The first wave of investments in artificial intelligence was directed towards industries that already had a large amount of software, such as code development, marketing, customer support, and intellectual labor. The next wave will be more substantial as it penetrates physical industries where narrow margins, manual operations, and every unit of efficiency matter.
The reason for this is obvious but critical: the smaller a company's margin, the more the profit from each saved unit of currency is multiplied. Logistics is the most striking example. Consider the economics: the impact of cost savings on profit is inversely proportional to the margin. Thus, the same increase in efficiency yields the greatest benefit where the margin is smallest.
Take an operator with an 8% margin: savings equivalent to two percentage points of revenue increase profit from 8% to 10%, which is a 25% growth. If the same saving is applied to a software-based business with a 25% margin, profit will increase to 27%, which is only an 8% increase. The exact same improvement is three times more valuable in a low-margin business, and in freight operations operating at 3–4%, the same two points double the profit.
The first wave of AI focused on high-margin industries, where savings have the least impact on profit. Low-margin physical industries present the opposite picture: there are no reserves, so almost every dollar saved through AI goes directly to net profit.
This saving must come from somewhere, and in logistics, there are plenty of opportunities. This industry is valued at approximately $10 trillion and still operates based on Excel, email, WhatsApp groups, phone calls, and knowledge accumulated in operators' heads. Drivers communicate with dispatchers, teams monitor warehouses, the finance department manually reconciles tariff cards and delivery documents, and managers constantly solve problems with late trucks or missed slots. The most expensive element of this network is not the trucks themselves, but the operational knowledge stored in the minds of people who leave with every experienced dispatcher.
This manual intermediary layer represents a vast and fragile cost base, which explains the scale of potential savings. Until recently, these savings were unattainable because the work was in unstructured formats that existing programs could not process. Artificial intelligence changes this situation: it is capable of analyzing messages, calls, GPS signals, invoices, and delivery documents, transforming them into structured decisions and acting upon them, transforming logistics from a record-keeping system into an action system.
This is where a sustainable advantage is formed, because the model itself is not a defensive barrier. Advanced models become commodities; the decisive factor for an agent's success is the context: specific business operational objects, exceptions, and feedback loops. A static agent is easy to copy, but one that has accumulated experience from millions of real episodes and corrections over months of operation—not so much. A company that manages to capture this operational knowledge before it leaves creates an insurmountable advantage that no better model can provide.
Furthermore, this changes the sales economy in the logistics sector. When software only records what happened, it is priced by the number of users. When it performs the work, pricing shifts from access to accountability, focusing on achieved results—from resolved exceptions to processed invoices. In practice, deal sizes multiply many times because clients pay for demonstrated business value, not for a set of features, which represents a different and more effective approach than the classic SaaS model.
This also affects customer growth: instead of increasing the number of coordinators, drivers, and auxiliary staff as volumes grow, the operator can handle more shipments without scaling all functions, turning productivity into a lever for growth, rather than just a cost reduction.
For the investor, these three aspects combine into a rare combination: a low-margin, $10 trillion industry where savings disproportionately affect profit; a protected position based on proprietary operational data, not a licensed model; and a pricing model that expands with the provided value. The winners will not be chatbots attached to outdated logistics software, nor general AI companies hunting for vertical integration. They will be platforms that understand the physical operation deeply enough to link data, decisions, and actions, and prove savings at scale.
The desktop era allowed logistics systems to record the movement of goods. The AI era will provide systems to manage the entire network, and in one of the world's largest and most manual industries, this represents one of the biggest opportunities for implementing enterprise AI in the coming decade.