Much like with gold rush of the 19th century, as the world has embraced e-commerce in early 2000s, many achieved success (i.e. Amazon, Warby Parker, etc.) but even more lost fortunes. And much like in the 19th century, the only group that stands solidly in black are the support merchants. One such company is UPS, a global provider of logistics and small package delivery services.
While historically UPS grew by delivering relatively stable volume of packages to businesses across the US, the e-commerce has changed that forever. In the world of social media where items can go viral in the matter of hours, Prime day can attract hundreds of millions of shoppers in one day, consumer do their holiday shopping later and later, demand for deliveries has become extremely volatile and difficult to predict . With traditionally fixed delivery infrastructure and staffing, UPS has dealt with delivery issues for a number of years by simply hiring armies of temporary workers who were not only expensive, but also unproductive, if staffed in wrong facilities (and often they were) . E-commerce has also given consumers significant power to dictate what their delivery experience should look like, as a result generating an explosion of phone calls and online queries about delivery status and re-delivery requests. Finally, with millennials choosing cities over traditional suburbia, it is not uncommon to see 2 or 3 UPS trucks lurking on the same street at any time, causing ire of criticism from environmentalists and city planners who rightfully blame delivery trucks for most of the congestion issues. It is estimated that delivery trucks drive less than 7% of all miles, but are responsible for 18% of all congestion costs .
Realizing that it could no longer address these issues by relying on traditional tools, UPS has turned to machine learning and AI to understand how it could tackle inefficiencies of its delivery network. While UPS’ first version of ORION system plans drivers’ routes once a day, its next iteration relies on machine learning to make predictions about best route after each individual delivery, taking into account where to best park, which streets to avoid at different times of the day, etc. It is estimated that not only will it save the company $300-$400M, but will reduce driving by several million miles and thousands of metric tons of carbon emissions annually.
As demand becomes more volatile, by working closely with key customers and by relying on AI-fed recommendations, UPS is now trying to predict where exactly in the delivery system delivery demand will be most prominent. Ultimately, AI will be able to decide real time whether packages should be expedited to the air network or can be moved to rail and still be delivered on time with minimal human intervention.
In an effort to improve the responsiveness to consumer queries, UPS has joined a variety of companies in deploying AI-powered chatbots . When customers start interacting with UPS website or app, chatbot is often able to give customers quicker and more accurate answers. UPS’ MyChoice award-winning platform has garnered more than 52M users and actively collects data on shipping patterns, redirection / hold requests, etc.  With the power of machine learning, UPS will relatively soon be able to suggest customers optimal delivery location depending on the delivery day, weather conditions, etc.
While UPS has already achieved some results by employing machine learning, it has traditionally done so by relying on in-house development efforts. The company can go further and faster by partnering with providers who have already made significant advances in this area. For instance, XPO Logistics has also been developing applications and employing hundreds of scientists to develop AI that will be able to better match loads and capture better pricing as market conditions change, it has made significant strides in partnering with vendors who offer AI-powered robotics for sorting and fulfillment (i.e. GreyOrange) . In addition, UPS should also consider using AI for improving its product offering. For instance, AI-powered algorithms may be able to turn packages that are picked up during the day into same-day deliveries (if shipped within the same area). Not only will it reduce travel (driver will typically take the package back to local sorting center only to see the same package end up back on his/her truck out for delivery next day), it will actually allow UPS to sell same-day deliveries (which tend to be higher-value than traditional overnight deliveries).
As we think of the applications that UPS has relied on in its use of AI, few questions remain to be answered:
- Where does greater machine learning opportunity lie for UPS: product-related or operationally-focused areas?
- Should the AI be restricted to UPS-only use or should it turn into an open platform for its customers and partners to improve the entire supply chain?
(Word Count: 799)
 Washington Post, “Amazon, UPS offer refunds for Christmas delivery problems”, Dec 26, 2013, https://www.washingtonpost.com/business/economy/amazon-ups-offer-refunds-for-christmas-delivery-problems/2013/12/26/c9570254-6e44-11e3-a523-fe73f0ff6b8d_story.html?noredirect=on&utm_term=.f78a32815803 , accessed Nov 2018
 Wall Street Journal, “UPS Didn’t Have a Happy Holiday”, Jan 23, 2015, https://www.wsj.com/articles/ups-warns-of-disappointing-results-1422021124 , accessed Nov 2018
 Texas A&M Transportation Institute, “ 2015 Urban Mobility Scorecard”, August 2015, https://d2dtl5nnlpfr0r.cloudfront.net/tti.tamu.edu/documents/mobility-scorecard-2015.pdf, accessed Nov 2018
 UPS Press Release, “UPS Accelerates use of Routing Optimization Software to Reduce 100 million Miles Driven” March 2, 2015, https://www.pressroom.ups.com/pressroom/ContentDetailsViewer.page?ConceptType=PressReleases&id=1426329559785-791, accessed Nov 2018
 MarketWatch, “UPS’s ORION on pace for up to $350M in savings”, Oct 27, 2016, https://www.marketwatch.com/story/upss-orion-on-pace-for-up-to-350-million-in-savings-2016-10-27, accessed Nov 2018
 Forbes, “The Brilliant Ways UPS Uses Artificial Intelligence, Machine Learning, Big Data”, June 15, 2018, https://www.forbes.com/sites/bernardmarr/2018/06/15/the-brilliant-ways-ups-uses-artificial-intelligence-machine-learning-and-big-data/#24d38405e6da, accessed Nov 2018
 Logistics Management, “UPS Announces Global Expansion Plans for UPS MyChoice”, Oct 11, 2018, https://www.logisticsmgmt.com/article/ups_announces_global_expansion_plans_for_ups_my_choice, accessed Nov 2018
 XPO Press Release, “ XPO Logistics to Deploy 5,000 Collaborative Warehouse Robots in North America and Europe”, Oct 3, 2018, http://investors.xpo.com/phoenix.zhtml?c=204615&p=irol-newsArticle&ID=2370062 , accessed Nov 2018