Forecasting and AI
In forecasting future demand, the most important aspect in the creation of these estimates is accuracy. With such a wide range of factors contributing to potential variation, it can be difficult for companies to consider all possible variables, along with the additional issue of constantly incorporating real-time data. This is where AI can play a crucial role, with the emergence of, what retail companies are calling, “Forecasting 2.0.”
In the past, traditional forecasting relied primarily on historical data, using Excel spreadsheets or other programs to find future estimates from past results. A frequent issue was uncertainty with the release of new products; often times, such forecasts were based on what companies considered “like” items or goods. This can become problematic as the calculated forecasts would be extremely limited and simplistic, with only available data factoring into the estimate. Forecasting 2.0 fixes this problem, as not only is it able to process much larger amounts of data, but, by using complex machine learning algorithms, it can factor in any sort of erratic event or circumstance. AI algorithms can take weather patterns, seasonality, multiple channel exposure, and more variables into consideration, and are then capable of providing multiple “what-if” scenarios depending on different inputs. Also, the AI are able to analyze all old product data, and then decide which existing goods’ sales are closely related enough in making forecasts of the new product. Perhaps the most significant aspect about Forecasting 2.0 that makes it so dynamic is the ability of AI to constantly adapt to changes in the product mix, changes within the company, or changes in the markets. In other words, the algorithms are self-learning and as new data comes in, the algorithms are able to forecast future results with greater and greater accuracy.
Ultimately, this development is obviously a significant implementation for all companies forecasting demand. As previously stated, the flexibility and learning nature of AI have the ability to make a range of forecasts for any sort of spontaneous event as well as the insertion of newly acquired data into the future estimates. Looking at all aspects of the production process, this development would have extremely positive effects on the company. There would be less chance of surpluses or shortages of products and as a result, companies would see greater profits. Looking at some reasonings as to the importance of forecasts we discussed in class, greater accuracy would be correlated with better supply-chain relations and better management of the number of employee within the company or how to delegate work to different products. It should be noted however, that the second article related to AI notes the importance of still needing a human presence in the forecasting process. Managers within the company can use these scenarios obtained from AI algorithms to perform analyses to make more educated internal decisions, better than simply relying on historical data. AI is emotionless, and there are many matters within a company that deal with intangibles (or factors with no easily calculated numeric value) that an AI couldn’t process wholly into algorithms or estimates. Nonetheless, utilizing AI in forecasting is an extremely powerful asset that, as technology progresses further, will be more and more prevalent within companies. Still, it is not a singular solution to the inevitable uncertainty that comes with forecasts, and human analysts are still needed to help make the best estimates and any resulting decisions stemming from this data.
Sources:
http://multichannelmerchant.com/ecommerce/making-sense-of-forecasting-2-0-and-the-role-of-ai/
https://www.salesforce.com/quotable/articles/how-AI-is-transforming-forecasting-for-the-better/
Any improvement in forecasting is definitely good for companies. As we have learned, there are many ways for companies to forecast, and many of these ways give different answers. I will be following this subject closely as I am very curious to see how the new forecasts that use AI will do. You made some very good points about the fact that AI will not be able to account for the qualitative factors and will only be able to make forecasts purely by a numbers standpoint. This article (https://www.hrtechnologist.com/articles/taxation/4-ways-ai-will-revolutionize-the-tax-function/) mentions how AI will use predictive analytics in order to improve forecasting, further backing up the main point of your article. This article mostly applies the forecasting to the tax industry and predicting when taxes will come in and be paid. Since AI is going to have such a large impact on the future of our economy, I will be very curious to see how AI impacts forecasting as a whole.
Like Zach, I am interested to see how Al turns out. As a society, I believe that we focus too much on algorithms and ignore the importance of straight uncertainty. For example, right now in March Madness there are so many games that are breaking history and beating algorithms. For something as simple as March Madness this is not that big of a deal, however as it starts to affect the future of our economy, this starts to matter. I do, however, find it very interesting that Al uses things like weather patterns and seasonality. For something like predicting tax revenue, this is not something that I would consider being a factor. However, including such detailed information can only serve to make the algorithm smarter and hopefully more effective in the long term.
As artificial intelligence becomes increasingly involved in our daily lives and the business landscape, it is no surprise that algorithms have been implemented in the supply chain process. I think the most critical aspect of AI in supply chain management is the attribute of machine learning. Certain algorithms allow AI to acquire knowledge about the process and commit new strategies to memory. With this data, AI can predict the best methods to forecast future inventory based on trends and exogenous variables. Another interesting trait AI brings to supply chain processes is the future potential of autonomous trucks for delivery. TESLA and Uber do not seem too far off from understanding and putting self-driving trucks into practice. It is only a matter of time until AI really benefits distribution, which, when standardized, could really take away some firms competitive advantages in distribution. In other words, it will be interesting to see when many companies use standardized AI to distribute their products and how that uniform concept affects each company. Nevertheless, AI is a really interesting development that will affect both supply chain processes and overall operations.
Source: https://medium.com/@KodiakRating/6-applications-of-artificial-intelligence-for-your-supply-chain-b82e1e7400c8
As some commenters have pointed out, while AI is an extremely helpful and powerful tool in the new era of business, it should not be totally depended on. Justin mentioned at the end of his post that humans are still the ones making the final decisions, and thus even though a model can spit out some figure, it does not reveal the perfect choice. Another aspect of these AI models is the role of chance. Even though the model may be very predictive in most cases, there can still be times when they fail to predict the truth. Laura mentioned March Madness as an example of the randomness of events. Another example is the 2016 election when most models predicted Hillary Clinton to win, but that obviously did not pan out. FiveThirtyEight, one of the top analytics websites, noted that in the aftermath of the election, many people did not seem to understand the reality of what models do. Even though their model gave Clinton a 70% chance at victory, people can overstate that margin as a certainty. In forecasting models, this same issue can occur where a poor manager could put too much weight in an AI figure and not consider other alternatives.
http://fivethirtyeight.com/features/why-fivethirtyeight-gave-trump-a-better-chance-than-almost-anyone-else/
I think it is interesting how the AI forecasting techniques could be applied to Just-in-Time processes and relates to the idea of jidoka. The Toyota production philosophy also focuses on the ability to reduce waste through producing the correct amount of vehicles and parts. This can be done through customer generated numbers using a pull process of demand but AI might give companies the ability to more accurately predict a customers orders and therefor reduce waste to a further degree. Jidoka emphasizes automating a process as much as possible but still maintaining the human touch and ability to stop the process to prevent defects. I think this connects very well with your point about needing human influence when making these predictions. As technology advances companies will have access to more powerful tools but must not lose sight of the human element of making decisions.