SupplyChainTalk: Empowering demand planners with AI tools

On 10 July 2023, SupplyChainTalk host Alastair Charatan was joined by Hari Srivastava, Senior Lecturer, Auckland University of Technology, Business School; Piero Donaggio, VP & Managing Director, FibonacciLab; Tim Long, Global Industry GTM Lead - Manufacturing, Snowflake; and Vinay P Sharma, Vice President - Product Management Group, Blue Yonder.

Views on news
The European Union has raised tariffs on Chinese electric vehicles, as Brussels takes action to protect the bloc’s motor industry. The new tariffs on individual manufactures range from 17.4% to 37.6%, which is on top of a 10% duty that was already in place for all electric cars imported from China. This is rather controversial, also because the biggest EV imports coming from China are Tesla cars. How this affects supply chains depends on whether Europe can fulfil the demand that cheaper Chinese cars have met up until now.  

The upside of incentivising Chinese EV manufacturers to set up operations in Europe can be a cut on shipping costs. In a tariff free world, competition from China could also incentivise manufacturers in Europe to invest more in R&D to improve their technology. These types of tariffs are not new to demand planning. When changes of this scale and impact are on the horizon, AI-based simulation tools can come in very handy. They can model what-if scenarios, calculate probabilities of what’s going to happen next and suggest ways of how the company can meet new challenges. 

The changing role of demand planners
Demand planning is a complex procedure relying on data from order fulfilment, production control, inventory management, shipping and overall expenses. To make more accurate predictions, businesses should leverage the power of AI. For precise demand planning outcomes, collaboration across functions is essential too. If the AI bit is sourced out to expert firms, a knowledge gap may emerge within the company. But advanced solutions such as control towers can only bring actionable outputs if the data they are fed is properly structured, clean and comes from multiple sources in real time. So, now, for demand planner roles, the skill of using excel sheets is no longer enough but having some understanding of AI and how it arrives at different recommendations is becoming Prt of the job description too. Forecasts in an AI-driven world aren’t made monthly or quarterly but daily – or even multiple times per day. 

There is no ONE AI tool that fits the need of every business.

The right choice depends on the specific problem that the company wants to solve. Prior to the widespread deployment of AI tools, there already were great solutions that could help with interpreting data, spotting trends and exceptions and making reliable forecasts. The novelty about LLMs or ChatGPT is that they don’t rely on human labelling. The new models are now leveraged to make time series forecasting too, which relies on historical data to predict future values. Demand planners today must also be able to interrogate and interpret data dynamically – and gen AI technology is a great tool for that. LLMs are also used to read public sentiment in unstructured data and make predictions about how that can impact demand. Demand planners address the reliability issue of GenAI tools by running different models simultaneously and comparing results. AI can be used by demand planners also as co-pilots. Metrics that can be used to assess outcomes and improvements in efficiency thanks to AI include stock reduction, which can increase up to 15% as a result of AI deployment. Another one is the ability to sell at a higher price at an earlier point along the product’s lifecycle.

Also, you can measure your current outcomes and then use it as a baseline that you can compare your AI-driven outcomes against. In this case, you might want to experiment with low-risk products and use randomised samples. Forecast accuracy is yet another important metric. An increase of even as much as 5% of forecast accuracy can lead to considerable savings –  and “backcasting” (checking whether former forecasts proved accurate against reality) is a useful and easy method to monitor it. Data required for making this “reality check” is available on any company’s ERP system. 

The panel’s advice

• As they say, your job won’t be replaced by AI but by a human who knows how to use it. 
• The use of AI has sustainability implications. Microsoft’s emissions increased by 30% as a result of the company’s growing AI use. However, this can be offset by, for example, using materials with smaller carbon footprints or the efficiencies gained thanks to AI.