Is AI the Answer to No Touch Demand Planning?

    

There is an increasing belief that machines can replace human activities. With supply chain becoming more complex, humans would need to make hundreds of thousands of decisions to stay competitive. Does this mean companies need AI and No Touch Demand Planning?

In this article, we met with Siva Lakshmanan, EVP of Forecasting and Supply Chain Analytics for Antuit, to ask him this question and how it impacts Demand Planners. Siva is a well-known forecasting and supply chain expert who frequently meets with global business leaders.

INTERVIEWER: We’ve heard you speak before and you often mention how decisions to fulfill customer demand can range from repetitive tasks, like validating inventory, to intensive data-based decisions, like predicting demand. Traditionally, we know the tasks supported by ERP and planning tools require humans to make final decisions for both planning and execution activities. What change do you see coming for companies and demand planning roles with the introduction of AI? 

SIVA:  Lots of change is occurring for companies and Planners. Time and time again, we see company executives acknowledging that they are facing greater complexity throughout the entire supply chain. Yet, those same companies continue to rely on Planners who insist on using spreadsheet-based processes and intuition to make most of their decisions. These processes rarely capture the various drivers of demand, as well as the complex interactions between the drivers. Often, only after implementing an AI-powered solution can the real source of supply chain problems become evident. 

However, as algorithms and machine learning can replace many of the mundane tasks and make planning decisions that were once made by humans, from my observations, Planners will continue to play a pivotal role in demand planning. They must arrive at the baseline demand, obtain support from statistical forecasting, and then collaborate with various functions to get everyone to agree on a finalized demand for downstream supply and inventory planning purposes.

INTERVIEWER:  Many of your clients are CPG companies and having an accurate demand plan and forecast is crucial. It’s not the same as retail where they can clearance merchandise, ship it to an outlet store, or return it to the CPG supplier.

SIVA: That is true, CPG companies face different challenges, especially when running promotions. We had a client that ran a BOGO (Buy-One-Get-One) along with a display for one product category. Using a simple, manual approach, the client couldn’t capture the right demand uplift of the promotion or the corresponding cannibalization on the other products. The CPG company never knew what was really happening at the store level, and when they did, it was too late to react. As a result, the client ended up with excess inventory going to waste in some stores and out-of-stocks at other locations. It was a mess. Which is why companies, especially CPGs, are seeking AI/ML solutions to help increase their human planners’ accuracies.  

INTERVIEWER: What is the accuracy of human Planners?

SIVA: Our team has observed that most of the time, the value human Planners added to the statistical demand is a mixed bag. In some cases, the Planners did a better job of providing the right level of insights to improve the forecast; more often, Planners made the forecast worst.

INTERVIEWER: Why is this the case?  

SIVA: For Planners, it’s the large volume of decisions to be made – reviewing 1000s of SKU/location combinations every planning cycle – monthly or weekly. Evaluating high volumes of data is humanly impossible as Planners must go through all the demand drivers for each combination before arriving at the final demand. Because of decision fatigue, planners often make blanket adjustments across categories, or use the classic, “I’ve always done it this way, so will continue to do so.”

You know what can make the problem even worse? Our team has noticed that many demand planners fall into the trap of making their demand plan the same as their budgeted target. They never ask what the customer wants to purchase or how much they will buy. The net result is negative planner value add, which helps no one.

INTERVIEWER: So, we’re back to the million-dollar question – if the value add tends to be negative in most organizations, do we still need Planners?

SIVA: Yes, of course, companies still need Planners.

By incorporating statistical models using AI and machine learning, our clients experience almost instantly a 10 – 15% accuracy improvement. But the accuracy improvement doesn’t automatically replace the role played by demand planners in the supply chain; it does change the nature of their role.  Planners provide real insight based on their field experience and how the forecast will respond. So, their focus will be on market shifts and handling exceptions that are identified by the AI/ML solutions.

INTERVIEWER:  Do you believe companies can achieve No Touch Planning?

SIVA: Yes. Technology is rapidly disrupting and fundamentally changing the business landscape for everyone. I think as artificial intelligence and machine learning continues to transform supply chain planning and forecasting, many of the traditional career positions will evolve alongside the technology.

Companies need human intelligence to enrich the forecast and safeguard against factors that are not systematically captured in the forecast. Likewise, artificial intelligence provides a safeguard against human bias and provides a what-if mechanism to evaluate various drivers and their impact. That’s the sweet spot for companies – taking the best of what artificial and human intelligence can provide to create No Touch Demand Planning.

INTERVIEWER: Thank you Siva, for your insight and time.