Artificial intelligence for supply chain management
AI has come of age.
Behaviors can be predicted once sufficient data has been collected, to then consistently predict and suggest responses and next steps.
These concepts can now even be applied to the supply chain.
So what is Artificial intelligence for supply chain management?
- Any device that can perceive its environment and takes actions that maximize its chance of success at some goal is engaged in some form of artificial intelligence (AI). AI is a loosely defined term that can refer to several technologies. But operations researchers will tell you there is a tendency to refer to algorithms, embedded in a technology, that are less mature as AI, whereas other branches of math and statistics that are mature tend not to be labeled as AI even though they fit the definition.
- In the supply chain realm, machine learning is where most of the activity has been focused. Adeel Najmi, chief product officer at Symphony RetailAI, has a definition of machine learning I like. “Learning occurs when a machine takes the output, observes the accuracy of the output, and updates its own model so that better outputs will occur. Any machine that does this is using machine learning. It does not matter if data science methods are used or not. It does not matter if neural networks or some other form of supervised or unsupervised learning technique is being used. It’s important not to get bogged down on the specific technique. What matter is if the machine is itself capable of learning and improving with experience.”
- When you look at machine learning this way, AI is supply chain management is nothing new. Machine learning has been used to improve demand forecasting since the early 2000s. Demand planning applications rely on a series of algorithms to take historical shipment data and turn it into a forecast. One algorithm works better for promotions, another for end of life products and so forth. The machine looks at the forecast, compares it to actual shipments, and suggests when it may be time to move from one algorithm to a different one for a certain stock keeping unit or product family.
- Over time, many more data inputs have been introduced into the demand planning process, and many companies are doing far more forecasts. For example, instead of just doing a monthly forecast in the eastern half of the country, some companies are doing forecasts at the product/store level at daily, weekly, monthly and longer time frames. For a product being forecast daily at the store level, it may be that algorithms applied to the point of sale data stream have the most predictive power. Forecasting that same product at the warehouse level on a monthly basis, an algorithm applied to warehouse shipment history and warehouse ordering patterns has more predictive power. A forecasting engine with machine learning, just keeps looking to see which combinations of algorithms and data streams have the most predictive power for the different forecasting hierarchies.
- But based on the volume of data used, and volume of forecasts across diverse time horizons and ship to locations, some companies have gone with a fully AI-based solution that takes humans out of the loop.
- Black box solutions are controversial. With a black box solution, planners cannot see into the machine and understand how the forecasting engine is generating the forecast. They must trust the output. AI solutions are more likely to be black box than traditional solutions. Harish Iyer, the vice president of industry solutions marketing at Kinaxis, strongly opposes these types of solutions. “Someone needs to be accountable.” If something goes wrong, and users have just accepted a plan from a black box, how do you hold the planner accountable? And the further up the hierarchy you go the more unacceptable this is. If a public firm misses their numbers in the current quarter, “can you imagine a CEO telling Wall Street they don’t know why?”
- Whether a solution is black box or not, part of the power of AI and machine learning will come from joining it with more traditional BI and business process management technologies. This allows the AI insights to be embedding into a business process, allowing a user to only see the insights that matter to them, and allowing planners to drill down and view supporting information.
- More recently, demand planning applications are working to use machine learning to better incorporate competitor pricing data, store traffic by day of week, weather data to improve demand forecasts, and potentially many other factors.
- Demand planning is a great application for machine learning because these systems have a natural feedback loop. In a demand management application, the system is continuously monitoring forecasting accuracy. That accuracy data in the system allows for the learning feedback loop. Applying machine learning to supply planning is more difficult.
- But machine learning is being applied in many areas in supply chain management. Indeed, there is something of an arms race to leverage machine learning in supply chain applications.
- Suppliers of transportation management systems are also envisioning using weather data to improve their transportation planning. But the promise of machine learning in TMS is broader.
- Work is being done to look at the critical parameters in supply planning, like lead times, and use machine learning to update these parameters. AspenTech is working to use predictive analytics on when key machinery in a refinery will break down to create alternative production schedules.
- Some customers see machine learning as something of a magic wand. They ask suppliers of supply chain applications how machine learning is being used to solve supply chain problems they are facing. But machine learning is just one tool in the supply chain technology tool box. It is not always the right tool. It works best when there is a feedback loop with a measure of success that is clearly defined, when there is big data, and when there are quick feedback loops which speeds learning. In short, machine learning functions best when it is narrowly focused on well-scoped problems.
- One ROI issue associated with machine learning and AI is that we can use AI to search through mountains of data, including new sources Internet of Things (IoT) data, to find new variables that might add some predictive power to formulas. But in many instances, the new variables that are discovered will add relatively little to the accuracy of forecasts. This is one main reason why machine learning problems need to be narrowly scoped.
- AI is also being used in hardware solutions that have relevance to logistics professionals. The autonomous mobile robot (AMRs) is growing very fast. More than half a billion dollars in venture capital has been invested in this market in recent years. AMRs are used to help automate ecommerce fulfillment. AMRs do not follow a predetermined path but can navigate around obstacles. Because they don’t rely on extensive infrastructure – like RFID tags implanted in a warehouse floor – they can be implemented quickly. They rely on an AI technology known as simultaneous localization and mapping.
Want to know more? contact us.