Skip to main content

Tim Cook, CEO of Apple stated the following in Apple’s financial press release on January 2nd, 2019: “We knew we had an unprecedented number of new products to ramp during the quarter and predicted that supply constraints would gate our sales of certain products during Q1”. Supply constraints as one of the most important reason not meeting the revenue targets and subsequent market expectations. The key question: when did Apple know and could they have known earlier?

This is where AI comes in. At many very large organizations I observe a reluctance and misunderstanding of AI and how AI could help optimizing the supply chain. Organizations are in a default ‘firefighting mode’ and constantly surprised by changes in demand and/or supply. In today’s world we observe complex supply chains including thousands of various actors, continuously changing customer and consumer buying habits and preferences, raw material shortages, geopolitical events, heavy competition across multiple channels and an increased pace of innovation. All those individual elements impact the supply and demand, but the elements are also correlated to each other. One element can make the impact of another element stronger or weaker.

A CPG company with revenues around USD 10b employs on average between 500 and 1,000 planners. Those planners are all working on demand and supply plans and are trying to make sense of all the data. They generate massive collection of Excel spreadsheets that is disconnected and standalone, while most decisions are based on historical information or best estimates from Sales, Procurement, Logistics and Marketing. Very often, those planners work in silos, resulting in information-asymmetry and existence of multiple plans without clear reconciliations. No wonder that organizations are not able to give predictable estimates to the market; they just do not understand what is happening… Next to unreliable estimates this also results in suboptimal capacity planning, excess stock, missed revenues, expedite costs, and so forth.

How to solve? This is where AI forecasting comes in. AI, by definition, outperforms humans on many aspects. First and foremost, AI is not biased. As Daniel Kahneman outlines in his best-selling book “Thinking Fast and Slow”: people are biased. Kahneman touches overconfidence, “What we see is all there is”; our brain primarily deals with ‘known knowns’ but is not good in estimating ‘known unknowns’ or ‘unknown unknowns’. Then there is the planning fallacy, as we tend to overestimate benefits and underestimate costs. Other relevant biases are the availability bias (people will reconstruct a story around past events to underestimate the extent to which they were surprised by those events), the way we incorrectly assign cause to random chance and our illusion of understanding. In addition to biases, people cannot work 24/7, they need vacation, they have cognitive limitations with respect to the amount of data and information they can absorb, and the list goes on. Conclusion: human driven forecasts are prone to errors, significant errors.

In 2017-2018, AI made remarkable progressions and beat humans in the following examples (source: AI Index 2018 Annual Report and Forbes)

  • In a 2017 Nature article, Esteva et al. describe an AI system trained on a data set of 129,450 clinical images of 2,032 different diseases and compare its diagnostic performance against 21 board-certified dermatologists. They find the AI system capable of classifying skin cancer at a level of competence comparable to the dermatologists.
  • A Microsoft machine translation system achieved human-level quality and accuracy when translating news stories from Chinese to English. The test was performed on newstest2017, a data set commonly used in machine translation competitions.
  • Google developed a deep learning system that can achieve an overall accuracy of 70% when grading prostate cancer in prostatectomy specimens. The average accuracy achieved by US board-certified general pathologists in study was 61%. Additionally, of 10 high-performing individual general pathologists who graded every sample in the validation set, the deep learning system was more accurate than 8.
  • The average passmark for the MRCGP exam, which trainee general practitioners take to test their ability to diagnose, has been 72% over the past five years. An AI start-up, Babylon Health got 82%. AI had beat human doctors handily.

Back to the supply chain. Think about a CPG company selling cosmetics such as make-up. The demand is impacted by local events (e.g. fashion week in Amsterdam), weather, competitor actions and pricing, marketing campaigns and promotions, competitor ratings, macro-economic developments such as GDP, social media sentiment, new product introductions, new fashion trends, and many more. Humans lack the cognitive capabilities to assess the impact of all these ‘demand drivers’ in a short period of time (if at all). This is where AI is needed to understand the impact and to get early warnings from the market. Similarly, assume there is a constraint somewhere upstream in the supply chain. This could be a transportation route being constrained, a factory not being able to produce to its full capacity or a raw material supplier not being able to deliver. How to manage such a situation taking into consideration a company’s demand priorities (one customer can be of higher priority versus another, or SKU A might be more strategic than SKU B), supply chain networks, demand variability, current orders, desired service levels? Again, this is where AI comes in as AI is able to analyze, in seconds, the impact of all these factors and will recommend the best possible solution based on a company’s requirements (for example the trade-off between service levels versus cost).

The number of benefits and possibilities that AI can bring continues. Imagine yourself assessing the predicted gap versus your target 3 months from now. Wouldn’t it be awesome to have AI proposing which marketing campaigns to run to close the gap? Or to advise what pricing to apply to gain market share? Or which shipments to expedite while in the end still adding to your bottom-line?

o9 Solutions ( has developed a cloud-native and AI-powered supply chain platform that brings real AI to the marketplace. We have unique Use-Cases with some of world’s largest organizations where we apply AI to provide them with a competitive advantage. Many market forecasts predict 50% of the current S&P 500 companies to be replaced in the next 10 years. If your supply chain is still a primarily manual operated process, you will be in trouble. It is time to upgrade your enterprise systems and to embrace AI within your organization. Have a chat with o9 Solutions and let us create that competitive advantage for you! Before it is too late…

Stephan de Barse

EVP at o9 Solutions, driving digital transformation with some of the leading Fortune-500 companies. Intrinsically motivated to solve some of the most difficult challenges with technology with the aim to deliver business value.