Being the most noticeable âproductâ of the digital revolution, e-commerce could not but follow the rapid revolution-driven steps forward. Unsurprisingly, therefore, online merchants have been investing in AI technologies designed to improve consumer experience and increase sales. Interestingly, the intended âconsumer experienceâ improvement mainly translates into making shopping less burdensome for consumers, which in turn boils down to easing or even delegating certain tasks inherent in shopping to software. Examples in the media are numerous and confirm that the consumer no longer shops unaided. More specifically, having referred to âadvancements in AI like deep learning, in which software programs learn to perform sometimes complex tasks without active oversight from humansâ. Kelleher (2017), writes the following:
Gilt deploys it to search for similar items of clothing with different features like a longer sleeve or a different cut. Etsy bought Blackbird Technologies last fall to apply the firmâs image-recognition and natural-language processing to its search function . . . Adobeâs marketing tools are also incorporating deep learning into offerings for retailers, using AI to predict customer behavior. Shoppers can choose to receive suggestions based on their shopping lists â a belt that matches a pair of pants, painting supplies to help with a can of paint, a wine paired to a dinner recipe.
More generally, the prevalent trends seem to be (a) natural language processing that enables precision in search results,1 (b) personalized product recommendations that rely both on past browsing or purchasing activity and on the analysis of an abundance of relevant data from across thousands of sources2 and (c) smart shopping assistants using visual filters3 or image recognition4 to understand consumer preferences and return accurate product results and information as to availability in real time.
1 âYou can search for phones or other electronics goods just like you would ask your friend for a recommendation and not like a robot that has been fed a list of terms to search byâ (Choudhuri, 2017).
The latter seems to compose part of the so-called âomnichannel approachâ,5 which rejects the distinction between âofflineâ and âonlineâ consumers. Based on the fact that consumers increasingly shop through both offline and online channels, this approach mainly advocates the use of technology effectively connecting the website of a retailer to its brick-and-mortar store, ensuring consistency and best consumer experience (Piotrowicz and Cuthbertson, 2014). Giant online retailers such as Amazon turn to the offline world opening up traditional stores (Fortune, 2017), and eBay seeks to integrate itself with offline, stores even directing to them consumers who cannot wait for delivery (Kopytoff, 2010; Rao, 2011). It seems that the omnichannel approach is bound to increase sales on or through the internet effectively encouraging resort to intelligent tools assisting consumers to navigate the vast online and offline marketplace.
Such tools include notification systems incorporating price-prediction algorithms that can notify consumers about upcoming price reductions effectively, advising them on the best time for a purchase (Master of Code Global, 2017) and e-commerce platforms that place unprecedented emphasis on assisting consumers effectively searching for and choosing the right product. A clothing and footwear platform, for example, prides itself for using massive amounts of connected data and machine learning to enable precise fit and size recommendations (True Fit Corporation, 2018). This seems to mark an important shift from âproductâ searches to âsuitable productâ searches enabled by the provision of very detailed purchase-related information.
Comparison tools returning to consumers a list of competing offerings relating to a specified product enabling them to locate the best prospective deal, have been around for quite some time, yet they remain a trend. Not only they have become mobile and thus, in the form of apps that are accessible anywhere and everywhere but also their use has been boosted due to the omnichannel approach mirroring consumers engaging in comparisons on their mobile devices while in store (Research and Markets: 2015, cited in Business Wire, 2016). Moreover, they are increasingly equipped with machine learning and natural language processing capabilities, thus becoming smarter and more accurate in their output (Choudhuri, 2017). They are also able to assist consumers in finding a specific product (such as a good mobile phone with long battery life) rather than merely comparing prices. This involves the relevant tool using AI to sift through and make sense of âover 50,000 phone specifications, 5,000 phone variants, and over one million user reviewsâ (Raj, 2015).
2 âWhat sets Watson Trend apart from other product recommendation resources is the variety and magnitude of the data it sources. The app pulls in product information from more than 10,000 sources, from major ecommerce sights, blogs, product reviews and social mediaâ (Clark, 2016).
3 âInstead of typing into a search bar or checking filter boxes, the tech learns what the customer likes by analysing the products they click on. Selecting a pair of heeled boots, for instance, brings up more heeled boots. Judging by the images that attract the customer, the retailer can then determine that she might be interested in black rather than brown shoes, or shoes with laces rather than ones without, endlessly bringing up more styles that match their preferencesâ (Sentient, 2015)
4 âFor instance, a budget-savvy customer at a brick-and-mortar store sees a pair of pants she likes but wishes to compare prices online. However, she doesnât know what those specific pants are called or where to start. Instead of wasting time manually searching for them, she can simply take a photo using her phone and upload it to a retailerâs online store or mobile app. AI technology automatically searches for the item and quickly pulls it up on the customerâs phoneâ (Angeles, 2017).
5 On the history and development of the omnichannel approach, see Frazer, M. and Stiehler, B.E., 2014, January. Omnichannel retailing: The merging of the online and off-line environment. In Global Conference on Business & Finance Proceedings (Vol. 9, No. 1, p. 655). Institute for Business & Finance Research and Lazaris, C. and Vrechopoulos, A., 2014, June. From multi-channel to âomnichannelâ retailing: review of the literature and calls for research. In the 2nd International Conference on Contemporary Marketing Issues (ICCMI).
One observes that most of the emphasis of the recent consumer-oriented technological developments in e-commerce has been placed on discovery, i.e., the easy and effective search for a desired product and the merchant whom to buy from. This is logical: âAs the web itself does not provide for search functionality, search engines have become an essential gateway to everything that is available on the web. Without them, the web is like a library without a cataloguing system and librariansâ (Kulk, 2011, p. 7). The same is true of the retailing area of the web inhibited by the abundance of e-shops. It is even true of certain e-shops themselves, as some sell an almost unnavigable wealth of products. Thus, most e-shops prominently display a search tool on their pages. Without a search aid, it is virtually impossible for consumers to find everything related to a need or interest online. It is thus only natural that technology (i.e., natural language processing, image recognition, rich-data-driven product recommendations and comparison tools) has mainly been geared towards easing the consumer search-and-discovery experience.
One also observes that despite their AI capabilities, all of the aforementioned e-commerce advancements still necessitate consumer involvement; the consumer remains tied to a screen having to point and click or touch throughout the process in order to find and eventually buy a product. Interestingly, such beginning-to-end consumer involvement (or active participation) seems to be a constant variable through the evolutionary journey of userâmachine interaction; what changes is the nature of said involvement and, inherently, the effort required on the part of the human consumer. De Kare-Silver (2011, p. 18) describes the said evolutionary journey nicely:
So we are moving away from Point and Click. We are firmly into an era of âTouch and Go!â. It is only a matter of time before further voice integration takes us into âTouch and Talkâ. But is that the end of the journey?
He answers this question in the negative, describing upcoming technologies, amongst others, allowing machine control through effortless hand gestures (De Kare-Silver, 2011, pp. 18â19, 21). Those, however, still require a consumerâs active involvement.
Indeed, the current e-commerce technological trends lack automation in the actual contract conclusion. Such automation involves the machine not only looking up and finding a product, its price and the retailer but also actually buying it, thereby concluding a contract in the name of the consumer. This is the missing piece of a puzzle picturing a buying process in which the consumer is (fully) substituted by software. Importantly, as illustrated later in this chapter,6 there is strong evidence to suggest that marketplaces in which selling and buying software will conclude real automated contracts binding upon their human users will soon be the next big thing, which the media will be talking about. Relevant prototypes have existed in labs for more than two decades,7 and it often takes much longer for a technology to be commercialized, and even longer to become widely available.8
6 See infra at p. 12
7 Ibid.
8 De Kare-Silver (2011, p. 17) refers to the first voice recognition technological component being developed in the 1964. The relevant technology became widely available in 2003 when Microsoft developed its own such technology and incorporated it in Office 2003.
Software designed to negotiate and contract on behalf of human users has been mentioned in seminal technical literature of the late 1990s. MIT researchers identified, amongst others, five main stages of the consumer buying process: need identification, product finding, merchant finding, negotiation and purchase (Guttman, Moukas and Maes, 1998, pp. 148â149) and pinpointed to technology, specifically software agents, which could assist or represent consumers during all those stages. Notification agents notify consumers when a specified product becomes available, while recommendation agents recommend products that a given consumer may be interested in. Shopping agents (or comparison tools) play a vital role during the product- and merchant-finding stages and finally, negotiation agents undertake the final stages of negotiation and purchase.
As a result, during the late 1990s and 2000s, there was a growing body of literature on the technical and legal aspects of software (or intelligent) agents.9 Such literature has become scarce in the 2010s, yet one readily notices that the so-called notification, recommendation and shopping agents simply aimed at doing what has earlier been described to be the core of current e-commerce trends, namely, intelligently notifying consumers, offering product recommendations and comparing product offerings. It would seem, therefore, that albeit by a different name, 10 the technology described in the 1990s is now being commercialized and widely discussed. Obviously, this reinforces the view that more advanced technology capable of taking over the whole of the buying process, thereby completing a purchase (i.e., what was called a ânegotiation agentâ in the 1990s), will probably be ready for a commercial roll-out in the (near) future.
It also follows that the current scarcity of literature on software agents does not mean that they have ceased posing legal challenges, amongst others, relating to the protection of consumer economic interests and privacy that merit (further) examination. Actually, these challenges have never stopped being discussed in relevant (legal) literature and major studies, the difference being that the relevant discussion has largely been unfolding by reference to their underlying processes rather than to a technological product label, such as ârecommendation agentsâ.
Thus, relying on the collection and use of detailed data,11 notification and personalized product recommendation tools pose issues of data protection and privacy heavily discussed in literature on the so-called âbig dataâ12 and its legal ramifications. Yet, long before the birth of the âbig dataâ concept,13 the consumer was known to be tracked and monitored on the web though the collection and analysis of data on her browsing behaviour. On the basis of these detailed consumer profiles, the consumer was served personalized advertising (often in the form of âproduct recommendationsâ). Big data is very much about profiling-enabled personalization, as one readily understands from recent works on big data and advertising.14 These issues and the arising legal ramifications in the areas of consumer protection and data protection have been discussed by the present author elsewhere, specifically in reference to notification and recommendat...