This article is based on the presentation by Asparuh Koev, Founder and CEO at Transmetrics, at the “Logistics meets Innovation” conference. The topic of the conference was “AI in Logistics: from Theory to Practice” and you can read the full overview of the event here: “AI in Logistics: from Theory to Practice” – Transmetrics’ Conference Summary.
Artificial Intelligence (AI) is already a widespread term in logistics, thanks to the implementation of exciting AI technologies like automated warehouses by Amazon, autonomous trucks by Einride, drone deliveries by Zipline, last-mile delivery robots by Starship, and more. These are the types of technologies that can potentially replace some manual jobs that don’t require complex skills (e.g. warehouse sorting, last-mile delivery person, truck driver, etc.).
However, the situation is different when Artificial Intelligence is used to deal with high-skilled positions like logistics planners. That’s where Augmented Intelligence enters the scene. By combining human intelligence with AI, companies can save time, reduce operating expenses, and eliminate manual errors, while human employees can focus more on analytical and complex duties. In this article, we will uncover what exactly Augmented Intelligence is and how it can enhance logistics planning capabilities.
Current Limitations of Logistics Planning in Excel
A logistics planner is not a trivial or repetitive job, but a position that requires a lot of knowledge and experience about what works and what doesn’t in the industry. These employees need to understand complex customer requirements, business rules, and exceptions, handle data quality issues, and much more. Their job also often involves using common sense, which is not the strongest trait of Artificial Intelligence, let alone more fundamental questions about responsibility when AI makes a mistake – is the developer responsible, the dispatcher that used the AI’s suggestion or the manager who implemented the system? Therefore, in its current state, it is best that Artificial Intelligence technologies focus on empowering such high-skilled employees instead of trying to replace them.
Decisions are often made based on manual calculations, the planner’s experience, and gut feeling. That leaves about 10% for potential optimization – not something to ignore in a low-margin industry.
There is plenty of room for technological improvements in the logistics planning area. Today, planners typically use 2-3 monitors with their Transport Management System (TMS) on one screen and Excel on another and that is exactly where the “magic happens.” Decisions are often made based on manual calculations, the planner’s experience, and gut feeling. That leaves about 10% for potential optimization – not something to ignore in a low-margin industry.
While logistics is getting more and more dynamic, Excel doesn’t move fast enough anymore. An actual Excel sheet for planning includes models with 10-20+ worksheets which are, to a large extent, known by one person only, thereby exposing the whole organization to serious risks in case this person leaves the company for any reason. Excel doesn’t let planners refresh their plan very often and it can only process part of the data that exists within the logistics company. The worksheets don’t provide transparency on decision trade-offs and are prone to decision errors which is why logistics companies have to keep capacity reserves — resulting in too many empty trucks, empty containers, and unnecessary line hauls among others. Furthermore, these Excel sheets allow local, not global, optimization decisions – planners in different locations have their own Excel document and decide locally what to do with their service, their warehouse, their depot, and more. At the end of the day, global optimization is not possible.
Intelligent Predictive Alerts for Logistics Planning
To improve logistics planning, companies should use AI, though not in the sense of “Artificial Intelligence”, but rather “Augmented Intelligence”. Augmented Intelligence combines inputs from human planners (experience, responsibility, customer service, flexibility, common sense, etc.) together with Artificial Intelligence technology which is left doing the repetitive and tedious work (using all available data, updating models with recent data, providing fall-back decisions if the planner is not available) and more. For example, complex processes such as optimal carrier selection may take 10 minutes or longer when completed by humans alone, who must sort through hundreds or thousands of routes and schedules. But when that process is turned over to AI, the sorting can be done in just seconds and the final choice is left up to the human operator. In this way, Augmented Intelligence offers logistics planners the best of both worlds.
In its current state, AI helps planners in real logistics operations by using intelligent alerts based on predictive analytics. For example, from sources like MarineTraffic companies can get information about real-time positions and Estimated Time of Arrival (ETA) for every ship in the world based on satellite data. To add real value to the logistics business, AI algorithms can extract relevant portion of data, link it to in-house data which can be very complicated because of low-quality data, detect which events matter out of huge data quantity, formulate a decision proposal, and inform planners about the suggested decision at the right time and place (e.g. as alerts in the system where decisions are taken and implemented, not via email as it often happens.
Other examples of intelligent predictive alerts might include predicting time of arrival for trucks based on traffic conditions, predicting container repair requirements based on GPS tracker shock detection, predicting goods damage and insurance claims based on temperature sensors, predicting high future shipping demand based on multiple variables, predicting warehouse employee sick days based on public holidays and weather, and more.
The next step for AI in logistics planning is the so-called “Human-in-the-loop AI” approach. Transmetrics already offers this with its predictive optimization solutions for logistics. It works by using historical data to train AI algorithms. The decisions from these algorithms are offered only as suggestions to the planners who then need to make a choice whether to accept them or to change them. Afterwards, the algorithms capture final decisions from the planners, compare the outcomes between human and AI suggestions, and use this input in order to further train the AI. This is how companies can keep improving the AI performance and eventually upgrade it to versions 2.0, 3.0 and so on.. With this approach, logistics providers can keep their experienced employees in the planning process, leave them in control, and use their knowledge to improve the AI.
Transmetrics uses the “Human-in-the-loop AI” approach to help a market-leading chemical company with matching customer orders to concrete ISO tanks. This matching happens based on a number of variables such as time to next maintenance, expected round-trip time, type of tank needed, stacking/accessibility, and more. The process consists of three steps: extraction and cleansing of the historical data from the TMS, forecast modeling, and predictive optimization. The AI-driven software called AssetMetrics provides a set of up to 20 optimal tank assignments per order and logistics planner inside SAP chooses which ISO tanks to use. There is a continuous daily learning loop, so yesterday’s decisions from the planners become historical data and therefore a company can compare them to what the optimization suggested and AI learning can happen on the fly. As a result of the project, about 10% fleet reduction benefit was estimated.
The “Human-in-the-loop AI” still has its challenges, which are mainly related to planners accepting the software. Even if the AI suggestions are optimal from the statistical point of view, people still tend to over-correct them, set up excessive capacity safety cushions, add more rules to the algorithms, and try to turn the system into their old Excel. In such a case, AI software will eventually start working exactly as a planner, taking away the potential benefits of AI decision-making and limiting all the extra features it offers. To overcome these issues well-designed AI tools allow unlimited adjustments without pressuring logistics planners to use software suggestions, but at the same time, they closely monitor the efficiency of planners vs. AI algorithms. Typically, acceptance gradually comes in six to twelve months of experience with the system based on Transmetrics records.
“Pilot in the Plane Cockpit”
The ultimate goal for AI in planning is the so-called “pilot in the plane cockpit” concept, in which AI automatically does all the calculations and suggestions and planners only interfere when needed to account for unforeseen factors and strategy, providing additional feedback for AI training. In fact, Transmetrics is already testing this approach for several customers.
The ultimate goal for AI in planning is the so-called “pilot in the plane cockpit” concept, in which AI automatically does all the calculations and suggestions and planners only interfere when needed to account for unforeseen factors and strategy, providing additional feedback for AI training.
In one case, the system helps to optimize empty container logistics in the container shipping industry. The AI-driven software automatically calculates how many empty containers to load/unload at each port, how to allocate the optimal containers to each customer booking, get containers back at the right time at the right location in the right condition in order to give them to the next customer, plan proactively for maintenance, off-hires, grading, and so on. As a result of that, one of the leading shipping lines and a client of Transmetrics achieved 20+% savings through the reduction of costs for handling empty containers – in particular storage and transport costs and a 10% reduction in the number of containers used.
In another case, Transmetrics’ software optimizes dispatching for FTL/LTL orders and steers trucks based on actual and predicted orders for the next one to two weeks. The tool demonstrated the potential to reduce empty repositioning kilometers by 10% and to give 90% precise estimation to long-haul truckers regarding the end destinations for the next two weeks.
In summary, Augmented Intelligence is more powerful than human or machine alone and is likely the best approach for implementing AI technologies into highly-skilled jobs like logistics planning. Empowered with such tools, expert planners will become even more valuable to their companies. Until now, mainly the biggest logistics players could develop AI-driven solutions in-house. With the logistics technology providers like Transmetrics that offer AI predictive optimization tools on a monthly subscription basis, even small and medium logistics players can now get a chance to become more efficient and reduce wasted resources with the latest technologies. Presented concrete use cases and a 10% achieved efficiency gain only prove that.