Agentic process execution could change the P&L far beyond productivity, and the real gains for forwarders sit in execution quality, margin capture and a new kind of moat.
“Innovation and productivity are at the core of what we do.” That was the slogan by a global TMS provider in 2016. In the rear-view mirror, most of the innovation has barely moved the needle from a top-down perspective. Productivity of 40 to 60 files remains a common market result. Even if there were some gains, they might have been absorbed: more reports to be maintained, more customer-specific processes to be executed.
Ten years on, you will find that claim on every billboard again. However, within the last 24 months, first movers are already realising tangible upside. Large incumbents DSV and K&N do not shy away from concrete, public impact commitments of 3-digit million dollar amounts or individual players pretend to cross the 100 monthly files per person mark.
So it is worth reviewing the substance behind, and, more importantly, what it implies in the wider context of the industry.
Tackling the AI potential
Based on our own experience from building Forto and working with many of the leading global forwarders, we reviewed end-to-end processes of up to 100 steps each: the time each requires, its automation potential, and the time that remains manual: the automation estimates came out up to 82%. We see an evolution of the sources of AI automation over time currently (see graphic).
1. Standard automation through data entry into the system (month 0-12): The obvious one, and where many companies start. Examples: turning a customer’s booking email into a booking, updating events or documents or keying commercial invoice and bill of lading data into the system. It sounds simple, but companies will have to build conviction into AI with this first step in our experience, and can’t skip it.
2. Orchestration across fragmented systems and rules to deliver better customer processes (month 6-18): Here the work is not copying data but combining sources and applying logic. Quotation is the classic case. The agent pulls rates from carrier sheets and contracts, rate management systems, adds surcharges, applies customer-specific margin rules, checks capacity and assembles the quote. Other examples are cross-checking documents against the booking, and matching carrier invoices to accruals, orchestrating communication and documentation.
3. Coordinating activities that carry their own process knowledge and, hence, deliver customer value beyond (month 18 and beyond): These build on the first two, but they also need an understanding of the process itself, plus the judgment to run it. Examples: handling a delayed or rolled shipment end to end by informing the customer, orchestrating POs and SKU visibility, rebooking and adjusting costs, and monitoring demurrage and detention before charges accrue.
AI automation finally correlates with other business outcomes
Different than in the past, productivity and automation now works hand in hand with the other goals in the forwarder’s P&L.
Topline and the ability to compete on prices. Obviously faster and more accurate pricing execution has an impact on win-rates. That is not new. However, in the past, productivity gains have rarely translated into pricing leeway. With material AI cost gains, the automation lever has the potential to change the industry. In a tender where $10 on a submitted rate card decides on the actual nomination, a cost advantage of $50, $70 or $100 per file changes the ability to price
competitively.
Profitability. Typical gross profit margins range between 15-25%. That makes the gap between two kinds of forwarders decisive: those who extract the last dollar out of every file, and those who can’t afford to. A larger part of a file’s profit sits in execution, not in the quote or the final invoice. It comes from picking the right partner, watching demurrage and detention closely and keeping accruals actively managed, billing every surcharge. Today that is partly leakage. A well-instructed agent applies the same discipline to every file, it reads the full context from emails, TMS and then follows exactly your process every time, also in peak times or when human operators are flooded with additional work during transport disruptions.
Customer experience. The trade-off was customization versus automation. Agents execute what they are instructed to no matter how standardized the instructions are, so capturing a customer’s context now becomes a one-off task. It enables consistent execution on an individual level. We found in a recent analysis that up to 30% of relevant customer information comes from the customer’s context, such as SOPs, the actual eMail or documents. Building on this context allows forwarders to serve shippers whose requirements used to be too complex to serve profitably, or to replicate their best service levels across regions, segments and products. Even further, capabilities like SOP monitoring in combination with AI exception handling will enable forwarders to get a closer grip on their process and data quality itself.
Fast forward: Potential implications
Together these effects can potentially double the operating result and, therefore, change the economics of the business model, especially for those at the frontier who are implementing. Looking at a typical forwarder P&L, one can hypothesise that about 60% of the impact traces back to productivity itself. The other 40% sits on the top line and gross profit line, through better execution and less leakage.
1. A material EBIT upside that changes how forwarders are valued. The upside is large enough to change the logic by which companies are valued and evaluated, and the gap between frontier players and the rest will show up in the numbers. Should the productivity gain lead to pricing effects, there is no optionality for the broader market to optimize its operational cost base accordingly with AI automation.
2. The automation gains compound end-to-end. There is good reason to believe that productivity gains in this space will perpetuate themselves. Every automated process produces more context, cleaner data and sharper instructions, which makes the next process easier to automate. It also amplifies the risk profile of early choices: point solutions and the lack of an end-to-end view can become a liability, and what gets built in the early stage can turn into legacy surprisingly fast.
3. Standard process execution is not a differentiator. When players can execute reliably, excellent individual execution is no longer a craft that sets anyone apart. It also exposes niche customization: if you have been winning on a high-customization niche, expect competition from players who can deliver the same customization at lower cost.
4. The operator’s role will change. The focus moves from execution to judgment and expertise work. People spend their time on exceptions, as they do today. Many say freight forwarding is a “people business”, and this will continue. But the capability organisations need to build changes. The ability to abstract and conceptualise a process, which is what now goes into the prompts and instructions of the AI alongside them, becomes crucial for successful, widespread adoption.
5. The moat shifts to orchestration and context. From a technology perspective, the moat used to be data availability and a high-quality system of record. Now it is orchestrating fragmented data sources and abstracting process understanding and excellence to a level an agentic co-worker can execute. At the same time, customization becomes an imperative, and companies will invest in and extend its contextual layer. Much of what drives operations today is hidden in unstructured data: notes fields, long documents, tribal knowledge. It will become explicit.
Productivity has been the promise for ten years. What is new is that the gains compound, and the forwarders who build the layers and new capabilities first will have a clear advantage to set the economics going forward.
This article was sponsored by Zauber.
Want to hear more? Zauber founder Erik Muttersbach joined Charlotte Goldstone on The Loadstar Podcast to discuss where agentic AI really stands in forwarding today, why adoption matters as much as the technology, and how to separate genuine capability from the hype. Watch the episode here.
This article is © The Loadstar. Reproduction, rewriting, or derivative use requires a license. Contact [email protected] for licensing enquiries.