Defining Terms

As with most new technology, AI comes with a bundle of new terms to learn. With AI terminology, one of the key things to remember is that there’s a tremendous amount of overlap between terms - even experts debate the exact boundaries. Throughout the report, we’ll be using ‘AI’ as a catch-all term, but some of the more important words and phrases include;

• Artificial Intelligence - AI is an umbrella term for the field of computer science which attempts to emulate the cognitive functions of the human mind, such as learning, reasoning and problem solving. It is often divided into ‘weak’ or ‘narrow’ AI, designed for one particular task, and ‘strong’ AI, which has enough apparent cognisance that it can find solutions to even unfamiliar tasks - or even suggest tasks that users hadn’t thought of.
• Algorithm - At its simplest, an algorithm is a set of mathematical instructions that can be followed by a computer. Often visualised as a flow chart, these rules are followed in a particular order to process an input value into the desired output, whether that’s sorting data or solving a problem.
• Machine Learning (and the difference from AI) - Machine learning is the ability of software and its underlying algorithm(s) to evolve based on its own output - to change itself for the better, beyond what has been explicitly programmed by a human. It performs tasks repeatedly, measures its performance, and applies those data points to improve the process – much like a human learning from their mistakes.

Some of the more complex, nice-to-know terms include;

• Deep Learning - Deep learning is a type of machine learning which removes the need for ongoing human interaction from the equation. Rather than a programmer telling an algorithm what a ‘better’ result looks like, deep learning is able to make those assessments itself.
• Neural Networks - Inspired by the structure of the human brain, neural networks use ‘neurons’ arranged in layers to process data - A neuron is simply a mathematical function. Data is input at the first layer; elements of that data passes through subsequent layers until an output layer is reached. Neurons can activate or inhibit neurons in the next layer depending on the result of their own calculation. In this way, a path to output is found.
• Natural Language - Processing Natural language processing is the process of interpreting human language – written or spoken – so that it can be interpreted as a command by the AI. This creates a naturalistic interface that’s accessible to pretty much anyone, regardless of their technological skill and experience.

Introduction

Chances are, you’d heard of AI well before the past couple of years. The term has been around since 1956, and the concept considerably longer than that. It can be traced back as far as the 14 th century.1 The Lullian Circle, developed by Ramon Llul in the 1300s, is seen as the first instance a machine was used to perform useful reasoning. Depending on your age, maybe you first encountered the idea of AI as HAL 9000 in 2001: A Space Odyssey , through Tony Stark’s ‘Jarvis’ assistant in the Iron Man films, or in news stories about Deep Blue beating chess champs at their own game. More recently, Google’s voice assistant, Amazon’s Alexa, Microsoft’s Cortana and Apple’s Siri are overtly bringing smart computing, underpinned by AI, into our everyday lives.

Google CEO Sundar Pichai predicted that during the next 10 years, we will “shift to a world that is AI-first”. A report from Tractica predicts that annual AI revenues will soar from $1 billion in 2015 to $36 billion by 2025.

So why is AI suddenly such a hot topic in the marketing world? It’s partly because the technology has finally come of age - especially with the exponential growth of cloud computing, which makes it cheaper and easier to handle the processing that AI requires. But, at least as important to the rise of AI, is that it offers a solution to a problem facing almost every marketer today: no human can deal with the sheer amount of data out there.

Whether it’s purchase histories, social media posts, location data, user preferences, we now have access to an overwhelming amount of information about customers. By the end of 2016, it was estimated that global internet traffic surpassed 1 zettabyte . This is equal to 1 billion terabytes worth of data, or 1 trillion gigabytes. In order to be useful, all of this data needs to be processed. As data grows exponentially, AI is able to step in and take some of the strain. So let’s look at a selection of the ways it’s being used to do that in marketing today.

AI in Advertising

One form of AI is already well established within digital advertising. Programmatic trading utilises AI to optimise campaigns, by measuring the success of an ad and using self-improving algorithms to target ads and make intelligent choices about creative, timing, the amount to bid for in impression and more. It’s also used to analyse consumer data, tapping into the constant streams of data being created by consumers online

Christopher Pitt, Head of Marketing at search marketing specialist Vertical Leap concurs. “Where AI is key to marketing is liberating the creative elements from the manual, time-consuming drudgery of data collection. Previously, 75% of our time was spent on analysis, with 25% of our time spent on the actual work. AI has enabled us to switch that round and to achieve more effective and efficient results.” Vertical Leap have implemented AI in their business, and are seeing positive results. Pitt says that AI is “freeing up marketers to do more actual marketing with increasingly less time and budget”.

AI has also started to work its way into the creative side of advertising, in a variety of ways. In some cases, that’s by integrating AI functionality into the ads themselves. IBM’s Watson Ads launched on Weather.com last year with campaigns from the likes of Toyota and Unilever. This ‘cognitive advertising’ format lets consumers interact with the ad, asking for brand-relevant recipes, having a conversation about car features, or clothes that complete their outfit.

AI is even having an impact on creative decision-making. PicassoLabs is a company which uses AI to spot patterns between images that drive the most user engagement, helping marketers make informed decisions about which imagery they should use in campaigns. Through both its AdWords and display platforms, Google has been learning what makes good creative or selecting different creative based on customer traits for years - This is AI at work.

There are many more examples out there too: In 2015, Clear Channel and M&C Saatchi experimented with a ‘Darwinian’ digital out-of-home campaign, which created new creative variants based on the reaction of passers-by recorded with a camera, using only well received iterations to inform the next generation. In 2016, McCann Japan announced it had made its first non-human hire in the form of an AI creative director, which can take a client’s creative brief, consult its database of existing ads and provide basic direction for a human team to develop.

Across the myriad ways in which this technology is being used in advertising, the core idea is the same. AI enables brands to anticipate consumer’s needs and deliver more personalised experiences – the goal that advertising has been working towards since time immemorial, and especially since the rise of digital.

AI in Email

A channel as well-established as email might not seem like a hotbed of innovation, but it’s yet another space where you’ll find AI at work in almost every corner.

Maybe you’ve used Inbox, Google’s experimental cousin to Gmail, which uses AI to suggest automatic responses to emails that look like they’ve been written by a human rather than picked from a range of presets or categorises email better by interpreting message content.

A similar idea is being applied on the marketing side by companies like Conversica and Saleswhale, both of which offer a virtual sales assistants. These offerings can send AI-composed emails to potential customers, dealing with enquiries and warming up old or inactive leads with introductory mail before handing them off to human staff to seal the deal.

There are plenty of other solutions out there, covering practically every challenge facing email marketers today. Boomtrain’s personalisation platform uses AI to tailor marketing emails to each user, by pulling in content that is likely to interest them based on past behaviour and then sending each email out at the optimal time to increase its chance of being opened. MarianaIQ helps find new prospects by matching social, web and proprietary data to create in-depth personas that promise to go beyond a person’s job title. Persado and Phrasee help marketers find the perfect subject line, using machine learning to find the language that performs best and tailoring that to a brand’s existing tone of voice.

The ultimate goal of all this is simply; getting email marketing closer to a one-to-one interaction, treating each recipient as an individual - even in mass communication – without requiring a member of staff to handle conversations with each and every one of them. The technology is currently a long way from being able to handle the whole conversation without human input, but AI can certainly make the process easier.

Using all the information at your organisation’s disposal, to make real-time decisions about what and when to communicate is a key concept in modern marketing. This intelligent segmentation - categorising (and continuously re-categorising) types of consumers based on an evolving set of discrete data points started life in email software, and has since grown to be the backbone of many a marketing stack.

AI in Commerce

Commerce is another area where AI is already well-established, perhaps more than you might realise.

Product recommendations, the ones that we’re all familiar with from Amazon, make use of AI recommendation engines to pick the items most likely to get people spending, based on their purchasing habits and browsing behaviour. In Amazon’s case, these recommendations are powered by DSSTNE (Deep Scalable Sparse Tensor Network Engine, or ‘Destiny’ to its friends), an AI framework which went open source in 2016 - in the hope that other developers and researchers can help it become even more sophisticated.

As AI evolves, it is helping online merchants of all shapes and sizes better anticipate their customer’s needs, based on past behaviour and the context of their current search. This helps create a more personalised shopping experience, where a customer sees the products they’re most likely to want or need and retailers are able to more accurately forecast demand and focus marketing spend.

One brand making slightly different use of AI is The North Face, which is testing an AI-powered search interface for its range of jackets using IBM Watson. Instead of using the kind of search filters you might normally expect to see on a retail site, it simply asks the customer a series of questions – starting with where and when they’re planning to wear it, so it can check the weather, before moving on to more specific questions, such as the activities they intend to do while wearing it. These answers, which can be given in natural language, are used to narrow down the search results, offer products suitable for the customer, and of course, allow TNF to suggest complementary products based on the data they’ve gathered. As well as showing the right products, AI also offers a way of selling them at the price a customer is willing to pay. Historically, dynamic pricing optimisation has been used most commonly by airlines - but, from Amazon to Uber, an increasing number of companies are using algorithms to adjust pricing based on demand, availability and competitor pricing.

Before arriving at any of those shopping decisions, however, AI can first help streamline the initial product selection process, helping retailers know what to put on their shelves in the first place. One notable example of this is Lesara, a German online retailer, which uses AI to collate data on what’s trending on fashion blogs and social media, in combination with its own analytics and sales data, to help inform its buying decisions. This process helped it discover trainers with built-in LEDs before other retailers - which went on to become one of its biggest-selling products. And the principle doesn’t just apply to retailers. Netflix famously used similar AI to decide on its first original series. Based on six years’ worth of viewing data of its shows and films, an algorithm concluded that a political show starring Kevin Spacey and directed by David Fincher would be a surefire hit – which led to the creation of House of Cards .

Want more like this?

Want more like this?

Insight delivered to your inbox

Keep up to date with our free email. Hand picked whitepapers and posts from our blog, as well as exclusive videos and webinar invitations keep our Users one step ahead.

sign up

By clicking 'SIGN UP', you agree to our Terms of Use and Privacy Policy

SIGN UP

By clicking 'SIGN UP', you agree to our Terms of Use and Privacy Policy

Virtual Assistants

Perhaps the best example of AI at work is the one that people are most likely to be familiar with: virtual assistants.

These cloud-connected assistants interpret voice commands from the user and carry out basic tasks on their devices. For most of us, our first exposure to assistants would have been Apple’s Siri, but today the poster girl is surely Alexa, the software behind Amazon’s Echo smart speakers, which according to Morgan Stanley, had sold an estimated 11m units by the end of the November 2016.

As previously mentioned, there are plenty of assistants out there: Google’s Assistant, (which powers it’s Home speaker and is found in many Android powered devices), Microsoft’s Cortana (which is embedded in the Windows platform), Samsung’s Bixby (one of the big selling points of its latest handsets), and many more. Most likely due to its impressive market penetration, the majority of marketers have been focusing on Amazon Alexa - the technology that powers it’s Echo home speakers and is starting to find its way not only into a myriad of Amazon’s own devices and apps but, importantly, third party hardware as well. The language processing is even available for on-demand use in the Amazon Web Services Cloud.

As you’d expect from an Amazon creation, Alexa very much has commerce at its heart. One of the core tasks that the assistant can carry out is making purchases straight from Amazon, while brands like Uber and Domino’s Pizza have got in on the act by creating their own transactional ‘Skills’ (Skills are Alexa’s answer to apps). Assistants also provide opportunities for brand building and content delivery. Campbell’s has created an Alexa Skill that helps users find recipes, while tequila brand Patrón has a similar one for making cocktails, and Johnnie Walker’s Skill walks users through a personalised tasting session of its whiskies.

Of course, virtual assistants of varying application and specificity are popping up in more than just consumer devices. Adobe’s recently introduced Virtual Analyst, powered by its Sensei AI, analyses data to detect anomalies for users of its Marketing Cloud suite. Adobe’s offering joins IBM’s Watson and Salesforce’s Einstein in the marketing industry alone. The not-too-distant future will undoubtedly see AI embedded into more applications in the form of virtual assistants, as vendors strive to anthropomorphise technology to make it more accessible.