“AI is one of the most important things humanity is working on. It is more profound than, I dunno, electricity or fire.”
Sundar Pichai, CEO, Google
We seem to be living in the age of AI. Everywhere you look, companies are touting their most recent AI, machine learning (ML), and deep learning breakthroughs, even when they are far short of anything that could be dubbed a breakthrough. “AI” has probably superseded “Blockchain”, “Crypto”, and/or “Fintech” as the buzzword of the day. Indeed, one of the best ways to raise VC funding is to stick ‘AI’ or ‘ML’ at the front of your company’s prospectus and a “.ai” at the end of your company’s website.
Separating AI fact from fiction is not easy, partly because “AI” might be the buzzword of all buzzwords. With its end-of-world overtones, it could, literally, result in the end of humanity, not just as we know it, but as it finitely is. With their wishful thinking undertones, some argue that it has the ability to save the world. Just about every software, semiconductor, tech, advertising, retail, marketing, analytics, CRM, and robotics company touts its cutting-edge AI know-how. The hype cycle has hit overdrive and, in times like this, caution must be taken, as many promises are being made in the name of encouraging sales rather than in honest appraisals of AI’s true capabilities.
Today, AI is being sold by companies of all stripes and sizes as a panacea for just about any business problem imaginable; want to automate away an expensive labor force – add some Robotic Process Automation (RPA); want to cut down on customer service calls – build a chatbot to handle repetitive and labor intensive client questions; want to increase personalization – build a deep learning model that captures a holistic view of a customer so that a business can understand him or her so intimately that it will be able to predict upcoming purchases, as well as any potential problems that might go with those purchases; want to catch credit card fraud in action – build real-time models that spot outliers fast enough to refuse questionable purchases; want to save the world, well AI can do it in 11 different ways, that is, at least, according to the Global Challenges Foundation, which sees AI as both mankind’s biggest potential global threat risk – 10% chance of extinguishing the human race – as well as the potential technological savior of problems like the governance of nuclear weapons, ecological collapse, catastrophic climate change, asteroid impacts, among others.
Major tech companies have embraced AI as if it was one of the most important discoveries ever invented; Google’s CEO compares AI to the discovery of fire and electricity, while leading what he calls an “AI-first” company; Amazon’s entire business is shaped by AI, from its customer personalization and loyalty programs, to its warehousing, robotics and logistics capabilities, even to Alexa, its voice-activated assistant; IBM has Watson and Big Blue is pushing what it calls “cognitive computing”; Facebook has AI and ML algorithms that test out which of its AI and ML algorithms are most effective and should be rolled-out company-wide; Adobe, a big player in the multi-channel marketing and personalization space, runs much of its Experience Cloud Marketing platform through its Sensei AI algorithms; even the analytics powerhouse SAS has recently announced that it will spend US $1B over the next three years on AI initiatives. With Burger King creating commercials that claim to be produced with AI but really aren’t, one could say the situation has actually reached the poking-fun-at-itself stage, which often means the end of a technological and/or entertainment cycle. However, I don’t think we’re at the end of the AI technology cycle, quite the opposite, we’re probably just at the end of the beginning.
So why should brands that aren’t software companies choose to go down the tricky and complex AI road? Well, in the article Artificial intelligence Unlocks the True Power of Analytics, Adobe explains the vast difference between doing things in a rules-based analytics way and an AI-powered way, including:
- Provide warnings whenever a company activity falls outside the norm. The difference7:
- Rules-based analytics: You set a threshold for activity (e.g., “200–275 orders per hour”) and then manually investigate whether each alert is important.
- AI-powered analytics: The AI analytics tool automatically determines that the event is worthy of an alert, then fires it off unaided.
- Conduct a root cause analysis and recommend action. The difference7:
- Rules-based analytics: You manually investigate why an event may have happened and consider possible actions.
- AI-powered analytics: Your tool automatically evaluates what factors contributed to the event and suggests a cause and an action.
- Evaluate campaign effectiveness7:
- Rules-based analytics: The business manually sets rules and weights to attribute the value of each touch that led to a conversion.
- AI-powered analytics: The AI analytics tool automatically weights and reports the factors that led to each successful outcome and attributes credit to each campaign element or step accordingly.
- Identify customers who are at risk of defecting7:
- Rules-based analytics: You manually study reports on groups of customers that have defected and try to see patterns.
- AI-powered analytics: Your tool automatically Identifies which segments are at greatest risk of defection.
- Select segments that will be the most responsive to upcoming campaigns7:
- Rules-based analytics: You manually consider and hypothesize about the attributes of customers that might prove to be predictive of their response.
- AI-powered analytics: Your tool automatically creates segments based on attributes that currently drive the desired response.
- Find your best customers7:
- Rules-based analytics: You manually analyze segments in order to understand what makes high-quality customers different.
- AI-powered analytics: Your tool automatically identifies statistically significant attributes that high-performing customers have in common and creates customer segments for you to take action on.
Beyond the impressive differences listed above, AI can also be used in website morphing, customer and media recommendations, programmatic advertising, purchase prediction, demand forecasting, social listening, and much, much more, as I later explain. In its Artificial Intelligence in Logistics, DHL Customer Solutions & Innovation explains how AI was used to pick up the fidget spinners trend: “The first videos of teenagers doing tricks with fidget spinners began trending on YouTube in February 2017. Hidden deep within online browsing data, YouTube video views, and conversations on social media, AI in its current state is able to identify both the quantitative rise in interest in a topic; as well as the context of that interest from semantic understanding of unstructured text. This enables predictions to be made about which fads could boom in a similar fashion to fidget spinners.”
In the AI Momentum, Maturity, & Models for Success, a consortium of companies consisting of Accenture, Intel, Forbes Insight, and SAS polled 300 executives from a wide variety of industries about AI and discovered that it appears likely that we are on the verge of a radical momentum shift for the technology. The responses submitted in the “What are the benefits of AI study” (see Figure 1) show “a level of enthusiasm and AI-focused activity that point to an explosion of AI adoption just around the corner, even as gaps in capabilities and strategy are revealed. Already, 72 percent of the organizations we surveyed have either deployed AI-based technology or are in the process of doing so.”9
“A large percentage of survey respondents report having real success with AI. When looking at only those who have reported having deployed AI, 51 percent say the impact of deployment of AI-based technologies on their operations has been ‘successful’ or ‘highly successful,’” claim the consortium.9
Figure 1: Benefits of AI
In early 2018, while sitting at the China Gardens restaurant at HKUST, I was giving an update of my recent AI talks to a friend, who is a professor there. After listing off a few of the exotic locales I had been invited to speak in, my friend chimed in with the rather dismissive statement, “You know AI is just a neural net, right?” I answered that I did, but his question stuck with me. The question wasn’t meant to be dismissive or mean-spirited, it was simply the question of an annoyed expert who had been training neural nets for over a decade and was bothered, as well as probably a little surprised, by all the attention AI was suddenly receiving. My friend’s question made me realize how the overhype machine had hit overdrive with AI, machine learning, and deep learning and that anyone delving into this technology should be ultra-cautious.
What exactly is AI?
The Artificial Intelligence and Life in 2030 study states that, “Artificial Intelligence (AI) is a science and a set of computational technologies that are inspired by—but typically operate quite differently from—the ways people use their nervous systems and bodies to sense, learn, reason, and take action.” The study adds that, “While the rate of progress in AI has been patchy and unpredictable, there have been significant advances since the field’s inception sixty years ago.”10
In its report Sizing the prize. What’s the real value of AI for your business and how you can capitalise, PWC believes “AI could contribute up to $15.7 trillion to the global economy in 2030, more than the current output of China and India combined. Of this, $6.6 trillion is likely to come from increased productivity and $9.1 trillion is likely to come from consumption-side effects.” According to PWC’s analysis, “global GDP will be up to 14% higher in 2030 as a result of the accelerating development and take-up of AI–the equivalent of an additional $15.7 trillion.”11 For PWC, the economic impact of AI will be driven by11:
- Productivity gains from businesses automating processes (including use of robots and autonomous vehicles).
- Productivity gains from businesses augmenting their existing labour force with AI technologies (assisted and augmented intelligence).
- Increased consumer demand resulting from the availability of personalised and/or higher-quality AI-enhanced products and services.11
One of the most important things AI brings to personalization marketing is the ability to have such a deep customer understanding that marketing becomes simple, seamless, and as close to automatic as possible. As behavioral economist Susan Menke explains in her paper Humanizing Loyalty, “Decision fatigue and cognitive fatigue are the opposite of flow and seamlessness. We are making too many decisions that tax our cognitive bank account. We dole it out on important things and not on things that are already operating well.” In her paper, Menke touches upon the concept of the psychological script — the idea that the mind doesn’t have to focus on many day-to-day activities as they can be handled without much thought.12 The more seamless a company can make the customer interaction and purchasing process, the more likely a customer will buy from it.12 Go into just about any McDonalds in the world and you’ll see that ordering has been reduced to one word, and a often a numerical choice, “one” for a Big Mac meal deal, or “two” for a Quarter Pounder meal deal, etc. AI can help make marketing and purchasing become so seamless that customers will become conditioned to spend their money in an almost automatic fashion.
Once a mostly academic area of study, twenty-first century AI enables a plethora of mainstream technologies that are having a substantial impact on our everyday lives. Computer vision and AI planning, for example, drive the video games that are now a bigger entertainment industry than Hollywood.10 Through its AWS platform, Amazon brings natural language processing (NLP), automatic speech recognition (ASR), text-to-speech (TTS), and neural machine translation (NMT) technologies within reach of every developer. Today, brands can utilize Amazon’s AI products like Lex, Transcribe, and Comprehend to produce multilingual content for their marketing efforts. Amazon even has a product called Polly that allows users to turn text into speech in multiple languages, in voices that sound eerily human and, in some cases, almost as good as professional actors. Not to be outdone, Chinese Internet search giant Baidu has demonstrated an AI-powered tool that translates Chinese into English in real-time, right as it is being spoken.
In their article Neural Networks in Data Mining, Singh and Chauhan explain that a neural network is:
“A mathematical model or computational model based on biological neural networks, in other words, is an emulation of a biological neural system. It consists of an interconnected group of artificial neurons and processes information using a connectionist approach to computation. In most cases an ANN is an adaptive system that changes its structure based on external or internal information that flows through the network during the learning phase.”
Neural nets are extremely good at finding patterns in data. A key feature of neural networks is that they learn the relationship between inputs and outputs through training.14 For marketing purposes, neural networks can be used to classify a consumer’s spending pattern, analyze a new product, identify a customer’s characteristics, as well as forecast sales.14 The advantages of neural networks include high accuracy, high noise tolerance and ease of use as they can be updated with fresh data, which makes them useful for dynamic environments.14
Today, success in AI is partly because huge advances in computers, chips, and memory allow state of the art software to quickly run the highly complex algorithms that AI is based upon. Today, AI has the potential to radically alter the world of aviation, retail, medicine, automobiles, telcos, airlines, manufacturing, finance, insurance, government, gaming, as well as a whole host of other industries.
For marketers, AI can be broken down into five prominent segments (see Table 1)—sound, time series, text, image, and video. Areas such as CRM, customer loyalty, data governance, marketing automation, personalization, social marketing, and social listening will all be radically affected by AI and ML.
|GENERAL USE CASE||INDUSTRY|
|Voice recognition||UX/UI, Automotive, Security, IoT|
|Voice search||Handset maker, Telecoms|
|Sentiment analysis||CRM for most industries|
|Flaw detection||Automotive, Aviation|
|Fraud detection||Finance, Credit cards|
|Log analysis/Risk detection||Data centers, Security, Finance|
|Enterprise resource planning||Manufacturing, Auto, Supply Chain|
|Predictive analytics using sensor data||IoT, Smart home, Hardware manufacturing|
|Business and Economic analytics||Finance, Accounting, Government|
|Recommendation engine||E-Commerce, Media, Social Networks|
|Sentiment analysis||CRM, Social Media, Reputation mgmt.|
|Augmented search, Theme detection||Finance|
|Threat detection||Social Media, Government|
|Fraud detection||Insurance, Finance|
|Facial recognition||Multiple industries|
|Image search||Social Media|
|Machine vision||Automotive, Aviation|
|Photo clustering||Telecom, Handset makers|
|Motion detection||Gaming, UX, UI|
|Real-time threat detection||Security, Airports|
Table 1: A.I. use cases
Above all else, AI is a problem solver. One of the examples I give in my talks on AI is the idea of giving AI a goal to solve and then setting it off on its own to get an answer. For example, when it comes to marketing, the question for AI to solve might be, “How can I send an offer to a customer to assure its use?” Now, the variables to include here would be things like, what is the best offer to send, what is the best time to send it, what is the best channel to send it on, and perhaps there is a way to add social activity to increase the odds of offer use? Perhaps the customer has a potential for tweeting in the evening and the system notices that, when the customer tweets in the evening, the tweet is often followed by the opening of the brand’s marketing offers. The system discovers a strong correlation between the tweets and the opening of email offers in the past. The natural conclusion could be that the customer arrives home, sits down at his or her computer, goes through his or her email, and jumps on his or her social channels to communicate with family, friends, and, potentially, followers. The AI marketing system then watches for the evening tweet and, once it sees it, sends out the latest offer, fully expecting it to be opened within minutes of being sent out. Obviously, this is a much more powerful way to doing things than in a rules-based analytics way.
In his 15 Applications of Artificial Intelligence in Marketing, Dave Chaffey maps out the most effective AI technologies for marketing across the customer lifecycle. All the techniques are ‘AI’ in the sense that they involve computer intelligence, but Chaffey brakes them down into 3 different types of technology—Machine Learning Techniques, Applied Propensity Models, and AI Applications.15 The steps are broken down into the customer lifecycle RACE framework (see Figure 2), which contains four separate groups—Reach, Act, Convert, and Engage.15
Figure 2: A.I. for marketers across the customer lifecycle
Source: Smart Insights15
“Each different application has major implications for marketers, but the applications have different roles to play across the customer journey. Some are better for attracting customers, whilst others are useful for conversion or re-engaging past customers,” says Chaffey.15
According to Chaffey, reach “involves using techniques such as content marketing, SEO and other ‘earned media’ to bring visitors to your site and start them on the buyer’s journey.”15 “AI & applied propensity models can be used at this stage to attract more visitors and provide those that do reach your site with a more engaging experience.”15
Programmatic Media buying—the algorithmic purchase and sale of advertisements in real time—“can use propensity models generated by machine learning algorithms to more effectively target ads at the most relevant customers.”15 AI can ensure programmatic ads don’t appear on questionable websites and/or remove them from a list of sites that the advertiser doesn’t want them to appear on.15
In his Where Did All the Advertising Jobs Go? Derek Thompson explains that, “The emergence of an advertising duopoly has coincided with the rise of ‘programmatic advertising,’ a torpid term that essentially means ‘companies using algorithms to buy and place ads in those little boxes all over the internet.’” Thompson adds that, “advertising has long been a relationship-driven business, in which multimillion-dollar contracts are hammered out over one-on-one meetings, countless lunches, and even more-countless drinks. With programmatic technology, however, companies can buy access to specific audiences across several publishing platforms at once, bypassing the work of building relationships with each one.”16 Because advertising has become more automated, more ads can be produced with fewer people.16 AI should be a part of this programmatic advertising process.
In Programmatic Advertising 101: How it Works, Sara Vicioso states that, “programmatic advertising is the automated process of buying and selling ad inventory through an exchange, connecting advertisers to publishers.” This process uses AI technologies “and real-time bidding for inventory across mobile, display, video and social channels — even making its way into television.”17 Vicioso adds that, “Artificial intelligence technologies have algorithms that analyze a visitor’s behavior allowing for real time campaign optimizations towards an audience more likely to convert. Programmatic companies have the ability to gather this audience data to then target more precisely, whether it’s from 1st party (their own) or from a 3rd party data provider.”17
Programmatic media buying includes the use of demand-side platforms (DSPs), supply-side platforms (SSPs) and data management platforms (DMPs).17 DSPs facilitate the process of buying ad inventory on the open market, as well as provide the ability to reach a brand’s target audience due to the integration of DMPs.17 “DMPs collect and analyze a substantial amount of cookie data to then allow the marketer to make more informed decisions of whom their target audience may be,” explains Vicioso.17 “On the publisher side of things, publishers manage their unsold ad inventory through an SSP,”17 which reports such clickstream activity as how long a visitor was on a specific site or how many pages were viewed per visit.17 Vicioso explains that, “SSPs will ultimately be in charge of picking the winning bid and will serve the winning banner ad on the publisher’s site.”17
The second step of the RACE framework is “Act”. Brands must draw visitors in and make them aware of the company’s product and/or services. Machine learning algorithms can build propensity models that predict the likelihood of a given customer to convert, the price at which a customer is likely to convert, and/or what customers are most likely to turn into repeat customers.15
Machine learning algorithms can run through vast amounts of historical data to establish which ads perform best on which people and at what stage in the buying process.15 Using this data, ads can be served to them with the most effective content at the most effective time.15 By using machine learning to constantly optimize thousands of variables, businesses can achieve more effective ad placement and content than traditional methods.15 However, humans will still be needed for the creative parts.15
As Allie Shaw notes in her article AI could save television advertising with advanced personalization, “In short, AI programs draw from data pools to make decisions about where and when to buy or sell ad space according to demographic and cost-versus-benefit information.” “Essentially, your TV can learn about your habits in the way your web browser already does, allowing advertisers to present you with ads based on that information—so you’ll see fewer repetitive ads that you don’t care about. This means you and your neighbors may all be watching the premiere of The Walking Dead but seeing different ads based on your unique interests,” explains Shaw.18
“Thanks to programmatic TV advertising, advertisers can know how many people have viewed their ads, where these viewers are located, and what their viewing history looks like—with information updating by the minute,” says Shaw.18 “They’re also able to get more accurate data about an ad’s cost per impression (CPM, or the cost for each 1,000 people who see the ad), allowing for more relevant and cost-efficient targeting,” she explains.18
Although programmatic advertising was designed to be scalable, efficient, and precise, some brands have been reluctant to embrace it. However, as Giselle Abramovich argues, it is imperative for companies to have both a single view of the customer as well as a single view of their media.
It is not just on the client side where things are changing. There is a radical realignment going on in the advertising industry right now as well. As Derek Thompson points out in his article The Media’s Post-Advertising Future Is Also Its Past, it might be tempting to blame media’s advertiser problem and the current state of its demise as the inevitable end game of the Google and Facebook’s duopoly because the two companies already receive more than half of all the dollars spent on digital advertising, as well as command 90 percent of the growth in digital ad sales in 2017.  However, it isn’t just Facebook and Google, Thompson states20, adding that, “just about every big tech company is talking about selling ads, meaning that just about every big tech company may become another competitor in the fight for advertising revenue.”20
Amazon’s ad business exploded in the 201820; “its growth exceeded that of every other major tech company, including the duopoly,” notes Thompson.20 Wanting to move beyond just selling people iPhones, Apple is shifting its growth strategy to selling services not just phones. Meanwhile, “Microsoft will make about $4 billion in advertising revenue this year, thanks to growth from LinkedIn and Bing.”20 Thompson adds that, “AT&T is building an ad network to go along with its investment in Time Warner’s content, and Roku, which sells equipment for streaming television, is building ad tech.”20 Thompson concludes that, “In short, the future of the advertising business is being moved to technology companies managing ad networks and media companies making branded content—that is, away from the ad agencies.”16 These are cross-currents that brands need to be aware of because they are not just revolutionizing the marketing landscape but also offering huge marketing opportunities to brands willing to embrace them.
Becoming more self-sufficient may not be a bad thing for brand marketers. Marketing departments should be more self-reliant, and AI can help with that. Software that does everything from automating marketing campaigns, to inexpensively segmenting customers, to simplifying the mundane and repetitive processes of producing and categorizing content can help marketers speed up the creative process enormously. As Bill Marino and Ben Vegneron explain in their Trending Now: AI in Advertising video from the 2019 Adobe Summit, AI models can be used to recommend the optimal budget mix across advertising portfolios. AI can build landscapes for each individual programmatic advertising “bid unit”, which could be a keyword.23 A.I. can also help predict aggregated elasticity between ad spend and revenue, as well as recommend optimal budget mix across portfolios.23 AI can also propose the most optimum media mix, pace advertising in real time to ensure best pricing, and also evaluate bids hourly for improved pacing on highly seasonal days.23 AI can identify the point of diminishing returns and stay away from it, ensuring budgets are spent optimially.23
In his article Lookalike modeling breathing new life into old channels, Jordan Elkind says that lookalike marketing modeling isn’t new, “it’s been a mainstay of the ad tech industry for years, used to help advertisers expand digital audiences while maintaining relevancy of targeting.” “The principle is simple,” says Elkind, “Brands want to attract new visitors to their site. What better way to do this than to identify prospects who resemble existing visitors (or customers)?”24 “What is new,” Elkind claims, “is the dazzling variety of ways in which digital marketers are deploying lookalike modeling techniques to enhance the return on investment across marketing channels—both online and offline.”24
“With more data than ever before on user journeys and behaviors, increased adoption of platforms (like customer data platforms and data management platforms) to centralize and analyze that data, and growing ubiquity of machine learning tools and techniques, lookalike modeling is breathing new life into old channels,” says Elkind.24 “Customer-centric businesses have long recognized that the best way to acquire new visitors is to focus on users who resemble their existing visitors (or better yet, high-value customers),” explains Elkind.24 “For digital marketers looking to drive traffic and conversions, this means identifying and purchasing media against audiences based on a small number of static demographic attributes. Your recent site visitors are statistically more likely to be females, aged 18-29? Perfect—serve display advertisements to similar audiences elsewhere on the web!” says Elkind.24
“The problem,” as Elkind sees it, “is that demographic segment-based targeting, while enabling advertisers to reach audiences at scale, isn’t a great proxy for relevancy. Women aged 18-29 are a diverse demographic, only a subset of whom are likely to be interested in a brand’s offering. As a result, performance can tend to show a steep drop-off as audience size increases.”24 “Enter lookalike modeling, a form of statistical analysis that uses machine learning to process vast amounts of data and seek out hidden patterns across pools of users,” says Elkind.24 “Lookalike modeling works by identifying the composition and characteristics of a ‘seed’ audience (for example, a group of recent site visitors or high-value customers), and identifying other users who show similar attributes or behaviors,” he says.24 “By analyzing not just demographic but behavioral similarities—e.g., users who have demonstrated similar browsing patterns—lookalike modeling enables advertisers to leverage powerful and complex data signals to find the perfect audience,” says Elkind.24
“Lookalike modeling is a trusty tool in the digital media arsenal—and it’s quickly becoming indispensable to other channels as well. The convergence of ad tech and CRM—powered by platforms that enable advertisers to go well beyond cookies and CRM professionals to gain visibility into the digital journeys of known users—has made it possible to build lookalike audiences of unprecedented sophistication,” says Elkind. AI and machine learning can add even more sophistication to the process, including contextual, geolocation, social, and perhaps even emotional data.
With a single source of customer data spanning online and offline engagement, a brand can unify disparate signals of purchase intent from many customer touch points, including onsite and transaction behavior, email engagement, offline purchases, app usage, call center contacts, product reviews and more.24 This provides a rich and highly accurate view of customer Lookalike audiences can also be found on social channels like Facebook. “’Facebook Lookalike Audiences’ enables marketers to build a seed list based on pixel audiences (e.g., users who have recently visited the site or browsed a particular page) or a custom list of users,” says Elkind.24 For example, a fashion retailer “could use a platform to identify all customers with a predicted affinity—based on dozens of behavioral data points — for haute couture, and simply transfer that audience directly to Facebook. Marketers can then indicate how targeted vs. broad they would like the lookalike targeting to be.”24
For search, “getting in front of high-potential prospects when they’re in-market — searching or doing price comparison for a relevant category — is every marketer’s dream. The introduction of Similar Audiences through Google Customer Match enables marketers to automatically optimize bidding strategies around key lookalike audiences.”24
In the first quarter of 2019, LinkedIn also added lookalike marketing to its offerings. In his article LinkedIn brings lookalike audiences to B2B marketing, Andrew Blustein notes that, after a year of beta testing LinkedIn added lookalike audiences to its ads.25 “If someone searched for an article on digital marketing trends, that would map them to a category of being interested in marketing,” explained LinkedIn’s director of product, Abhishek Shrivastava.25
In his article Adobe adds new features to its data management platform, Barry Levine explains that Adobe’s Audience Manager “can now subtract traits in a lookalike model and report impressions by user segment.” “Lookalike models are often developed from the attributes of a group of users a brand wants to find more of. A model of the common attributes of the best customers, according to this thinking, can help find other users with similar attributes, who are more likely to become customers of this particular product or service,” says Levine.26 “One problem, Adobe says, is that when a brand creates a model from attribute data — either the brand’s own data or third-party data from a provider — there might be attributes that could bias the model in the wrong direction.”26 “For instance, the attributes creating the model might include visits to the brand’s site or other specific sites, when those site visits aren’t useful for finding lookalike users.”26 The new Trait Exclusion capability “lets marketers remove selected traits, and it employs Adobe’s Sensei machine learning to help make the subtraction,” says Levine.26 “In addition to removing traits that don’t add value, like site visits, Adobe said the new feature helps marketers focus on influential traits. When the brand has to comply with specific privacy regulations, the model can exclude certain demographic attributes,” he adds.26
AI – a personalization engine on steroids
Today, “Personalization”—the process of utilizing geo-location, mobile app, Wi-Fi, and OTT technology to tailor messages or experiences to an individual interacting with them — is becoming the optimum word in a radically new customer intelligence environment. Even though this personalization comes at a cost — privacy — it is a price most consumers seem more than willing to pay if a recognized value is received in return. For a marketer, “personalization” requires an investment in CRM, marketing, analytical, and social media software, but businesses should recognize that this price must be paid because highly sophisticated consumers will soon need an exceptional customer shopping experience to keep them from visiting a competitor (who will, undoubtedly, offer such services). This kind of personalization also gives the business powerful data to build optimization models that can reduce cost and increase productivity.
In her article 3 AI-driven strategies for retailers in 2019, Giselle Abramovich claims that, “Personalization is table stakes for today’s retailers, who are increasingly competing to be relevant in the hearts and minds of shoppers.” This is a great analogy as personalization will soon be the base level upon which strong customer relationships will be built, an ante that needs to be tossed in to even be in the game.
We live in an instant gratification world and the companies that will thrive in this new environment will be the ones who can both keep up with the requirements of their discerning and demanding customers as well as predict what their customers’ need throughout their customer journeys. Today, companies need every competitive advantage they can get so they can provide better service than their counterparts as the marketing landscape is radically changing underfoot. In his article How Real-time Marketing Technology Can Transform Your Business, Dan Woods makes an amusing comparison of the differing environments that marketers have to deal with today, as compared to what their 1980s counterparts might have faced28:
“Technology has changed marketing and market research into something less like golf and more like a multi-player first-person-shooter game. Crouched behind a hut, the stealthy marketers, dressed in business-casual camouflage, assess their weapons for sending outbound messages. Email campaigns, events, blogging, tweeting, PR, ebooks, white papers, apps, banner ads, Google Ad Words, social media outreach, search engine optimization. The brave marketers rise up and blast away, using weapons not to kill consumers but to attract them to their sites, to their offers, to their communities. If the weapons work, you get incoming traffic.”
AI needs to be a part of this frenetic and fast-moving environment, because connecting with people on a personal level would be completely impossible without it. It is obvious that creating a consolidated customer view is a necessary component of personalization, but, unfortunately, as Andrew Jones explains in his article Study finds marketers are prioritizing personalization…but are further behind than they realize, “most marketers today are working with customer data that is decentralized, spread across the organization in multiple databases that are updated in batch processes.” “To find success, marketers must prioritize consolidating data into a single database,” argues Jones.29 Another important step to bringing personalization efforts up to a user’s expectation level will be by using behavioral data. “In order to create these types of customer experiences, marketers must strategically collect and utilize customer data, including real-time signals of intent, which are typically not captured today,” argues Jones.29
behavior, which could power high-performing lookalike models.24
Psychographics — the study and classification of people according to their attitudes, aspirations, and other psychological criteria, especially in market research — is an important element of personalization and will become more significant as more and more data collection occurs. These days, all the large tech companies seem to be following what I call the “A-B-C-D-E’s” of data collection, i.e., Always Be Collecting Data Everywhere, all with an attitude of privacy be damned. The Cambridge Analytica-Facebook scandal is only now starting to reveal how powerful this kind of psychographic detail can be, but, because of the extensive fallout from that scandal, Facebook has tighten access to its user information. However, Twitter, with its publicly Tweet data available to all, customer psychographic detail can be gleaned from a Twitter handle. Figure 2 shows a breakdown of my Twitter profile after it has been run through IBM Watson’s Personality Insights service, which can be found at https://www.ibm.com/watson/services/personality-insights/.
Figure 3: The author’s Twitter profile run through IBM Watson™ Personality Insights
According to IBM, the IBM Watson™ Personality Insights service derives insights about personality characteristics from social media, enterprise data, or other digital communications. “The service uses linguistic analytics to infer individuals’ intrinsic personality characteristics from digital communications such as email, text messages, tweets, and forum posts,” says IBM.30 “The service infers, from potentially noisy social media, portraits of individuals that reflect their personality characteristics. It can also determine individuals’ consumption preferences, which indicate their likelihood to prefer various products, services, and activities,’ adds IBM.30
The IBM Watson™ Personality Insights service infers personality characteristics based on three primary models, the Big Five personality model, the Needs model, and the Values model. The Big Five personality model represents “the most widely used model for generally describing how a person engages with the world.”30 It includes five primary dimensions, agreeableness, conscientiousness, extraversion, emotional range, and openness.30 The Needs model “describes which aspects of a product are likely to resonate with a person.”30 The model includes twelve characteristic needs: Excitement, Harmony, Curiosity, Ideal, Closeness, Self-expression, Liberty, Love, Practicality, Stability, Challenge, and Structure.30 The Values model describes “motivating factors that influence a person’s decision making. The model includes five values: Self-transcendence / Helping others, Conservation / Tradition, Hedonism / Taking pleasure in life, Self-enhancement / Achieving success, and Open to change / Excitement.”30
In his article We Know How You Feel, Raffi Khatchadourian profiles Rana el Kaliouby, co-founder and CEO of Affectiva, a startup that specializes in AI systems that sense and understand human emotions. Affectiva develops “cutting-edge AI technologies that apply machine learning, deep learning, and data science to bring new levels of emotional intelligence to AI.” Affectiva is the most visible of a host of competing startups that are building emotionally responsive machines.31 Its competitors include Emotient, Realeyes, and Sension.31
Khatchadourian explains that, “Our faces are organs of emotional communication; by some estimates, we transmit more data with our expressions than with what we say, and a few pioneers dedicated to decoding this information have made tremendous progress.”31 “Since the nineteen-nineties a small number of researchers have been working to give computers the capacity to read our feelings and react, in ways that have come to seem startlingly human,” explains Khatchadourian.31 Researchers “have trained computers to identify deep patterns in vocal pitch, rhythm, and intensity; their software can scan a conversation between a woman and a child and determine if the woman is a mother, whether she is looking the child in the eye, whether she is angry or frustrated or joyful.”31 “Other machines can measure sentiment by assessing the arrangement of our words, or by reading our gestures. Still others can do so from facial expressions,” says Khatchadourian.31
In his book Architects of Intelligence32, Martin Ford interviews Kaliouby, and her thesis is “that this kind of interface between humans and machines is going to become ubiquitous, that it will just be ingrained in the future human-machine interfaces, whether it’s our car, our phone or smart devices at our home or in the office.”32 She sees a world where, “We will just be coexisting and collaborating with these new devices, and new kinds of interfaces.”32 “I think that, ten years down the line, we won’t remember what it was like when we couldn’t just frown at our device, and our device would say, ‘Oh, you didn’t like that, did you?’” says Kaliouby.31
Afectiva’s signature software, Affdex, tracks four emotional “classifiers”—happy, confused, surprised, and disgusted.31 “The software scans for a face; if there are multiple faces, it isolates each one. It then identifies the face’s main region—mouth, nose, eyes, eyebrows—and it ascribes points to each, rendering the features in simple geometries,” explains Khatchadourian.31 “Affdex also scans for the shifting texture of skin—the distribution of wrinkles around an eye, or the furrow of a brow—and combines that information with the deformable points to build detailed models of the face as it reacts,” says Khatchadourian.31 The algorithm identifies an emotional expression by comparing it with countless others that it has previously analyzed. “If you smile, for example, it recognizes that you are smiling in real time,” Kaliouby says.31
“People are pretty good at monitoring the mental states of the people around them,” says Kaliouby.32 “We know that about 55% of the signals we use are in facial expression and your gestures, while about 38% of the signal we respond to is from tone of voice. So how fast someone is speaking, the pitch, and how much energy is in the voice. Only 7% of the signal is in the text and the actual choice of words that someone uses!”32A multi-billion-dollar industry that tracks people’s sentiments about this product or that service has been built within just a couple of years ago, which is amazing when you think that all of these tweets, likes and posts only account for about 7% of how humans communicate overall.32 “What I like to think about what we’re doing here, is trying to capture the other 93% of non-verbal communication,” contends Kaliouby.32
Kaliouby believes Affectiva’s technology has the potential to monetize what she calls an ‘Emotion Economy’. 31 “Tech gurus have for some time been predicting the Internet of Things, the wiring together of all our devices to create ‘ambient intelligence’—an unseen fog of digital knowingness,” explains Khatchadourian.31 Emotion could be a part of this IoT.31 Kaliouby predicts that, in the coming years, mobile devices will contain an “emotion chip,” which constantly runs in the background, the way geolocation currently works on phones now.31 “Every time you pick up your phone, it gets an emotion pulse, if you like, on how you’re feeling,” Kaliouby says.31
Affectiva has filed a patent for “a system that could dynamically price advertising depending on how people responded to it.”31 However, when they did, they discovered that they were not alone; more than a hundred similar patents for emotion-sensing technology existed, many of them, unsurprisingly, also focused on advertising.31 Companies like AOL, Hitachi, eBay, IBM, Yahoo!, and Motorola are also developing technology in this space.31 Sony had filed several patents; “its researchers anticipated games that build emotional maps of players, combining data from sensors and from social media to create ‘almost dangerous kinds of interactivity,’” notes Khatchadourian.31
For Affectiva, there is now plenty of interest in its Affdex solution.31 The company has conducted research for Facebook, experimenting with video ads.31 Samsung has licensed it and a company in San Francisco wants to give its digital nurses the ability to read faces.31 A Belfast entrepreneur is interested in using it at night clubs.31 A state initiative in Dubai, the Happiness Index, wants to measure social contentment: “Dubai is known to have one of the world’s tightest CCTV networks, so the infrastructure to acquire video footage to be analyzed by Affdex already exists,” says Kaliouby.31 All-in-all, Affectiva could be showing us the future of customer engagement. Although somewhat Big Brotheresque, all of this data collection is incredibly seamless, which means it will probably be popping up in all kinds of technology in the coming years.
Morphing is one of the ways a brand can hyper-personalize the customer shopping experience. So, what exactly is morphing? In their article Website Morphing, Hauser et al. state that, “’Morphing’ involves automatically matching the basic ‘look and feel’ of a website, not just the content, to cognitive styles.” Hauser et al. use Bayesian updating to “infer cognitive styles from clickstream data.”33 They then “balance exploration (learning how morphing affects purchase probabilities) with exploitation (maximizing short-term sales) by solving a dynamic program (partially observable Markov decision process).”33
In a world of deep personalization, website design becomes a major profit driver, contend Hauser et al. 33 As the writers see it, “Websites that match the preferences and information needs of visitors are efficient; those that do not forego potential profit and may be driven from the market.”33 The authors believe that “retailers might serve their customers better and sell more products and services if their websites matched the cognitive styles of their visitors.”33 I’d argue it is not just retailers who would profit from this, most B-2-C companies who have a big web presence would.
Keeping with the themes of simplicity and seamlessness, Hauser et al. do not believe personal self-selection — the process in which a customer is given many options and allowed to select how to navigate and interact with the site — is viable.33 “As the customer’s options grow, this strategy leads to sites that are complex, confusing, and difficult to use,” they argue.33 The second option, which requires “visitors to complete a set of cognitive style tasks and then select a website from a predetermined set of websites”33 is just as problematic. Website visitors probably won’t see value in taking the time to answer these questions and there is always the problem of self-bias hindering any potential results.33
Hauser et al. propose another approach: “’morphing’ the website automatically by matching website characteristics to customers’ cognitive styles.”33 A cognitive style is “a person’s preferred way of gathering, processing, and evaluating information” It can be identified as “individual differences in how we perceive, think, solve problems, learn and relate to others.” “A person’s cognitive style is fixed early on in life and is thought to be deeply pervasive [and is] a relatively fixed aspect of learning performance,” contend Riding and Rayner.
The “goal is to morph the website’s basic structure (site backbone) and other functional characteristics in real time.”33 “Website morphing complements self-selected branching (as in http://www.Dell.com), recommendations (as in http://www.Amazon.com), factorial experiments (Google’s Website Optimizer), or customized content .”33
For Hauser et al., cognitive styles dimensions “might include impulsive (makes decisions quickly) versus deliberative (explores options in depth before making a decision), visual (prefers images) versus verbal (prefers text and numbers), or analytic (wants all details) versus holistic (just the bottom line).”33 For example, “a website might morph by changing the ratio of graphs and pictures to text, by reducing a display to just a few options (broadband service plans), or by carefully selecting the amount of information presented about each plan. A website might also morph by adding or deleting functional characteristics such as column headings, links, tools, persona, and dialogue boxes.”33 There are, literally, hundreds of thousands or even millions of ways a website can morph to better serve its customers and AI has to be an integral part of this process simply because of the enormous amount of data involved.
“If a picture is worth a thousand words, visual search—the ability to use an image to search for other identical or related visual asset—is worth thousands of spot-on searches—and thousands of minutes saved on dead-end queries,” says Brett Butterfield in his Adobe blog See It, Search It, Shop It: How AI is Powering Visual Search. Butterfield argues that visual search could become a big part of a buyer’s shopping future. With visual search, a potential buyer doesn’t need to try and guess the brand, style, and/or retail outlet something was purchased on, the buyer can simply snap a picture of the item she likes, upload the image, and immediately find exactly the same item or ones like them, and purchase them directly from the site where they are sold, all incredibly seamlessly.39
“That spot-it/want-it scene is common, and good for business. It could be a shirt on someone walking down the street, an image on Instagram, or a piece of furniture in a magazine—somewhere, your customer saw something that made them want to buy one, and now they’re on a mission to find it,” explains Butterfield.39 Making something easy to find doesn’t have to be that hard and AI can help.
“While it’s a seemingly simple task, in many cases the path from seeing to buying is a circuitous and friction-filled route that leads to a subpar purchase—or no purchase at all. Just one in three Google searches, for example, leads to a click—and these people come to the table with at least a sense of what they’re searching for,” notes Butterfield.39
“Like text-based search, visual search interprets and understands a user’s input—images, in this case—and delivers the most relevant search results possible. However, instead of forcing people to think like computers, which is how the typical text search works, visual search flips the script,” adds Butterfield.39 “Powered by AI, the machine sees, interprets, and takes the visual cues it learns from people. After applying metadata to the image, AI-powered visual search systems can dig through and retrieve relevant results based on visual similarities, such as color and composition,” explains Butterfield.39 Visual search is another technology that can facilitate better, more frictionless retail experiences that can help buyers find what they want faster.39 Visual search is another good reason why companies need to get their metadata house in order. Properly tagged retail images and videos will help reduce fruitless buyer searches, as well as assist buyers to find the products they clearly want.
“One early adopter of visual search is Synthetic, Organic’s cognitive technology division, an Omnicom subsidiary,” notes Butterfield.39 “Synthetic’s Style Intelligence Agent (SIA) — powered by Adobe Sensei — uses AI to help customers not just find specific clothing items, but also find the right accessories to complete their new look.”39 To use SIA, customers simply upload an image, either from a website, from real life or even from an ad in a magazine and from there, “Adobe Sensei’s Auto Tag service extracts attributes from the image based on everything from color, to style, to cut, to patterns.”39 SIA’s custom machine-learning model then kicks in, correlating those tags with a massive catalog of products.39 “SIA then displays visually similar search results as well as relevant recommendations—items with similar styles, cuts, colors, or patterns, for example.”39 Just as importantly, SIA then “uses these visual searches to build a rich profile for that customer’s preferences and tastes—a much deeper profile than what could be built from text-based searches alone.”39 Here you are getting customer preferences on steroids, an enormous of amount of personalized data that can then be used in marketing.
“In delivering such a simple, seamless experience, AI-powered visual search removes the friction from traditional search-and-shop experiences,” says Butterfield.39 “No longer do customers have to visit multiple retailers or sites and strike out. They can now find virtually anything, anywhere, even without knowing exactly where to find it,” he adds.39 This is another important moment for marketers because if brands invest in visual search, they can propel their brand up the Google rankings and get a solid leg up on their competition.
To get started with image search, brands should focus on solving customer problems and getting their own visual assets in order.39 They shouldn’t try to make their visual search workflows all about advertising.39 Instead, brands should “aim to have solid metadata on products so that searching is easier and more natural.”39 From there, brands should work towards “visual search processes that are real time and increasingly intuitive, creating a positive customer experience that keeps people coming back,” recommends Butterfield.39
The moving picture, i.e. movies and video, will also be radically affected by AI. In his article The Future of Video Advertising is Artificial Intelligence, Matt Cimaglia sees a video advertising world that is completely different to the one we currently see. He describes it as such: “Meanwhile, somewhere in another office, in that same year, a different team is creating a different digital video. Except they’re not shooting a single video: They’re shooting multiple iterations of it. In one, the actor changes shirts. In another, the actor is an actress. In another, the actress is African American. After finishing the shoot, this agency doesn’t pass the footage off to a video editor. They pass it off to an algorithm.”40
Cimaglia states that, “The algorithm can cut a different video ad in milliseconds. Instead of taking one day to edit one video, it could compile hundreds of videos, each slightly different and tailored to specific viewers based on their user data.”40 “As the video analytics flows in, the algorithm can edit the video in real-time, too—instead of waiting a week to analyze and act on viewer behavior, the algorithm can perform instantaneous A/B tests, optimizing the company’s investment in a day,” claims Cimaglia.40
Cimaglia believes this is what is happening right now.40 “We are witnessing a moment in video marketing history, like moments experienced across other industries disrupted by the digital revolution, where human editors are becoming obsolete,” contends Cimaglia.40 This is the next evolution of advertising—personalized advertising, i.e., tailoring content to individuals rather than the masses40; surgically striking relevant offers to a market of one, rather than blasting a shotgun of offerings to the uninterested many.
“Savvy agencies are turning to artificial intelligence for help making those new, specialized creative decisions,” says Cimaglia.40 “It’s the same logic that’s long overtaken programmatic banner and search advertising, machine learning and chatbots: There are some things computers can do faster, cheaper and more accurately than humans,” contends Cimaglia. “In this future of data-driven dynamic content, viewers’ information is siphoned to AI that determines aspects of the video based on their data,” explains Cimaglia.40
Cimaglia sees advertising being tailored towards individuals.40 “The options for customization extend beyond user data, too. If it’s raining outside, it could be raining in the video,” easily done by the agency plugging in a geolocating weather script.40 Similarly, if a user is watching the video at night, the video could mirror reality and be a night scene filled with cricket sounds.40 For Cimaglia, “This is a logical progression for a society already accustomed to exchanging their privacy for free services.”40 The video could also be in multiple languages thanks to tools like Amazon Polly.
In its article The Magic of AI in a content-driven world. Using AI to create content faster, the Adobe Enterprise Content Team argues that we are currently in the midst of a content explosion. Perhaps because of this, it is also a time when “Consumers expect to have personalized, relevant experiences at all times, in all places, and on all platforms.”41
When thinking about what is needed to create this kind of content for thousands or even millions of customers at the near real-time speed that is necessary, doing it manually is impossible.41 Adobe’s State of Creativity in Business 2017 survey found that “40 percent of creatives are using AI in photo and design retouching,”41 so it’s already being heavily utilized. Currently, it can take hours for a designer to find just the right image to use in a piece of marketing collateral, and that’s not counting the time required to manipulate the image, to crop it, to find the right layout scenario, and then to publish it to an online catalog and/or social media channel.41 Serving the right content to the right person at the right time adds even more time.41 The cost for all this work adds up, as does the cost of photo shoots to create new assets.41
According to the Adobe Sensei Team, “AI can help you create more relevant content and more engaging experiences across the customer journey at the speed your customers expect. On the creative side, AI can speed up all kinds of tedious tasks, from identifying and organizing assets to adjusting and refining for specific channels.” “On the audience level, AI can help you better understand which audiences respond to which content, or how often people prefer to receive emails, so you can deliver the experiences your customers want while respecting their preferences and privacy.”42
“Designers simply don’t have time to tag the hundreds of images uploaded from every photo shoot. Even if they did, the list of keywords probably wouldn’t be as exhaustive as it should be. But when a photo isn’t tagged, it’s virtually impossible to find by searching in an image bank of thousands,” contend the Adobe Experience Cloud team.42 “According to IDC, marketers report that one-third of marketing assets go unused or underutilized with the average organization creating hundreds of new marketing assets each year.”42 Repurposing images is unlikely, which means ROI suffers. To try to tackle this issue, Adobe has created “Auto Tag”, an Adobe Sensei capability that automatically tags images with key words.42 For example, a marketer might have a picture of a young girl on a beach under a clear blue sky, which could be tagged with keywords like “beach”, “girl”, “dancing”, “sundress”, “blue sky”, “white sand”, or even a place like “Aruba.”42 Using the Sensei framework, marketers could train their AI and machine learning models to create their own unique auto tags.42 This includes identifying brand characteristics like the company logo, so the designers adhere to specific brand standards, or training it to identify a company’s products so that they can be tagged in pictures on social media, which helps identify true reach.42
Custom auto tagging not only has the potential to increase a marketing team’s efficiency, but it could lead to image-based shopping.42 If a customer who is looking for a new couch uploads a photo of one they like and then shop for something similar based solely on the image, that’s metadata the brand can use.42 “Auto tagging identifies what is in the photo and finds the best matches for the customer. Auto tagging also allows brands to gain a deeper understanding of their audience and can help uncover market trends on social media, without the brand having to rely on tags and text. “If you run a social media feed through Adobe Sensei, it will tag places your brand is pictured—even if it’s not mentioned or tagged—allowing you to see what is trending,” explains the Enterprise Content Team.42
The third step of the RACE framework — “Content” — is one of the most important steps and it includes dynamic pricing, re-targeting, web and app personalization, and chatbots.15 “Much like with ad targeting, machine learning can be used to establish what content is most likely to bring customers back to the site based on historical data,” says Chaffey.15 By building an accurate prediction model of what content works best to win back different customer types, machine learning can help optimize a brand’s retargeting ads to make them as effective as possible.15
Another way to convert customers is with chatbots that mimic human intelligence by interpreting a consumer’s queries and potentially complete an order for them.15 Chatbots are relatively easy to build and Facebook is simplifying the process of developing chatbots for brands.15 Facebook “wants to make its Messenger app the go-to place for people to have conversations with brand’s virtual ambassadors.”15 Facebook has created the wit.ai bot engine, which allows brands to train bots with sample conversations and have these bots continually learn from customer interactions.15
In her article 3 AI-Driven Strategies For Retailers In 2019, Giselle Abramovich claims that chatbots are probably “the most common AI-powered customer service application today.”27 “To date, bots have predominantly been used to provide search and discovery and product recommendations,” says Abramovich.27
There are several commercial chatbot development platforms that simplify the creation and maintenance of chatbots. These let users add more functionality to a bot by creating a flow, machine learning capabilities, and API integration. According to, Maruti Tech, chatbot platforms can be simple to use, and deep technical knowledge or programming skills are not required as many chatbot platforms come with drag-and-drop functionality.
Calling itself “the leading bot platform for creating AI chatbots for Facebook,” Chatfuel claims that 46% of Messenger bots run on its platform.44 No coding is required with Chatfuel, which “provides features like adding content cards and sharing it to your followers automatically, gathering information inside Messenger chats with forms.”44 Chatfuel also uses AI to script interactive conversations.44
“Let your bot chat like a human” is Botsify’s tagline. It is another popular Facebook Messenger chatbot platform that uses a drag and drop template to create bots.44 Botsify offers features like easy integrations via plugins, Smart AI, Machine learning and analytics integration.43 Botsify’s platform does allow seamless transition from bot to human as well, which isn’t always the case with the other platforms.43
Other chatbot development platforms include Motion.ai, Flow XO, Chatty People, Recast.ai, Botkit, ChatterOn, Octane AI, Converse AI, GupShup, and Microsoft’s QnA Bot, which can be integrated into Microsoft Cognitive Services to enable the bot to see, hear, interpret and interact in more human ways.43 QnA Maker seamlessly integrates with other APIs and can scale to be a know-it-all part of a bigger bot.43
With the availability of such platforms, Maruti Tech argues that anyone can create a chatbot, even if they don’t know how to code.43 However, to make an intelligent chatbot that works seamlessly, AI, machine learning and NLP are required.43 Chatbots will undoubtedly revolutionize the future of industries by their rich features.43 They will reduce human errors, “provide round the clock availability, eliminate the need for multiple mobile applications and make it a very seamless experience for the customer.”43
The final step of the RACE framework is “Engage”. As several studies have shown, it is far easier to sell to an existing customer than it is to attract new ones, therefore keeping current customers happy is of paramount importance.15 “This is particularly true in subscription-based business, where a high churn rate can be extremely costly,” advises Chaffey.15 “Predictive analytics can be used to work out which customers are most likely to unsubscribe from a service, by assessing what features are most common in customers who do unsubscribe,” says Chaffey.15 “It’s then possible to reach out to these customers with offers, prompts or assistance to prevent them from churning,” recommends Chaffey.15
Today, there is almost a FOMO or fear-of-being-left-out attitude to AI, but one needs to be extremely cautious when jumping onto the latest and greatest technology, as plenty of wiped-out crypto and blockchain investors can attest to. Anyone planning to utilize AI should be aware that the technology is not new, but many of the so-called experts are and their depth of knowledge exceedingly shallow.
In his MIT Technology Review article Is AI Riding a One-Trick Pony?, James Somers notes that, “Just about every AI advance you’ve heard of depends on a breakthrough that’s three decades old.” Of course, many recent advances in hardware and software technology have turned AI’s potential into reality, but still anyone wishing to jump into AI should have a good understanding of where it came from and how long it has been around, as well as the limitations inherent in the technology. In reality, the technology is decades old and might be near the end of its cycle, but it is still some of the most powerful technology around.
Not only is AI old, but it is also a difficult technology to implement. In its Conquer the AI Dilemma by Unifying Data Science and Engineering, Databricks says that only “1 in 3 AI projects are successful and it takes more than 6 months to go from concept to production, with a significant portion of them never making it to production — creating an AI dilemma for organizations.”
Dan Woods frenetic marketing world that includes “Email campaigns, events, blogging, tweeting, PR, ebooks, white papers, apps, banner ads, Google Ad Words, social media outreach, search engine optimization”28 would be impossible to operate with normal marketing automation technology. Lookalike modeling that includes contextual, geolocation, social, and perhaps even emotional data would overwhelm normal CRM systems as well. Personality insights services, website morphing, chatbot services, programmatic advertising, emotional, image, and facial recognition technology all require massive amounts of data to be crunched in real-time and would be almost useless without AI, machine learning, and deep learning.
Bill Gates has stated that “If you invent a breakthrough in artificial intelligence, so machines can learn, that is worth ten Microsofts.” AI sure seems to have a million uses, but we do need to keep an eye on the potential threat that AI holds. Hopefully, we’re just at the end of the beginning of AI’s technological curve and not at the beginning of the end of us all.
It has been a few years since I had that lunch with my friend at HKUST’s China Gardens restaurant, which is perched high above Hong Kong’s stunning Clearwater Bay, and I have come to realize that my friend’s question really went to the very heart of the simplicity and beauty that is AI. Perhaps he was one of those who had labored through the “AI winter” and found the technology so frustrating to work with because of the limitations of the tools he was using them on, but he had a good point. We often get caught up in the buzzword mumbo jumbo of technological hype. The professor is correct about AI, it is just a neural net that goes around and around improving upon itself as it learns more and more and more about a problem. But then again, what beauty and form can be created out of that little old neural net.
 Global Challenges Foundation. (2015). 12 Risks That Threaten Human Civilization. The Case for a New Category Risk. https://api.globalchallenges.org/static/wp-content/uploads/12-Risks-with-infinite-impact-full-report-1.pdf (Accessed 2 June 2019).
 Clifford, Catherin. Google CEO: A.I. is more important than fire or electricity. Cnbc.com. February 1, 2018. https://www.cnbc.com/2018/02/01/google-ceo-sundar-pichai-ai-is-more-important-than-fire-electricity.html (Accessed 6 June 2019).
 IBM. How to get started with cognitive computing. https://www.ibm.com/watson/advantage-reports/getting-started-cognitive-technology.html (Accessed 6 June 2019).
 Metx, Cadz. Building AI is hard—so Facebook is building AI that builds AI. Wired. May 6, 2016. https://www.wired.com/2016/05/facebook-trying-create-ai-can-create-ai/ (Accessed 7 June 2019).
 SAS. (2019). SAS announces $1 Billion investment in Artificial Intelligence (AI). March 17, 2019. https://www.sas.com/en_us/news/press-releases/2019/march/artificial-intelligence-investment.html (Accessed 2 June 2019).
 AdAge. Burger King’s A.I.-created campaign pokes fun at A.I. September 27, 2018. https://adage.com/creativity/work/burger-king-ai-ad-whopper/950441 (Accessed 10 June 2019).
 Adobe. Artificial intelligence Unlocks the True Power of Analytics. https://www.adobe.com/au/insights/ai-unlocks-the-true-power-of-analytics.html (Accessed 4 June 2019).
 UPS Customer Solutions & Innovations. (2018). Artificial Intelligence in Logistics, a collaborative report by DHL and IBM on implications and use cases for the logistics industry. https://www.logistics.dhl/content/dam/dhl/global/core/documents/pdf/glo-ai-in-logistics-white-paper.pdf (Accessed 29 May 2019).
 SAS, Accenture Applied Intelligence, and Intel, with Forbes Insights. (2018). AI Momentum, Maturity, & Models for Success. sas.com.
 Artificial Intelligence and Life in 2030. September 2016. https://ai100.stanford.edu/sites/default/files/ai_100_report_0831fnl.pdf (Accessed 1 June 2019).
 PWC. (2017). Sizing the prize. What’s the real value of AI for your business and how you can capitalise. https://www.pwc.com/gx/en/issues/analytics/assets/pwc-ai-analysis-sizing-the-prize-report.pdf (Accessed 9 June 2019).
 Humanizing Loyalty A road map to establishing genuine emotional loyalty at scale. Olson1to1.com. https://go.icf.com/rs/072-WJX-782/images/Humanizing%20Loyalty%20-%20Olson%201to1.pdf (Accessed 6 June 2019).
 Kharpal, Arjun. Baidu Challenges Google with AI That Translates Languages in Real-Time. CNBC. October 23, 2018. https://www.cnbc.com/2018/10/24/baidu-challenges-google-with-ai-that-translates-languages-in-real-time.html (Accessed 10 June 2019).
 Singh Y, Chauhan AS. Neural Networks in Data Mining. Journal of Theoretical and Applied Information Technology. 2009; 5:37-42 http://jatit.org/volumes/research-papers/Vol5No1/1Vol5No6.pdf (Accessed 20 November 2017).
 Chaffey, Dave. (2019) 15 Applications of Artificial Intelligence in Marketing. January 31, 2019. https://www.smartinsights.com/managing-digital-marketing/marketing-innovation/15-applications-artificial-intelligence-marketing/ (Accessed 18 March 2019).
 Thompson Derek. (2018). The Atlantic. Where Did All the Advertising Jobs Go? February 7, 2018. https://www.theatlantic.com/business/archive/2018/02/advertising-jobs-programmatic-tech/552629/ (Accessed 1 June 2019).
 Vicioso, Sara. Seer Interactive. Programmatic Advertising 101: How it Works. August 27, 2015. https://www.seerinteractive.com/blog/programmatic-advertising-101-works/ (Accessed 6 June 2019).
 Shaw, Allie. (2017). VentureBeat. AI could save television advertising with advanced personalization. October 28, 2017. https://venturebeat.com/2017/10/28/ai-could-save-television-advertising-with-advanced-personalization/ (Accessed 5 June 2019).
 Abramovich, Giselle. (2017). CMO. Future of Advertising: Automated, Personalized, and Measureable. 26 September 2017. https://www.cmo.com/features/articles/2017/9/25/the-future-of-advertising-automated-personalized-and-measurable-.html (Accessed 3 June 2019).
 Thompson, Derek. (2018) The Atlantic. The Media’s Post-Advertising Future Is Also Its Past. December 31, 2018. https://www.theatlantic.com/ideas/archive/2018/12/post-advertising-future-media/578917/ (Accessed 1 June 2019).
 Ballentine, Claire. (2018) New York Times. 12 August 2018. Google-Facebook Dominance Hurts Ad Tech Firms, Speeding Consolidation. https://www.nytimes.com/2018/08/12/technology/google-facebook-dominance-hurts-ad-tech-firms-speeding-consolidation.html (Accessed 1 June 2019).
 Mickel, Tripp & Wells, Georgia. (2018) Wall Street journal. Apple Looks to Expand Advertising Business with New Network for Apps. June 1, 2018. https://www.wsj.com/articles/apple-looks-to-expand-advertising-business-with-new-network-for-apps-1527869990?utm_source=newsletter&utm_medium=email&utm_campaign=newsletter_axiosmediatrends&stream=top (Accessed 1 June 2019).
 Elkind, Jordan. (2017). Lookalike modeling breathing new life into old channels. Martechtoday. 2017. August 30, 2017. https://martechtoday.com/lookalike-modeling-breathes-new-life-old-channels-203250 (Accessed 2 June 2019).
 Blustein, Andrew. (2019). LinkedIn brings lookalike audiences to B2B marketing. The Drum. 21 March 2019. https://www.thedrum.com/news/2019/03/21/linkedin-brings-lookalike-audiences-b2b-marketing (Accessed 8 June 2019).
 Levine, Barry. (2018). Marketingland.com. Adobe adds new features to its data management platform. December 3, 2018. https://marketingland.com/adobe-adds-new-features-to-its-data-management-platform-252944 (Accessed 2 June 2019).
 Abramovich, Giselle. (2018). CMO.com. 3 AI-driven strategies for retailers in 2019. 11 November 2018. https://www.cmo.com/features/articles/2018/11/15/3-ai-driven-strategies-for-retailers-in-2019.html#gs.COMQdxdU (Accessed 3 June 2019).
 Woods, D. (2011, May 6). How Real-time Marketing Technology Can Transform Your Business. Retrieved from Forbes.com: http://www.forbes.com/sites/ciocentral/2011/05/06/how-real-time-marketing-technology-can-transform-your-business (Accessed 3 June 2019).
 Jones, Andrew (2015, December 15). Study finds marketers are prioritizing personalization…but are further behind than they realize, http://venturebeat.com/2015/12/14/study-finds-marketers-are-prioritizing-personalization-but-are-further-behind-than-they-realize/ (Accessed 6 June 2019).
 Khatchadourian, Raffi. (2015). The New Yorker. We Know How You Feel. January 19, 2015. https://www.newyorker.com/magazine/2015/01/19/know-feel. (Accessed 2 June 2019).
 Ford, Martin. (2018). Architects of Intelligence. The truth about AI from the people building it. Packt Publishing; 1 edition. November 23, 2018.
 Hauser, John R., Urban, Glen, Liberali, Guilherme, Braun Michael. (2009). “Website Morphing.” Marketing Science 28.2 (2009): 202-223. © 2009 Informs. https://www.researchgate.net/publication/41822749_Website_Morphing (Accessed 16 February 2019).
 Hayes, J., C. W. Allinson. 1998. Cognitive style and the theory and practice of individual and collective learning in organizations. Human Relations 31(7) 847–871.
 Witkin, H. A., C. Moore, D. Goodenough, P. Cox. 1977. Field dependent and field-independent cognitive styles and their educational implications. Rev. Educational Res. 47(1) 1–64.
 Riding, R. J., S. Rayner. 1998. Cognitive Styles and Learning Strategies Understanding Style Differences in Learning and Behavior. David Fulton Publishers, London.
 Ansari, A., C. F. Mela. 2003. E-customization. J. Marketing Res. 40(2) 131–145
 Montgomery, A. L., S. Li, K. Srinivasan, J. Liechty. 2004. Modeling online browsing and path analysis using clickstream data. Marketing Sci. 23(4) 579–585.
 Butterfield, Brett. (2018). Adobe. See It, Search It, Shop It: How AI is Powering Visual Search. 12 December 2018. https://theblog.adobe.com/see-it-search-it-shop-it-how-ai-is-powering-visual-search/ (Accessed 20 January 2019).
 Enterprise Content Team. The Magic of AI in a content-driven world. Using AI to create content faster. https://www.adobe.com/insights/the-magic-of-AI-in-a-content-driven-world.html (Accessed 5 June 2019).
 Adobe Sensei Team. Adobe. AI: Your behind-the-scenes marketing companion. https://www.adobe.com/insights/sensei-ai-for-marketers.html (Accessed 7 June 2019).
 Maruti Tech. 14 Powerful Chatbot Platforms. https://www.marutitech.com/14-powerful-chatbot-platforms/ (Accessed 8 June 2019).
 Somers, James. (2017) MIT Technology Review. Is AI Riding a One-Trick Pony? 29 September 2017. https://www.technologyreview.com/s/608911/is-ai-riding-a-one-trick-pony/ (Accessed 8 June 2019).
 Databricks. Conquer the AI Dilemma by Unifying Data Science and Engineering. https://databricks.com/company/newsroom/press-releases/databricks-conquers-ai-dilemma-with-unified-analytics (Accessed 8 June 2019).