In- Trends and Innovations in Marketing Information Systems, Tsiakis, Theodosios, (Publisher- IGI Global, 2015)

Edward Forrest

University of Alaska-Anchorage, USA

Bogdan Hoanca 

University of Alaska-Anchorage, USA


Among the emerging innovations in marketing information systems, there are arguably none greater than those applications and technologies utilizing Artificial Intelligence (AI). The advent of AI agents has placed new tools into the hands of marketers and consumers alike, and subsequently has begun the redefinition of the roles and rules of the marketing game. AI agents have begun to proffer the marketer unprecedented marketing research and communication capabilities. However, at the same time AI applications are empowering the consumer to bypass or question the corporate marketing message. Virtual Personal Shopping Assistants (VPSAs) can learn consumers’ tastes, predict their needs, and optimize their product purchases. VPSAs are able to instantly match a consumer’s need against all accessible products that meet the consumer’s expectations and price points. It is becoming increasing important for marketers to recognize the primary role that the consumer’s VPSA will play in the marketing game and adjust their marketing efforts accordingly.

Keywords: Artificial-Intelligence Marketing Agent, Virtual Personal Shopping Assistant, IBM’s Watson, Watson Engagement Advisor, Siri, Google Now, Google Glass, Microsoft Cortana, Marketing Recommendation Systems



The idea of being able to call upon an all-knowing and all-powerful entity to answer your most important question or grant your greatest wish can be traced back through the centuries from the tales of Aladdin to the Oracles of ancient Greece. However, not until the emergence of the computer and its convergence with the internet, would the dream of instant answers to every personal query and concern move beyond wishful thinking. With the advent of the computer in the 1940’s, visionaries such as Vannevar Bush foresaw the utility of amassing, collating and interlinking all of the information that was important and relevant to one’s research and interests. In his 1945 treatise, "As We May Think," Bush conceived the development of a memory index or extender which he named the “Memex” – in which one could access, retrieve, annotate, interconnect and share the world’s record base. In 1958, John McCarthy expanded on Bush’s idea of ready and hyperlinked access to a global database with his suggestion that software should be written and designed logically and follow the lines of deductive reasoning. In his 1958 paper “Programs with Common Sense,” McCarthy argued that a computer program should be able to “draw immediate conclusions from a list of premises,” the conclusions should be “either declarative or imperative sentences” and “when an imperative sentence is deduced the program takes a corresponding action.” McCarthy’s paper is regarded as the first to propose that software design, based upon “common sense reasoning ability,” was the key to artificial intelligence. However, it would take four more decades before such AI applications would be realized. It was the mid 1990’s, with the commercialization and diffusion of the Internet, exponential growth in computing power and advanced software sophistication that fully functional intelligent agents would come to pass. And, it was at this time that the nature and function of the “intelligent agent” was formally defined:

An intelligent agent is software that assists people and acts on their behalf. Intelligent agents work by allowing people to delegate work that they could have done to the agent software. Agents can, just as assistants can, automate repetitive tasks, remember things you forgot, intelligently summarize complex data, learn from you, and even make recommendations to you. (Gilbert, 1997)

Among the companies that pioneered the development and marketing application of intelligent agents were Firefly, Amazon, Apple, Yahoo and Blockbuster. Results from the first generation of intelligent agents were mixed. Blockbuster’s use of in-store kiosks that used a filtering algorithm to make film recommendations to members based on their rental history “made such interesting recommendations as a pornographic film to kids and Teletubbies to grandparents living in the same household... Magnet the first intelligent agent for the Macintosh ....silently threw everything it found in the trash when a user mistyped the name of the destination folder”    (Ansari, Essegaier, & Kohli, 2000). Since these initial trials, intelligent agents have come to play an increasingly important role in marketing strategy and customer relationship management. Today, it is hard to imagine anyone who has purchased online who has not been presented with recommendations from a marketer. The ability to understand and predict subscriber’s tastes drives Netflix’s user recommendations system, which in turn translates into customer satisfaction and subscription maintenance. In 2009 Netflix awarded its now famous Netflix Prize of $1,000,000, for the best collaborative filtering algorithm that could, based only on a user’s previous ratings and the ratings by the general user community, predict the user’s ratings for other films the user had not viewed.

 It has become standard practice to analyze a consumer’s past and present consumption patterns and to make suggestions concerning current and future sales as well as cross- and up-selling. Advanced recommender systems are also taking into consideration more of the consumer’s unique context, in terms of current location, current status and immediate needs. Yet, as sophisticated as the current recommender systems may be, the next generation of AI agents represent a paradigm shift in AI Marketing applications.  Rooted in cognitive computing, advanced sensor technology and natural language interfaces, the next iteration of the AI Marketing agent is just beginning to emerge, and it fulfills the promise of that all-knowing and all-powerful entity that can address the consumer's ultimate command --"Tell me which brand is best to buy and why!"  

This chapter’s evaluation of the impact of the next iteration of AI technologies on Marketing will be multifaceted in scope.  First, in order to clearly explain and fully prove the proposition that "Artificial Intelligence is Marketing’s game changer," three factors will be addressed: 1) What is Artificial Intelligence?  2) What is Marketing—or more specifically-- what is the Marketing Game? and, 3) What is changing? Herein, the ramifications of the emerging AI technologies will be examined from the marketing practitioner’s, as well as the consumer’s perspective. Most notably will be a discussion of the impact of Virtual Personal Shopping Assistants (VPSAs) and their role within the ongoing fundamental shift in marketing’s orientation from a marketer-driven and dominated arena to a consumer-centered and controlled marketing information system. Finally, the chapter will conclude with delineation of the research implications and social concerns of advanced AI agents’ impact on marketing and commerce. 

 Artificial Intelligence: Marketing’s Game changer

What Is Artificial Intelligence?

First coined at a 1956 Dartmouth conference, the term AI has evolved over a half century of steady progress in building computers that can manipulate symbols (both logical and linguistic). According to the Google N-gram Project, the use of the term “artificial intelligence” peaked in 1989, even though AI systems would not enter mainstream media until the late 1990’s. Before the 1990’s, advancements in AI resulted mainly in academic papers and a small number of proprietary back office expert systems designed to automate key business decisions in logistics, procurement, operations management.

A number of developments brought AI into wider use in the 2000’s.  First, search engines evolved to include not just keyword searches, but more of the context (recent searches, longer term history of search preferences, social searching – aggregating searches geographically or by moment in time, searching for synonyms). Additionally, much work has gone into refining search results to make them more relevant for users, best exemplified by many of the Google proprietary algorithms. Search engines form part of the back end, the engine inside the AI system.

Another relatively recent development, natural language processing (NLP) systems are the front end to AI systems, and they allow users to interact with a computer in conversational language, not requiring a rigid syntax as do computer programming languages. NLP systems have matured beyond textual input, to the point where they can carry out reliable voice recognition, and they can parse and process natural language queries (for example in a search engine). Processing natural language takes considerable computing power, but continuous advances in computing technologies via Moore’s Law have finally made sufficient computing power available in a cost effective package. Dedicated systems like IBM Watson rely on local hardware that occupies considerable physical space, but cloud technologies are increasingly used to move the computing power behind the scenes, allowing users to access it via handheld mobile devices. As an example, Apple’s virtual personal assistant, Siri, relies on a voice or text interface on a thin client (a smartphone or tablet) that captures the raw user input and relays it to servers in the cloud for processing. The result of the processing is returned to the user in textual or sound format, giving the appearance that Siri resides in the mobile device, but without the hardware bulk apparent in the Watson system. Moore’s Law allows the size of AI systems to shrink exponentially over time, and IBM’s Watson has gone in two years from taking an entire master bedroom to the size of a pizza box   (Upbin, IBM's Watson Now A Customer Service Agent, Coming To Smartphones Soon, 2013).

Moreover, intelligent agents are developing the sensors that early AI versions lacked. Using microphones, cameras, accelerometers and GPS in smart phones, AI systems are aware of the context around the user. They can detect emotions via voice analysis or facial expressions, and can detect whether the user is moving, stationary, or in a vehicle. They can tell whether the user is on an airport or in a coffee shop   (Dwoskin, 2014). As GPS positioning accuracy continues to increase to possibly centimeter range    (Boyd, 2014), AI systems will be able to detect exactly where the user is located in a particular store or room in a building, and will have increasingly accurate information about the context. This will allow marketers to pinpoint the customer’s location down to the shelf or the individual product level. Finally, AI systems can tap into browsing histories, into contacts list and into data from other applications to create a data-rich profile of the user’s preferences, current context and to be able to predict the most relevant response.  

In addition, other advancements have made AI systems robust enough for general use. The development of AI systems had been plagued for many years by the so called Moravec’s Paradox (Brynjolfsson & McAfee, 2014): high level reasoning was easy to automate, while basic sensorimotor skills (recognizing a face in an image, being able to pick up an egg from a basket without breaking it and the other eggs) were virtually impossible for a computer to do. A major advance of modern AI systems over the expert systems of the 1990’s is the ability of intelligent agents to learn in real time. The expert systems developed in the 1990’s were based on a set of predefined rules that were hardwired into the system and that had to be updated explicitly to avoid obsolescence. Instead, current AI systems learn from both structured and unstructured data, and are able to adapt and relearn as new data points are added. While expert systems tended to become obsolete as the environment changed, modern AI systems remain current as long as they have access to current data sets. Thanks to this ability to learn from data, it is easy for an AI system such as IBM’s Watson --which can process 500 gigabytes of information (one million books) per second   (Kawamoto, 2011)-- to branch out into a new field or to automatically stay current on the most recent knowledge.

To date, AI’s greatest public recognition has been registered by IBM’s Deep Blue victory over the world chess champion Gary Kasparov in 1997, and IBM’s Watson victory over Jeopardy’s all-time grand champions in 2011. It is now an establish fact that AI applications can meet and exceed human capabilities in many aspects where previously humans excelled. It is predicted that within the next decade the world will witness an AI “machine” pass the test laid down in 1950 by Alan Turing. That test deems an AI agent as intelligent when a human engaged in a “blind” conversation simultaneously with another human and the AI agent will not be able to discern which one is the AI agent.  Accordingly, AI applications continue to emerge as a logical, economic and demonstrably superior alternative to its human workforce counterpart. Indeed, the on-going obsolescence of the human worker by computers and AI applications is advancing at such a pace that a recent Oxford University study   (Frey & Osborne, 2013) found that out of 702 occupations evaluated, nearly half (47%) are at high risk of being fully or substantially computerized in the near future. Those first in line include most workers in transportation, logistics, office and administrative support and production-line-labor. The occupation with the absolute highest risk (at 99% probability of replacement) are “telemarketers”-- whose jobs “will become fully automated” as artificial intelligence, machine learning and interactive voice response systems are integrated and implemented. Ergo the question arises: What other impacts will AI have on Marketing and those that practice the profession?

What is Marketing and more specifically the "Marketing Game?"

Whether one examines the official American Marketing Association definition of Marketing (as “the activity, set of institutions, and processes for creating, communicating, delivering, and exchanging offerings that have value for customers, clients, partners, and society at large”) or any of the 71 other definitions of Marketing that have been proffered  (Cohen, 2011), it is clear that Marketing – in concept and practice—has been defined primarily from the seller’s perspective. However, if one considers that the defining act of the marketing process is the phenomenon of exchange between a buyer and a seller, then Marketing is best defined as a game, wherein the buyer (consumer) and seller (marketer) attempt to maximize their cost/benefit ratio in the exchange of money for goods/services.  Herein, the role and perspective of the consumer is equally as important as that of the marketer. The fact that the concept of Marketing has been relatively one-sided is attributable to the perceived passivity of the consumer in the marketing game. Marketing was something done to the consumer. Products were pushed. Communication was one way and intrusive.  Jay Walker-Smith, president of the marketing firm Yankleovich estimated that the average consumer is bombarded with 5000 ads per day    (Johnson, 2006), a number that makes it unlikely that most of the ad messages can stick in the mind of the recipient. Today, consumers are redefining the rules and their role in the game. Marketers no longer can simply buy consumers’ interest, it must be earned. Howard Rheingold, a visiting professor at Stanford University and a worldwide authority on the social impact of the communication media, sees human attention as the scarce resource in a world where information will be increasingly commoditized  (Rheingold, 2012). Communication should be interactive. Consumers increasingly rely on each other for product information and recommendations. Rather than the customer being targeted by the seller, it is the seller that is often sought out by the buyer.

It is already clear that the Internet and social media have radically altered not only the roles and relationships, but also the differential advantages held by marketer and consumers in the marketing game, as consumers can now find (via Google) and buy (via Amazon) anything, anytime, anywhere on the recommendation of friends and family (via FaceBook) or other customers (via Yelp) and tell the world about it in 140 characters or less (via Twitter). Artificial intelligence has already enabled marketers to understand qualitative input from surveys, in addition to the quantitative survey results. Using tools from companies like Luminoso Technologies and Kanjoya, Inc, marketers can automatically analyze text input on surveys and social media to understand meaning as well as the emotions of the subjects   (Rusli, 2014). What is less clear, but arguably more significant, is the degree to which the next disruptive technology – Cognitive Computing -- will be the ultimate game changer.  As the next and ultimate iteration of computing, Cognitive Computing is ushering in “the smart machine era” that (short of the singularity envisioned by Ray Kurzweil) “will be the most disruptive in the history of IT,” as it “fulfills the earliest visions for what information technologies might accomplish — doing what we thought only people could do and machines could not”   (Austin, 2013). It will “ process sensory as well as transactional input, draw inferences from past experience, understand uncertainty, draw initial conclusions, interact with people in a natural, human-language like way and then modify conclusions according to feedback”… It will “solve problems like a human, have… advanced artificial intelligence algorithms and applications that mimic the brain’s abilities for perception, action and cognition”….It will “interact with Humans… using advanced technologies like voice, gesture, and touch” and, it will “interact with other computers: leveraging the Internet of Things …and respond to networks of other smart machines and devices that talk to each other, make decisions, and get work done autonomously”    (Chamberlin, n.d.).

What is changing on the Marketers side: Artificial Intelligence Marketing Agents (AIMAs)

 Marketers are just now gaining access to an array of advanced AI platforms, services and technologies. The role and scope of marketing applications that AI is usurping ranges from the relatively simple adjustment of the style and delivery of promotional messaging to the complete coverage of customer relationship management services and interactions. Already Apple’s Siri is a familiar companion to users of the iPhone. Google’s AI stables include Google Now, a predictive assistant available on both mobile and desktop platforms. Microsoft’s Cortana is the intelligent agent for Windows Phone 8.1 (unveiled in April 2014). The most recent addition to the AI family, Cortana inherits the friendliness and humor of Apple’s Siri and the ability to anticipate the user’s needs from the Google Now predictive assistant    (Marshall, 2014). Cortana will learn the user’s preferences and will learn to anticipate the user’s needs, based on previous interactions, on geo-location information, and on current actions (for example, having a boarding pass ready as the user approaches the airport, when Cortana is aware of a previously made flight reservation). The search capabilities are based on Microsoft’s Bing search platform. Unlike most other AI systems, Cortana will have a Notebook, where user preferences, reminders and contacts will be stored, and where the user will be able to review, add and delete information as an explicit privacy interface.

To date, rudimentary applications of artificial intelligence have been used to adjust images to make photos of human faces more memorable, more attractive or friendlier    (Khosla, Bainbridge, Torralba, & Oliva, 2013),  as well as to understand what makes images popular online   (Khosla, Sarma, & Hamid, 2014). Accordingly, by monitoring social media responses, AI agents can modify campaign materials in real time, to maximize consumer interest and message retention. In addition, AI has been applied to the real time analysis of human emotions. Advanced image processing has matured to the point where emotion recognition can be done either via a webcam or using a camera in a physical setting (and emotion detection in verbal interactions has an even longer history)    (Eaton, 2012). Using AI techniques, the user’s reaction to a product or to an ad can be interpreted, and the stimulus can be modified accordingly. For pleasant or attractive stimuli, the viewer can be cajoled into completing a transaction, while for negative reactions the viewer can be offered more explanations (if the emotion detected is confusion or a blank stare), a different set of options (if positive and negative emotions are displayed) or a different offering altogether (if strong negative emotions are encountered).

Companies have already started to mine social media to understand emotions and preferences as expressed on social media sites. Bluefin Labs, acquired by Twitter in 2013, was one of the companies that started to mine Twitter postings to help companies understand how their customers perceive various product offerings. CBS was surprised to find out that lower rated shows sometimes generate more comments on Twitter, partly because the audience might be more inclined to use social media (or even because the type of show is more conducive to starting a discussion: competition shows tend to foster a more active audience)   (Talbot, 2011). Bluefin also helped Procter & Gamble understand that the context is important for the way the message is perceived. The company aired the same product ad in two shows with similar viewer demographics and with similar ratings. The surprise outcome was that one show generated eight times more comments on social media than the other    (Talbot, 2011). Bluefin Labs used AI to correlate the Twitter comments and public Facebook comments with the live feed of several TV channels, to understand which channel the posts refer to, as well as the type of content being shown. In 2011, Bluefin was able to process 1400 hours of TV programming per day, and more than 166.7 million comments    (Arndt, 2011). By June 2012, the number of social media comments exceeded 100 million per month   (Jannarone, 2012).The AI is able to distinguish between regular broadcasting (sports, reality TV etc.) and commercial breaks. The software attempts to understand emotions, and is able to handle the slang, shorthand text, as well as emoticons that are frequently used in social media   (Talbot, 2011). On the other hand, the social media comments are far from representative of the entire TV audience. For example, in July 2012, Bluefin encountered comments from 8 million people on social media, out of the 115 million household TV sets then in the US    (Jannarone, 2012).

Extrapolating on the marketing implications of emotion and facial recognition through webcam technologies, Burns    (2013)asked a retail technology expert and expert in facial recognition technology to identify key applications for Google Glass. In this initial exercise to discern ideas that “would win you customers, intelligence, or other competitive benefits,” an impressive range of applications were envisioned. To wit:  If you are interested in marketing research, “forget focus groups,” you will be able to share or rent someone’s feed and spend an hour or day with the wearer, see what they see, how they live, how they work, shop or recreate.   “With Google glass… you can sell remotely… log into a sales-person’s Glass feed and view the object for sale – be it a house, car, product demo, campus tour…or any sale that requires a “walk around;” Tourist Feeds: take virtual vacations or participate in events through purchasing Google Glass feeds... sight-see, hike, attend fairs, amusement parks, concerts or any entertainment venue... Google Glass will enable products to interact (Talking Products) with shoppers by processing NFC [near field communication] tags, QR [quick response] codes or RFID [radio frequency identifier tags] to provide videos, coupons, promotions and suggestions …all by just picking up the product and viewing it… a wine bottle can recommend a food pairing, a shirt can suggest a tie, a food item can give a recipe and cooking lesson”   (Burns, 2013).

Pricing is another traditional marketing endeavor where AI has made inroads, and where more changes are afoot. In online settings, prices can be adjusted on the fly, and Amazon has come under scrutiny when it attempted dynamic pricing, trying to squeeze a higher margin out of some of its customers    (Gunderson, 2012). As some customers noticed, the prices for certain items would go up once the items were placed in the online shopping cart – and it is debatable whether this behavior depended on the item in cause, the identity of the shopper, or the time of the transaction (possibly related to a computer glitch or the practices of a third party selling the goods on Amazon). Amazon changes prices on a regular basis (in part to evaluate the price elasticity of the goods it sells), but now claims that it does so at the same time for all customers. Given the information the company has about particular shoppers (via their account history as well as the browser cookies and history), the company could use AI to adjust prices to possibly extract a higher profit margin out of the less price sensitive shoppers. While potentially lucrative, a practice like this is almost certain to lead to considerable backlash from customers. At the same time, more equitable price changes across the board could be done even in physical stores    (Nusca, 2013), for example using the electronic price labels used in the Kohl’s stores   (Stross, 2013). These electronic labels allow much more accuracy in posting prices, but they also allow a company to put in effect a sale at a moment’s notice. In particular, a product that is not selling well on a particular day, or an entire product category in a supermarket could be set on sale by an AI agent that monitors volumes of sales and introduces price promotions in real time. 

Perhaps the most advanced demonstration of AI technology can be found in its application to customer relationship management services and interactions. Two years after demonstrating its ability to out-perform the smartest humans ever to play the game of Jeopardy, IBM has proffered its Watson Cognitive Computing application to marketers in the guise of the “Engagement Advisor.” Envisioned as an intelligent technology whose services can apply to almost any industry, but especially those that receive many customer service inquiries, such as retailing, banking, insurance and telecommunications, the Engagement Advisor can supplement or supplant human customer-service representatives (Lohr, 2013). As described by Manoj Saxena, General Manager, IBM Watson Solutions:

Watson Engagement Advisor… available in hosted, on-premises, or cloud-based deployment models, learns from its interactions with consumers…”Watson can take questions in natural language, by typing it in or speaking, which makes for more natural conversations…. “Watson is also able to give you not just answers, but also the evidence and sources they're based on… Watson can provide omnichannel interactions; conversations begun on a smartphone, for example, can be picked up at the same point on a tablet later.    (Schiff, 2013)


Now considerably smaller in size and faster than when it won Jeopardy in 2011, Watson’s super Q&A capabilities are a perfect solution to the marketer’s primary objective of creating, sustaining and growing one’s relationship with their customers. Customers expect and demand immediate, knowledgeable and personal attention to their inquiries.  A single point change in customer satisfaction can yield significant gains “in share of wallet, profitability, revenue, and relationship growth”    (IBM Watson Engagement Advisor, Transforming the way people and organizations interact over the lifetime of their relationships, 2013). Engaged customers will spend up to three times more with a brand, and are two times more likely to recommend a company to others. Present levels of customer service are woefully inadequate, as half of the 270 billion customer service calls placed each year go unresolved  (Upbin, 2013). Given its ability to scan, analyze and interpret over 200 million pages of structured and unstructured data in a matter of seconds and subsequently enter into a natural language conversation based on its interpretation of and conclusions drawn from its analysis-- the IBM Watson Engagement Advisor provides consumers with a depth of  personalized and intelligent assistance that far exceeds what any human customer service agent could ever deliver.

Moreover, Watson’s capability to interact, answer questions, and assist consumers with complex purchasing decisions and/or solve problems a consumer might be having with products will revolutionize marketing on other fronts as well. Its ability to analyze structured and unstructured data from credit cards, sales databases, social networks, location data, web pages will enable marketers to ascertain trends in consumer preferences and sentiments, calculate sales probabilities and “proactively and intelligently test, measure and optimize digital content, ads, website pages”    (Fidelman, 2013). Applied to the task of media measurement, Watson is being utilized “to sift through and make sense of all the data advertisers receive about their Web, social media, print and television marketing campaigns... to optimize campaigns in real time across every marketing channel”    (Lohr, 2013).

Finally, in addition to being able to predict consumer trends and analyze consumer media consumption patterns, Watson has the ability to discern consumers’ personalities through Psycholinguistic Analysis of the language consumers use when posting on social media sites and predict major events that are likely to happen in a consumer’s life. “Amazingly, this tech doesn't even have to know a company's customer's social media accounts beforehand. It can figure them out on its own by sifting through what people post online with information in a company database“   (Russel, 2014).

What is changing on the Consumers side: Virtual Personal Shopping Assistants (VPSAs)

From the consumers’ perspective, the emergence of a virtual personal shopping assistant which can learn, predict and serve their tastes, needs and desires and optimize their product/service purchases is a welcome development. This virtual personal shopping concierge will be able to instantly match an immediate or imminent need against all accessible products that meet a consumer’s expectations and price points. The savings in time will be significant, given that consumers are hard pressed to keep up with all the latest trends, specials and sales. Moreover, when they do enter shopping mode “50 percent of consumers spend 75 percent or more of their total shopping time conducting online research… as the current state of online and mobile shopping yields a time-consuming, inefficient experience, as consumers comb through websites for information, link by link, or enter keywords into search engines and hope for the best”    (Natale, 2014).  With one’s own VPSA the global marketplace is monitored 24/7/365 and whether one is shopping or not, the consumer would be notified when any item of relevance is available. Once a potential product/service purchase is identified, the VPSA allows the consumer to “ask specific questions based on explicit needs and get expert, personalized, information-driven responses to guide buying decisions,” thus creating “the same experience we have in real-world stores with great sales reps every day and is what's missing from digital retail”    (Natale, 2014).  An agent like IBM’s Watson will be able to help a consumer through a complex purchase, for example a real estate transaction. Watson will be able to provide advice on the real estate market, including considerations about valuation, on the financing options available, on the architectural and regulatory restrictions on possible changes and improvements to the existing property, on the tax implications of the purchase, all with an eye to the consumer’s financial situation, investment plans, and family situation.

Access to a personal, omniscient, omnipresent shopping assistant which can engage in real-time conversation to learn the consumer’s needs, answer the consumer’s questions and guide the purchase decisions is currently being developed in the form of Fluid XPS. A partnership between of IBM, North Face and Fluid Inc. building on IBM Watson's natural language interface and cognitive computing analytical abilities, “Fluid XPS will draw on data, (including the brand's product information, user reviews and online expert publications through IBM Watson) to provide consumers with informed recommendations according to their needs and desires. Consumers will receive valuable insights in making smart, satisfying purchases and be able to interact with Watson on desktops, tablets and smartphones for the first time”  (Natale, 2014). In January 2014,” Fluid’s CEO presented a real life example comparing Fluid XPS with today’s online shopping experience:

  ... trying to help his son purchase some equipment on line, they read and read product descriptions, they comparison shopped to find a great price, and, after hours of exploration, they walked away from their computers without buying a thing. Instead they jumped in the car and went to REI. Why? Because they had questions and the only experts they had to consult with were themselves. Yelp isn’t interactive.    (Backaitis, 2014)

However, with Fluid XPS the experience for an individual “who plans to go camping in Patagonia and logs on to The North Face’s website to gear-up  ... wouldn’t be that different from what happens in a physical store:”

  ...the concierge might ask the shopper a little bit about her/himself and what he/she might [be] looking for. Watson would then show him/her some things he/she might enjoy or need. In the case of someone who is about to go winter camping, Watson could suggest a sleeping bag for frigid weather. The shopper would be able to tell Watson things like “I think I’d be too warm in that” and then a new suggestion would be made. It sure beats shopping on most of today’s retail sites that aren’t much more than on-line catalogs from which you can electronically order.    (Backaitis, 2014)

Of course, one of the concerns about a technology like Fluid XPS will be the level of trust the consumer needs to develop in the advice. Just as with a salesperson, an AI agent that seems pushy (up-selling or cross-selling) will not get a good response from the consumer. But with a person, the customer can tell if the salesperson is intending to manipulate, even with the most agreeable demeanor. In contrast, an AI agent will have no non-verbal clues that can engender trust, and will have more difficulty in building rapport and trust with the customer.


The Marketing Game: new tools, roles & rules

The advent of AI agents has placed new tools into the hands of marketers and consumers alike, and subsequently has begun the redefinition of the roles and rules of the game. On the marketer’s side, the standard approach of segmenting entire populations and assigning purchase probabilities and media exposure rates for such unrefined targets as “women 18-49 years old” are clearly antiquated. As Rugfelt  (2014) observed, “guessing what individual customers want, via researching large demographic populations, is no longer a viable marketing strategy...Amazon, Facebook, and Google are investing money in machine learning and artificial intelligence to predict what each of their customers want—even before they know... with the consumer more connected on more devices at all times of the day, there's more data available to collect and analyze.”

Simultaneously, AI agents increase the knowledge and insight marketers have regarding consumers’ product preferences and buying patterns, yet, render consumers less susceptible to the marketer’s message and manipulation. As virtual personal shopping assistants search, evaluate, recommend and finalize purchases, the actual consumer is removed from the purchasing process. As marketers and consumers begin to fully avail themselves of the advantages that artificial-intelligence agents proffer, the nature and process of the facilitating exchange in marketplace will be dramatically altered. On the consumer’s side, having a virtual shopping assistant that is fully informed of all available options, prices and evaluations greatly lessens the need or influence of the marketer’s commercials or promo-tools. As the marketer is increasingly disintermediated, it will become increasing important for marketers to recognize the primary role that the consumer’s VPSA plays in the marketing game and adjust their marketing efforts accordingly. Continuing to invoke the old moves in the new game will not only prove ineffective but counterproductive in gaining consumers’ trust and patronage. Again, as Rugfelt  (2014) observes, “today's marketers have more opportunities to analyze content and data to deliver campaigns that would benefit the marketer, but more important, the consumer.” Furthermore,

 [b]ecause users will want their assistants to help them optimize their lives, they'll be willing to share lots of information with it. This way, marketers will be receiving a current, steady stream of relevant information while the assistant collects and learns how consumers behave, what they're doing, and what they want. Once this assistant has proven its relevance, users will trust it as an authority on the possible opportunities for themselves.


Smarter marketing strategies ... should make room for more hypertargeted services offered to the consumer. Creating well-informed, relevant, and timely options that don't interfere with private digital spaces and respect the consumer's time and consideration will be a demand that will only increase.... We're already seeing marketing-driven apps powered and managed by individual consumers who self-serve by doing all the work a salesperson would help them with in-store. New AI-powered virtual assistants will exist to make consumers' individual lives more organized and improved—to take advantage of every discount, say, or to alert them when they're hitting their monthly spending limit. By using a Virtual Assistant app for advertising, the app can function as educational to consumers rather than feeling invasive, like other outdated forms of advertising, transforming the platform into a valuable service for the customer.   (Rugfelt, 2014)

We are just beginning to contemplate the ramifications of only the initial iterations of artificial intelligence agents in the marketing arena. In the nascent stages of this paradigm shift in the marketing game the impact and effects of AI agents will be more experienced than understood. Albeit, one thing already is clear. Consumers are gaining access to information processing technology and techniques that were formerly the preserve of the marketing professional and have begun to fundamentally alter the informational asymmetries which were inherent in the marketing game. In the past the marketer controlled the channels of communication and distribution, positioned the product and set the price.  The consumer was hit with over 3,000 messages a day that were placed in mass media vehicles that were targeted at a statistically similar gaggle of persons. The messages were intrusive and brief, designed as much to entertain as to inform, and provided the consumer with little more than a catch phrase in the guise of a slogan or jingle. With the advent of the internet and emergence of social media the communication pattern of the marketing game expanded from strictly “outbound” with the marketer doing all the talking, to “inbound” as well, wherein the consumer searches for, contacts, and transacts with the marketer. Social media has already fundamentally altered the pattern of behavior and communication that consumers employ as they search, evaluate and purchase products/services. Yet, however great and swift this change has been (given that social media is but a decade old), it could be said that the greatest change is still to come:

 “If you look out at the next five years, you'll see a super-convergence of major technologies coming together at one time—Big Data, the cloud, analytics, mobile, and social—creating a significant disruption in the market” (says Manoj Saxena, general manager of IBM Watson Solutions) who also noted the growing population shift toward consumers brought up amid digital technologies. “10,000 workers are retiring every day, soon to be replaced by millennials who have a whole different sense of how they want to interact [with brands].”    (Schiff, 2013)

In light of this shift in the consumers’ demographic and technographic profiles, a major re-ordering of marketing strategy is being evidenced. To assess the increased influence of social media, Simonson and Rosen  (2014) draw upon the Influence Mix model. The model measures the three components that influence a person’s decision to purchase a good or service: personal preferences, beliefs and experiences (P), influences from other people and sources (O) and the marketing message of the vendor (M). Since the three types of influence all compete for the customer’s attention, this is a zero-sum game, where an increase in influence in one area will result in lower influence in another. Thanks to social media and to the increased availability of online reviews, the purchasing decision in many industries is more and more influenced by the O component, at the expense of the M component. Simonson and Rosen view this trend as weakening the power of brands, weakening influence of even past purchases and experiences with a brand, and an overall weakening of the powers of positioning and persuasion, traditional tools of the marketers. This trend is not across the board, as few customers will check social media when deciding what brand of tissue to purchase (so called O-independent purchases), but will be more prevalent for goods that have a social aspect or that are purchased less frequently, and for larger amounts (called O-dependent purchases). Sadly for marketers, the types of goods that are O-independent are in low margin, low differentiation, almost commodity product areas.

On one hand, this loss of control over the marketing message is bad news for marketers. Consumers disregard the marketing message in favor of online opinions from other customers with similar backgrounds and interests, something the company cannot easily control. On the other hand, the increased influence of the O lends itself to valuable and freely available market research information. Instead of positioning and persuading, Simonson and Rosen view marketing as moving to a model where marketers keep an eye on competitors’ actions, view the reaction of the market, and make their own move in response. AI agents will review social media sites in real time, both about the company’s actions and about the competitors’, and will propose (or maybe even carry out) counteractions. A number of companies have positioned themselves to provide such services, for now driven less by AI and more by human brainpower: Bluefin Labs (acquired by Twitter), Radian 6 (now part of Salesforce), General Sentiment, Sysomos, Converseon, and Trendrr   (Talbot, 2011).  A company that has been successful in doing this type of quick response in the fashion industry is Zara    (Gallaugher, 2013). The company strategy is to identify promising high end fashion trends and to produce low cost replicas, using a highly vertically integrated operations system to produce imitations of the latest fashion within two weeks of their public appearance. To identify trends, Zara currently relies on human input from sales staff on the store floor who interact with customers. Using AI agents, the company would be able to gather similar information even faster, to further shorten the product lifecycle.

Another intriguing avenue for marketing messages is in automated group recommender (AGR) systems. While the O-dependent decisions discussed above involve individual consumption of products and services, an entire category of hedonic offerings involves group consumption, for example artistic performances or fine dining. The challenge for such situations is that the decisions that are optimal for an individual are not optimal for the entire group. When a group of people must decide what movie to watch together or what bottle of wine to consume together, suggestions from an AGR can lead to decisions that are better for the group than a human-derived choice. Even when members of the group make individual decisions (for example making separate food choices for a restaurant meal together), AGRs can still provide value if group members share high quality social relationships among themselves   (Hennig-Thurau, Marchand, & Marx, 2012). Such AGR tools could still be an avenue for marketers to aggregate preferences and to provide personalized advice the group could not otherwise gather on its own from online reviews.

From Mass Customization to Automated Mass Customization

Another area in which AI will change the marketing game is in product design and development. After almost a century of marketing to the mass of consumers (at best, segmented in large groups), information technology made possible a more personalized approach in the 1990’s. Pioneered originally by companies like Dell Computer, this new approach involved the customer in the product definition. Instead of conducting market research and targeting some “best guess” product configuration to an entire customer segment, mass customization provided a wide set of options and allowed the customers to custom design the product that best fit their needs and their budget. For example, rather than purchase a standard configuration of a personal computer from Dell Computer, a customer could interactively configure online and in real time the personal computer’s configuration: the CPU type and speed, the amount of RAM, the number and type of peripherals, until he or she found the best price/performance combination. More recently, Nike allowed customers to personalize athletic shoes using a vast menu of options. With the advent of AI agents, the customer could be abstracted from this process, because the AI agent will be able to carry out the transaction, personalizing the product based on its knowledge of the customer’s preferences. This new capability will allow for a much larger set of options, which would not have been practical for human interactions. Moreover, with the advent of new technologies like 3D printing, more of the design could be outsourced to the customer’s AI agent, not just a basic configuration with a set of standard options. For example, the AI agent could order a design for a coffee table that matches the style of other furniture in the customer’s possession, of size and design that are not provided by the vendor. Indeed, when one considers the confluence and integration of AI technologies with 3D printing the ramifications are nothing short of revolutionary.  When one combines the ability and convenience of producing one’s own customized goods with the savings accrued through the elimination of labor, re-tooling, assembly, shipping and inventory carrying costs, the consequences are most ominous for any and all engaged in traditional manufacturing and dependent on the relative cost-efficiencies of out-sourcing. 3D printing not only renders factories obsolete but threatens whole country’s economies as production is taken up by the consumer, distribution is de-globalized and marketing marginalized.  


Your Artificial-Intelligence Marketing Agent vs. My Virtual Personal Shopping Assistant

In the final analysis, the ultimate winner of the marketing game comes down to a battle of the “bots.” The marketer’s bot that can find, read and analyze a consumer’s every search, download, blogpost, picture, Facebook conversation, product evaluation, tweet, credit card purchase, media consumption preference and/or any and every other bit of information that can be gleaned in the infosphere about the consumer. With the ability to analyze the personality of the consumers, categorize their lifestyle and buying style, extrapolate all their information, the marketer can formulate the optimal marketing mix to sell, cross-sell and up-sell a consumer their product, as well as gain, sustain and cultivate consumers as life-long customers.

Meeting the Marketer’s Bot in the marketplace/ring is the Consumer’s Bot. In addition to knowing a consumer’s tastes, needs and desires, the VPSA knows all past purchases, current preferences and future plans. Moreover, the VPSA has a 360 degree view of the marketplace and can provide a ranking and recommendation of all competitive and comparative products. Rather than pushing a single brand, as a marketer’s bot, the VPSA more inclusively and objectively evaluates an entire evoked set of brands and proffers a recommendation aligned with and in the best interests of its master. No doubt, in the long run the VPSA should enable the consumer to gain the upper hand in the marketing game. This outcome appears inevitable, when one considers the latest advancement in Watson’s AI skills- the ability to present reasoned arguments pro and con.  In a recent demonstration (on April 2014), an example of Watson’s new found debating capacity was evidenced. After scanning through its database of 15 terabytes of information on Wikipedia, Watson was asked to present arguments for and against the sale of violent videogames to minors. Watson’s verbatim reply was:

"I would like to raise the following points in support of the topic. Exposure to violent videogames results in increased physiological arousal, aggression-related thoughts and feelings as well as decreased social behavior. In addition, these violent games or lyrics actually cause adolescents to commit acts of real-life aggression. Finally, violent videogames can increase children's aggression.

On the other hand, I would like to note the following claims that oppose the topic. Violence in videogames is not causally linked with aggressive tendencies. In addition, most children who play videogames do not have problems. Finally, videogame play is part of an adolescent boy's normal social setting."    (Borghino, 2014)

One can easily envision asking Watson for a reasoned argument for and against the purchase of any product. Watson having searched and analyzed every available customer review, user recommendation and Consumer Reports article in the world will deliver the consumer its verdict in a matter of seconds. For any purchase the consumer is considering all that is needed is a quick query of Watson: Watson what is the best brand? What is the best price? 


To date, the application of AI agents has been principally developed by and has primarily served marketing professionals to augment and extend the traditional marketing toolset. It is already fairly common practice to use data driven advertising, where click-through rate or revenue generated are used as a metric to identify which one of two ads is more effective   (Ayres, 2008). Using AI techniques, the ads themselves can be adjusted in real time to make them more memorable or possibly more effective at converting views into revenue, using techniques as outlined in   (Khosla, Bainbridge, Torralba, & Oliva, 2013). Another use of AI agents has been in harvesting and interpreting the vast amounts of customer feedback available online on social media, knowing full well that the sites can have a critical role in shaping the customers’ purchasing decisions. At outdoor equipment manufacturer LL Bean, existing marketing software flags products that have even a small number of one or two star ratings (out of five), for review by management. Products that have real flaws (as opposed to simply accidental flaws) are modified, discounted or discontinued   (Simonson & Rosen, 2014). As discussed earlier, more advanced AI agents can cull information outside of the company’s website, including social media sites, blogs, and other resources, to build a more complete picture of the perceived value of the company’s product mix. The importance of obtaining this more complete picture is premised on the fact that the Influence Mix model predicts that in O-dependent industries (influences from other people and sources) the information available online will be more important than the marketer’s commercial campaign.

This increasing realization that the influence and impact of the marketer’s commercial campaign is on the wane is further exacerbated by the fact that it may eventually be dismissed altogether when the consumer’s purchase cycle is fully maintained and conducted by a consumer’s VPSA. As VPSAs become available they will increasingly serve as the primary product purchasing platform for consumers. Accordingly, corporate marketing efforts may in the end be better directed at VPSAs, rather than at human decision makers. Corporate marketers will need to address the VPSA directly, instead, or in addition to the human target.

If consumers increase their reliance on VPSAs, this will greatly reduce the success of some of the marketing techniques that rely on human traits: impulse buying (preference for a product that is immediately available over one that is less available), cognitive heuristics and limitations (preference for a simple choice over a more complex choice), sensory overload (responding to appetizing smells or to arousing imagery) and irrational decision making (including the inability to carry out complex calculations in one’s head). In contrast, VPSAs will be closer to the homo economicus ideal, with essentially unlimited ability to uncover and compare choices, with the ability to follow the user’s preferences and not be swayed by the commercial message, and with the power to negotiate with several merchants at the same time.  

Future Research Directions

The role and scope of artificial intelligence in the marketing game is only in its nascent stage and its impact is predicted to increase exponentially. We have shown applications in marketing that will impact both corporate marketing systems (typically to improve on the traditional four P’s of marketing), as well as those on the consumer’s side.

Research on corporate marketing systems will need to focus on how to use AI to improve current marketing processes. A related, but important research area will involve the customer perceptions of these AI systems. Because AI agents will learn about consumers’ habits and will be able to infer behaviors, trends and preferences that might be rather private, corporations will need to establish privacy policies that will reassure customers. In 2008, a controversial marketing campaign at Target sent pregnancy related coupons and offers to women, based on information mined from the purchasing patterns of these women. Since the customers had not registered to receive pregnancy related materials, several (including a 16 year old girl still living with her parents) viewed the materials as an invasion of their privacy   (Hill, 2012). As more advanced AI systems are able to dig even deeper into the motives behind customers’ purchases, companies will need to at least give the impression that they did not pry too deeply: Target resorted to mixing pregnancy related offers with other seemingly random offers, leading the consumers to believe the company was not targeting them for particular purchases.

The rise of VPSAs will have a considerable impact on how customers make purchasing decisions, but the whole concept is still in its infancy. As rational decision making agents, VPSAs will not be susceptible to the same types of marketing messages currently aimed at humans. Marketers will need to understand how to position their products and how to target a mixed audience of both humans and VPSAs. Future research will also need to consider how VPSAs will be created and deployed to gain the trust of consumers. By their “personal” nature, VPSAs will have considerable insights into what drives consumers’ preferences. Will the VPSA’s data be owned by the consumer, or also accessible to the manufacturer of the agent? How will the consumer know that there are no backdoors (corporate or governmental in nature) that can leak out private information from the VPSA? How will the security of the agent be ensured, even if no backdoors are present? A third party breaking into a VPSA will gain tremendous information about the agent’s owner, private and confidential information about purchasing preferences, but also bank accounts and purchase histories that could give a thief access to valuable physical and virtual assets. Unless consumers are satisfied with the security levels available for VPSAs, the merits of the technology will not be compelling enough for it to become widely used.

Finally, a social concern is that the rise of VPSAs will lead for further social inequalities. Wealthier people, who will be more likely to afford a more expensive and capable agent, will have even more access to the best deals, the most up to date information, and the most efficient consumption processes. Despite the possible economic benefits of VPSAs, an unwise application of the technology has the potential to further deepen the digital divide, and to increase the inequalities that plague society. Further research might uncover ways in which such an outcome can be avoided.


Amazon Prime Air. (n.d.). Retrieved from Amazon:

Ansari, A., Essegaier, S., & Kohli, R. (2000). Internet Recommendation Systems. Journal of Marketing Research, 37(3), 363-375.

Apple - iOS7 - Siri. (n.d.). Retrieved from

Arndt, R. Z. (2011, November 7). Bluefin mines social media to improve TV analytics. Fast Company.

Austin, T. (2013). The Disruptive Era of Smart Machines Is Upon Us. Gartner. Retrieved from

Ayres, I. (2008). Super Crunchers: Why Thinking-By-Numbers is the New Way To Be Smart. Bantam.

Backaitis, V. (2014, January 12). Will IBM's Watson Change the Customer Experience? Beyond Customer Experience Management - What your CMS really needs to deliver. CMS Wire. Retrieved from

Borghino, D. (2014, May 9). IBM's Watson can now debate any topic. Gigmag,. Retrieved from

Boyd, J. (2014, April 23). Japan’s Plan for Centimeter-Resolution GPS. Retrieved from IEEE Spectrum:

Brynjolfsson, E., & McAfee, A. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. New York, New York: W. W. Norton & Company.

Burns, W. (2013, February 27). Nine Business Applications For Google Glass. Forbes. Retrieved from

Bush, V. (1945, July 1). As We May Think. The Atlantic. Retrieved from

Chamberlin, B. (n.d.). Cognitive Computing Trends to Watch in 2014. Retrieved from

Cohen, H. (2011, March 29). 72 Marketing Definitions. Retrieved from

Dwoskin, E. (2014, January 13). What Secrets Your Phone Is Sharing About You: Businesses Use Sensors to Track Customers, Build Shopper Profiles. Wall Street Journal.

Eaton, K. (2012, June 7). Does your phone know how happy you are? The emotion recognition industry comes giddily of age. Fast Company.

Epps, S. (2013, June 2). Five Ways Google Glass Will Change Marketing. Forbes. Retrieved from

Fidelman, M. (2013, September 4). IBM's Watson Set To Revolutionize Marketing. Forbes. Retrieved from

Frey, C. B., & Osborne, M. A. (2013, September 17). The Future of Employment: How Susceptible are Jobs to Computerisation? Retrieved from

Gallaugher, J. (2013). Information Systems: A Manager's Guide to Harnessing Technology, v. 2.0. Flat World Knowledge.

Gilbert, D. (1997). Intelligent Agents: The Right Information at the Right Time. IBM Corp.

GoodGuide Delivered To Your Phone. (n.d.). Retrieved from GoodGuide:

Gunderson, L. (2012, May 11). Amazon's 'dynamic' prices get some static. The Oregonian.

Hennig-Thurau, T., Marchand, A., & Marx, P. (2012, September). Can Automated Group Recommender Systems Help Consumers Make Better Choices? Journal of Marketing, 76, pp. 89-109.

Hill, K. (2012, February 2). How Target Figured Out A Teen Girl Was Pregnant Before Her Father Did. Forbes.

IBM Watson. (2014, January). Retrieved from IBM:

IBM Watson Engagement Advisor, Transforming the way people and organizations interact over the lifetime of their relationships. (2013, May). Retrieved from IBM:

Jannarone, J. (2012, September 17). When Twitter Fans Steer TV. Wall Street Journal.

Johnson, C. (2006, September 17). Cutting Through Advertising Clutter. Retrieved from CBS News:

Kawamoto, D. (2011, February 8). How IBM Built Watson, Its 'Jeopardy'-Playing Supercomputer. Retrieved from Daily Finance:

Khosla, A., Bainbridge, W. A., Torralba, A., & Oliva, A. (2013). Modifying the Memorability of Face Photographs. International Conference on Computer Vision 2013. Sydney.

Khosla, A., Sarma, A. D., & Hamid, R. (2014). What Makes an Image Popular? International World Wide Web Conference. Seoul, Korea.

Kotler, P., & Keller, K. (2011 ). Marketing Management, 14/E. Prentice Hall.

Lohr, S. (2013, May 21). I.B.M. Puts Watson to Work in Business. Retrieved from New York Times blogs:

Maitin-Shepard, J., Cusumano-Towner, M., Lei, J., & Abbeel, P. (2010). Cloth Grasp Point Detection based on Multiple-View Geometric Cues with Application to Robotic Towel Folding. International Conference on Robotics and Automation.

Marshall, G. (2014, April 3). Cortana: everything you need to know about Microsoft's Siri rival. Retrieved from techradar:

McCarthy, J. (1958). Programs with common sense. Symposium on Mechanization of Thought Processes (pp. 77-84). Teddington, England: National Physical Laboratory.

Natale, J. (2014, April 23). IBM Watson Group to transform the consumer shopping experience. Retrieved from

Nusca, A. (2013, October 2). The future of retail is dynamic pricing. So why can't we get it right? ZDNet.

Rheingold, H. (2012). Net smart : how to thrive online. Cambridge, Massachusetts: MIT Press.

Rugfelt, M. (2014, May 1). Artificial Intelligence's Impact on Marketing. Direct Marketing News. Retrieved from marketing/article/344554/

Rusli, E. (2014, April 2014). Firms Use Artificial Intelligence to Tap Shoppers' Views: Analyzing Focus Groups, Surveys, Online Forums and Social Media Provides New Data. The Wall Street Journal.

Russel, K. (2014, February 4). IBM Is Using Watson To Psychoanalyze People From Their Tweets. Business Insider. Retrieved from

Saxena, M. (2013, May 21). How Cognitive Technologies Can Help Transform the Way Companies and Consumers Connect. Retrieved from A Smarter Planet:

Schiff, A. (2013, May 21). Just Ask Watson: IBM Unveils the Watson Engagement Advisor. Direct Marketing News. Retrieved from

Seamans, R., & Zhu, F. (2014). Responses to Entry in Multi-Sided Markets: The Impact of Craigslist on Local Newspapers. Management Science, 60(2), 476-493.

Simonson, I., & Rosen, E. (2014). Absolute value: what really influences customers in the age of (nearly) perfect information. New York, NY: Harper Collins.

Stross, R. (2013, February 9). Digital Tags Help Ensure the Price Is Right. The New York Times.

Talbot, D. (2011, October 18). A Social-Media Decoder. MIT Technology Review.

Upbin, B. (2013, February 8). IBM's Watson Gets Its First Piece Of Business In Healthcare. Retrieved from Forbes:

Upbin, B. (2013, May 21). IBM's Watson Now A Customer Service Agent, Coming To Smartphones Soon. Forbes.

Wagner, M. (2011, January 19). What Does IBM's Watson Mean for Marketing? Retrieved from The CMO Site: