was moderated by Brendan Baker, partner at Rackhouse Venture Capital. Their discussion highlighted an important shift for the logistics industry. AI is moving from analysing operations after events have happened towards influencing decisions while goods are still moving through the supply chain.
That distinction matters because logistics operates in an environment where relatively small disruptions can quickly create expensive consequences. Weather, port congestion, customs delays, capacity shortages, geopolitical events and equipment failures can alter transportation costs and delivery schedules within hours. Traditional forecasting models remain important for pricing, route planning and demand prediction, but companies increasingly want systems capable of combining information from many different sources and responding as conditions change.
MODE Global, for example, operates across a supply chain involving numerous technology providers, carriers and other partners. A shipment travelling internationally may pass through ports, customs authorities, warehouses, transport companies and different technology platforms before reaching its destination. The challenge becomes particularly significant for temperature-sensitive products such as pharmaceuticals. Moving a container successfully between two ports does not necessarily mean the logistics operation has succeeded if the cargo subsequently remains in customs for several weeks.
AI therefore has potential value not simply in predicting the scheduled arrival of a shipment, but in identifying emerging problems throughout the journey and recommending alternative actions before those problems become expensive. The difficulty is obtaining reliable information quickly enough. Ilangovan argued that model development itself is no longer necessarily the most difficult part of logistics AI. Data quality and integration can present much larger obstacles.
International transportation remains highly fragmented. Some shipments provide sophisticated real-time information, while others move through systems where visibility remains incomplete. Different companies operate different enterprise platforms, transportation-management systems and data standards, creating gaps between individual stages of the supply chain. AI cannot eliminate those underlying problems simply by placing a language model over them.
For logistics companies, building an effective AI operation therefore starts with creating a reliable data foundation connecting transportation-management systems, enterprise software, customer information, carrier data and external information such as weather and market conditions. Freight pricing demonstrates how that information can be used.
ShipStation Global operates across parcel and freight services, connecting shippers with transportation capacity. Pricing a shipment requires balancing the rate charged to the customer with the amount paid to the carrier while maintaining sufficient margin for the intermediary. Historical transportation data can establish how particular routes have previously been priced, but current conditions can change the answer. Weather, available capacity, fuel costs, regional demand and broader economic conditions can all affect what it costs to move the same shipment at different times.
AI and predictive analytics allow logistics companies to combine more of these variables when determining prices. The potential result is an increasingly dynamic freight market where pricing resembles other digital marketplaces. Rather than relying primarily on relatively static tariffs and manual judgement, companies can continuously reassess supply, demand and operating conditions.
Warehouses provide another major opportunity. Distribution facilities already contain substantial automation, but the next phase increasingly combines robotics, computer vision, sensors and AI to monitor the physical operation of the building itself.
Srinivasan described previous work involving predictive maintenance in Amazon warehouse environments, including efforts to identify potential conveyor-system problems before equipment failed. The commercial logic is straightforward. Modern fulfilment centres contain extensive conveyor systems and automated handling equipment. When critical machinery stops, the interruption can affect thousands of orders and workers throughout the building.
Instead of waiting for equipment to fail, sensors and computer-vision systems can identify changes suggesting deterioration. Maintenance teams can then investigate before a breakdown interrupts operations. Advances in vision-language models could expand these capabilities considerably.
Cameras already installed throughout warehouses can potentially become intelligent sensors capable of recognising damaged equipment, incorrectly handled products, inventory discrepancies or unsafe conditions. Drones and autonomous robots could eventually perform some inspections without requiring employees to manually examine large facilities. This could change the economics of warehouse maintenance.
Historically, facility operators have often balanced preventive maintenance schedules against the risk of unexpected equipment failure. AI offers the possibility of moving towards maintenance based more closely on the actual condition of equipment. For warehouse owners and occupiers, that could translate into less downtime, more predictable operating costs and higher utilisation of increasingly expensive automation systems.
The significance for industrial real estate is substantial because warehouses themselves are becoming more technologically intensive. Location, clear height, loading capacity and transport access remain fundamental property characteristics, but occupiers increasingly also depend on power availability, connectivity, automation infrastructure, sensors and sophisticated software systems. A modern logistics facility is consequently becoming both a building and a technology platform.
Computer vision is also creating new possibilities for handling individual products. This is particularly relevant in businesses where goods cannot easily be treated as identical units. Luxury products, second-hand merchandise, customised orders and returns may require inspection, authentication or individual handling.
Ilangovan discussed his previous experience at Neiman Marcus, where specialised products and luxury goods created warehouse processes very different from conventional high-volume retail distribution. AI-enabled visual systems can potentially help identify product condition, classify items, verify characteristics and direct goods towards the correct handling process.
That could become increasingly important as reverse logistics expands. Online commerce has generated enormous volumes of returns, creating warehouses dedicated partly or entirely to inspecting products and determining whether they should be restocked, repaired, discounted, recycled or discarded. Computer vision could automate part of that decision-making process, particularly for apparel, electronics, luxury goods and other categories where condition affects resale value.
Yet one of the most important conclusions from the panel was that new AI applications cannot simply replace the underlying systems on which logistics companies depend. Transportation-management systems remain the operational backbone of much of the freight industry. They contain orders, carrier information, shipment records, pricing and other data required to move goods.
The problem is fragmentation. Large logistics groups may operate several transportation-management platforms because businesses have grown through acquisitions or because different systems specialise in different modes of transport. Parcel delivery, international freight, full truckload and less-than-truckload transportation can all have different technological requirements.
Some platforms are modern and accessible through APIs. Others rely on older integration methods that have existed for decades. Srinivasan described transportation-management systems as systems of record that are unlikely simply to disappear. What may change is the intelligence built around them.
Instead of replacing every existing platform, logistics companies can create a data and AI layer capable of accessing information across several systems. AI agents can then perform tasks that previously required employees to move manually between applications. That could gradually transform the role of the transportation-management system.
The traditional platform remains the authoritative source of transaction data, while an AI layer increasingly becomes the interface through which employees interact with that information and execute workflows. In practical terms, an employee might no longer need to open several systems to find a shipment, check its status, identify a carrier and create documentation. An AI agent could retrieve the necessary information, prepare the transaction and present it for approval.
Over time, some lower-risk processes could become increasingly autonomous. The panel repeatedly returned to the importance of human involvement, however. Logistics remains a relationship-driven industry, particularly in freight brokerage, where long-standing relationships between shippers, carriers and intermediaries influence how capacity is secured and problems are resolved.
Technology can automate routine processes, but replacing those relationships entirely could prove difficult. That makes organisational change almost as important as the technology.
Srinivasan argued that employees who currently perform processes targeted for automation need to be involved from the beginning. They understand the exceptions, informal rules and operational realities that may not appear in process documentation. If automation is developed separately and presented to those employees only after completion, adoption can suffer. If operational staff instead become testers and subject-matter experts during development, the resulting system is more likely to reflect how the business actually works.
Ilangovan described a similar approach based on connecting AI investment directly to measurable business problems. Rather than beginning with a technology and searching for somewhere to use it, logistics companies can identify expensive or inefficient processes and determine whether AI can improve them.
That distinction is becoming increasingly important as almost every enterprise-software vendor adds AI functionality to existing products. For logistics operators, the question is no longer whether a product contains an AI assistant or agent. The more useful question is whether the technology reduces transportation costs, improves margins, increases warehouse productivity, prevents downtime or provides better service to customers.
This emphasis on return on investment could favour logistics because many potential benefits are relatively measurable. A prevented equipment failure has a financial value. Avoiding an unnecessary truck journey saves money. Improving freight pricing affects margin. Reducing manual order processing lowers administrative costs. Predicting a shipment delay early enough to take corrective action can protect both inventory and customer relationships.
That makes logistics one of the industries where AI could move relatively quickly from experimentation towards operational deployment. Quantum computing represents a much earlier and less proven part of the discussion.
Dianat argued that quantum systems could eventually complement AI in optimisation problems involving very large numbers of possible outcomes. Logistics contains many such problems, including vehicle routing, fleet allocation, scheduling and network optimisation. The potential is significant, but commercial applications remain at an early stage. For logistics companies and property investors, AI, computer vision, robotics and conventional optimisation technologies are considerably nearer-term influences on operating models than large-scale quantum computing.
Even so, the discussion illustrates how rapidly the technological toolkit available to supply-chain companies is expanding. The longer-term consequence could be logistics networks that continuously adjust themselves.
Freight could be repriced as market conditions change. Shipments could automatically be rerouted around disruption. Warehouses could detect equipment deterioration before failure. Inventory could be repositioned according to predicted demand. AI agents could prepare transportation documentation and coordinate routine transactions between shippers and carriers.
The physical supply chain would remain composed of ports, warehouses, roads, railways, aircraft and trucks, but the intelligence coordinating those assets could become increasingly autonomous. For industrial and logistics real estate, this matters because the competitive performance of a warehouse will increasingly depend on more than the physical building.
The most productive facilities may be those capable of supporting dense automation, continuous data collection, high-capacity connectivity and increasingly intelligent equipment. This could gradually widen the operational divide between modern logistics properties and older facilities that were designed primarily as storage buildings.
Older warehouses can often be upgraded, but insufficient power, outdated layouts, poor connectivity or limited capacity for automation may restrict what occupiers can implement. AI could therefore reinforce the existing flight towards higher-quality logistics space.
At the same time, technology will not remove the fundamental importance of location. A sophisticated automated warehouse positioned far from customers, workers or transportation infrastructure still faces economic disadvantages. Instead, AI adds another layer to what makes logistics property competitive.
The warehouse of the future will need the right location and physical specifications, but increasingly it will also need the digital infrastructure necessary to connect robots, sensors, cameras, inventory systems, transportation platforms and AI agents. The transformation may ultimately be less about replacing logistics workers or existing software than about creating an intelligence layer over the physical supply chain.
Transportation-management systems will continue recording transactions. Warehouses will continue storing and processing goods. Trucks, ships and aircraft will continue moving freight. What changes is the speed and sophistication with which decisions connecting those assets can be made.
The companies able to combine reliable data, operational expertise and automation could gain an increasingly important advantage. In that environment, the logistics sector’s most valuable asset may no longer be simply its physical network of warehouses and transportation capacity, but the intelligence capable of coordinating that network in real time.
Source: CIJ.World Research & Analysis Team
