1、Important disclosures appear at the back of this report GP Bullhound LLP is authorised and regulated by the Financial Conduct Authority GP Bullhound Inc is a member of FINRA Subscribe to receive GP Bullhound Research and News on 2019 THE IMPACT OF AI AND DATA ON DIGITAL MARKETING AI AND THE SERVICES

2、 REVOLUTION CONTENTS 04 THE VIEW FROM GP BULLHOUND 06 PREFACE Setting The Stage: Key Definitions 08 CHAPTER I Artificial Intelligence Today 13 EXPERT VIEW Gurman Hundal, MiQ 14 CHAPTER II Artificial Intelligence In The Universe Of Digital Marketing 16 EXPERT VIEW Pete Kim, MightyHive Sir Martin Sorr

3、ell, S4 Capital 18 SECTION 1 Hyperpersonalization 26 EXPERT VIEW Raj Balasundaram, Emarsys 30 SECTION 2 Branding 35 EXPERT VIEW Nii Ahene, Tinuiti 39 EXPERT VIEW Nithya Thadani, Rain 40 SECTION 3 B2B / Sales when, where and how to reach them; what to tell them; and how to most efficiently allocate m

4、arketing budgets. In addition, there is already a bewildering labyrinth of companies offering AI marketing technologies, between which marketing teams will have difficulty differentiating, choosing and implementing. ROLE OF DIGITAL AGENCIES AND SERVICES PROVIDERS This is where digital marketing serv

5、ices providers including agencies and consultancies can play a critical role in the efficient use of AI. They can act as a trusted advisor who interprets the data and applies it to their clients marketing strategies. According to Pete Kim, CEO of S4 Capitals MightyHive, “future 2.0 marketing service

6、s firms will be subject-matter experts in advertising and data analysis, Machine Learning and AI.” However, all of this will require fundamental organizational and cultural shifts - both services providers and their clients will need to reorient their structures around a seamless flow of data betwee

7、n analysts, strategists and creatives with the ability to make tactical shifts in real time. In fact, one model that is emerging is AIaaS (AI as a Service). In this case, AIaaS not only refers to the ability to pay for an otherwise costly AI technology platform with recurring licensing fees, but als

8、o implies that a services provider will still add the necessary human element. Adam Hildreth, CEO and Founder of Crisp put it this way: “AI can do 99% of the heavy lifting to identify the needles in a haystack. But you still need a human to verify that it is indeed the needle you are looking for and

9、 the action thats required.” CATEGORIES OF AI IN MARKETING We have explored three themes on the topic of marketing AI: 1. Hyperpersonalization: the ability to target individual consumers with highly tailored content at the optimal time, in the optimal place and using the optimal medium but with dyna

10、mic, constantly changing variables. This is only possible at scale with AI. AI is supercharging established targeted marketing practices such as programmatic media buying, email, and search while adding new possibilities such as voice and visual search all with the promise of truly optimized omnicha

11、nnel marketing. 2. Branding: in addition to one-to-one marketing, AI will help marketers build stronger brands that stand out in a cluttered media landscape. With AI-enabled segmentation, marketers can better understand their audience and create new insights into their consumers using digital signal

12、s of demographic, psychographic and behavioral variables. In addition, new AI-enabled technologies such as Natural Language Processing help identify the best social media influencers, often with small but devoted followings, and enable brands to reach new consumers. 3. B2B (business-to-business): AI

13、 promises to augment and reshape B2B marketing, which is typically more sales force-driven with a more complex customer journey. AI is revolutionizing the B2B sales process, which can often take several months and involve numerous steps and parties, by uncovering patterns in customer behavior and pr

14、edicting best next steps, as well as discovering high value overlooked customers. During and after the sales process, AI-driven bots are able to respond efficiently to customer inquiries, freeing up human customer service personnel for more critical or nuanced tasks. Bots are likewise used by many c

15、onsumer marketers. We believe that we are at the beginning of a new marketing era driven by the need to connect vast amounts of disparate data, which can only be accomplished with AI. Gurman Hundal, Co-founder and CEO of MiQ, views the development of AI for marketing in two stages: first, the automa

16、tion of processes and data analysis, and second, rationalization by using AI to predict behaviors and recommend actions, which is where AI gets really interesting. Brands will benefit by more efficiently interacting with existing customers and uncovering new customers as AI opens up new marketing ch

17、annels such as voice and visual search and enables seamless omnichannel communications. Digital services providers including agencies and consultancies will benefit by providing critical advice in interpreting this data and executing these AI-driven strategies. No longer is half of the advertising b

18、udget being wasted. The View FROM GP BULLHOUND THE VIEW FROM GP BULLHOUND SIMON NICHOLLS PARTNER GREG SMITH PARTNER OLIVER SCHWEITZER EXECUTIVE DIRECTOR “ARTIFICIAL INTELLIGENCE.IS ALREADY SHAPING OUR EVERYDAY LIVES AND TRANSFORMING THE WORLD OF MARKETING.” Gurman Hundal, Co-Founder the trouble is I

19、 dont know which half” will overcome these obstacles, as organizations adapt to the needs, risks and opportunities of the digital economy. 54 Setting The Stage PREFACEPREFACE KEY DEFINITIONS ARTIFICIAL INTELLIGENCE (AI) Artificial Intelligence is a sector of computer science which works towards a hy

20、pothetical (as of today) state where machines can perfectly replicate intelligent behavior typical for humans. Full AI has not been achieved yet. Machine Learning and Deep Learning are subsectors and enablers of AI. NEURAL NETWORKS NATURAL LANGUAGE PROCESSING (NLP) AND GENERATION (NLG) In the most b

21、asic sense, NLP and NLG refer to machines understanding and speaking a language the same way a human would. Both NLP and NLG have only become possible with advancements in Deep Learning as neither conventional programming nor pioneering Machine Learning models could capture the complexity of human l

22、anguages (context, emotional connotations, errors etc.). MACHINE LEARNING (ML) Machine Learning is a sector within Artificial Intelligence. It describes an ability of a computer or software to adjust its own algorithms based on learnings from the data that has been processed with little to no human

23、intervention. Machine Learning algorithms learn by example rather than by exclusively following pre-defined rules. DEEP LEARNING COMPUTER VISION In nature, neural networks are systems of neurons in a brain connected to perform a certain function based on specific information they receive. An artific

24、ial neural network, consisting of algorithms, loosely serves the same purpose recognizing patterns in the data provided. Deep Learning describes Machine Learning using deep neural networks, or neural networks with more than one layer (feature, aspect) of data. Deep Learning models are by nature more

25、 sophisticated than a one-dimensional machine- learning model, but it also means they require more computational power and longer times to train them. Computer vision is quite like NLP and NLG the premise is to enable computers to see and interpret static and dynamic images similarly to how a human

26、eye would do it. The problem of deciphering images rather than words might be even more complicated for machines to master, considering the innumerable variations in the natural world around us. 76 CHAPTER I Sources: “The AI Index 2018 Annual Report”, AI Index Steering Committee, Human-Centered AI I

27、nitiative, Stanford University, Dec-2018 Artificial Intelligence Today ARTIFICIAL INTELLIGENCE TODAY ON THE VERGE OF A REVOLUTION 113% GROWTH IN THE NUMBER OF ACTIVE AI STARTUPS IN THE US BETWEEN 2015 AND 2018 9x ANNUAL GROWTH IN NUMBER OF ACADEMIC PAPERS ON AI PUBLISHED IN THE US IN 2017 COMPARED T

28、O 1996 350% GROWTH IN THE VC FUNDING FOR AI STARTUPS IN THE US BETWEEN 2013 AND 2017 34x GROWTH IN THE NUMBER OF JOB OPENINGS REQUIRING SKILLS IN DEEP LEARNING BETWEEN 2015 AND 2017 IN THE US 98 AI: A Retrospective FROM SCIENCE FICTION TO EVERYDAY REALITY Sources: based on “The History of Artificial

29、 Intelligence”, Harvard University, Blog of the Graduate School of Arts and Sciences on Artificial Intelligence, 2017 ARTIFICIAL INTELLIGENCE TODAYARTIFICIAL INTELLIGENCE TODAY Moores Law MULTIPLIED BY THE DATA Sources: Ray Kurzweil; Domo, “Data never sleeps 6.0”, 2018. I. COMPUTE /SOLVE II. REMEMBE

30、R /STORE III. ANALYZE /LEARN AI AMERICANS USE 3.1 MILLION GB OF INTERNET DATA GOOGLE CONDUCTS 3.9 MILLION SEARCHES AMAZON SHIPS 1.1 THOUSAND PACKAGES YOUTUBE USERS WATCH 4.3 MILLION VIDEOS INSTAGRAM USERS POST 49.4 THOUSAND PHOTOS Once a dream of the brave, Artificial Intelligence has come a long wa

31、y since the beginning of the 20th century. Through an unprecedented advancement of computer science, smart machines have become a staple in our everyday life feeding off the data we generate. I. The concept of programmable computers as in “machines to compute” existed for over a century with various

32、 machines successfully solving an array of simple tasks for humans, before the term “Artificial Intelligence” was popularized during the Golden Age of Science Fiction between 1938-1946. ACHIEVING ARTIFICIAL INTELLIGENCE II. Until 1949, one major problem of all computers was that not only were they e

33、xtremely expensive but also they could not store commands, only execute them. In the 50s-70s, however, computers became exponentially faster and smarter. So much so, in fact, that at one point AI scientists predicted machine intelligence on a level of a human being by the end of the 70s. The 80s wer

34、e marked by an overall disillusionment. Contemporary computers could not break out of a computational dilemma: they either could not store enough data or could not process it fast enough. III. Turns out, the secret was to wait. With both computational power and memory capacity of computers doubling

35、every two years, the problems from the 80s gradually disappeared. In the 21st century, humanity officially entered an era of smart machines. Artificial Intelligence thrives where there is enough data to process and enough power to process it. With computational power doubling approximately every two

36、 years and amounts of data generated by humanity growing exponentially, Artificial Intelligence might be entering its best years yet. In 1965, one of the founders of Intel, Gordon E. Moore, predicted that the number of transistors per microchip would double about every year. While he then revised th

37、is to every two years in 1975, he has been approximately right computing power has increased exponentially, and we are now all walking around with supercomputers in our phones. Today, the latest chips from Intel can run over 10 trillion calculations per second. This sheer power, coupled with other i

38、mprovements in microchip functionality and the possibility to reprogram chips on the fly, has opened previously unimaginable opportunities for computers to learn and recognize patterns in data that are fed through them. COMPUTER RANKING PER CALCULATIONS PER SECOND PER CONSTANT DOLLAR AT THE SAME TIM

39、E, EVERY MINUTE OF THE DAY 1E-09 1E-07 1E-05 0.001 0.1 10 1,000 1E+05 COMPUTING POWER (MILLIONS OPERATIONS PER SECOND) 1900 1910 1950 1940 1930 1920 1960 1970 2010 2000 1990 1980 2020 1E+09 1E+07 ANALYTICAL ENGINE HOLLERITH TABULATORIBM TABULATOR COLOSSUS ENIAC WHIRLWIND DEC POP-1 IBM 360 DATA GENER

40、AL NOVA IBM PC APPLE MACINTOSH POWER MAC IBM ASCH WHITE IBM BLUE GENE GTX 450 NVIDIATITAN X MECHANICALRELAYVACUUM TUBE TRANSISTORINTEGRATED CIRCUIT 1905 1915 1955 1945 1935 1925 1965 1975 2015 2005 1995 1985 2025 1110 Make Hay While The Sun Shines MARKETERS ARE RUSHING TO CAPITALIZE ON AI OPPORTUNIT

41、Y Source: Salesforce, “State of Marketing”, 2018 EXPERT VIEW: MiQARTIFICIAL INTELLIGENCE TODAY Artificial Intelligence can no longer be considered a technology of the future it is already shaping our everyday lives and transforming the world of marketing. According to Gartner, 30% of companies world

42、wide will use AI in at least one of their sales processes by 2020. And AI is increasingly powering customer experiences from real-time personalization to conversational voice interactions. But what do we mean when we talk about AI in marketing? At MiQ, we view AI as two concepts: automation and rati

43、onalization. Automation replaces specific human jobs and is the solution to the challenge of Big Data, which produces an immense amount of information. The automation of processes and interpretation of this data is right now the most effective use of AI in marketing. But then there is rationalizatio

44、n, which is where AI starts to get interesting. Rationalization is the use of AI to predict behaviors and recommend actions and it is where AI really excels. But before putting AI in the drivers seat in the decision-making process, brands need to fully understand the technology they are using. After

45、 all, algorithms are only as good as the code that governs them, and the data used to teach them. As such, the rationalization phase is currently somewhat overplayed. If a brand wants to trial AI, they should first employ an expert. Its not like picking up a product off the shelf. They need to inves

46、t in people to customize it to the needs of their business. Thats why many of the firms using AI effectively are, at heart, technology companies with experts on hand. Agencies and other businesses looking to follow in their footsteps will need to bring in technology experts before they put AI to wor

47、k for their marketing teams. In other words, before they integrate AI and data analysis, they need to integrate AI and data analysts. This could mean a change of leadership or acquiring a specialist company but those are the kind of significant changes brands need to make to prepare for the digital

48、era. One brand that has taken the leap is McDonalds. Earlier this year, McDonalds acquired AI start-up Apprente to implement AI voice-based technology in its US drive- throughs. It follows on from an earlier decision to invest in technology that can automatically alter individual drive-through menu

49、panels depending on factors such as the weather. For programmatic advertising, AI is now used for activation and insights. In fact, we should redefine programmatic what used to be about buying audience more effectively has now expanded to leveraging any online or offline data set in an automated and real time manner. Another area of marketing that AI is redefining is customer segme