Welcome to the new year, and to January 2017. To get this month started, we’ll take a step back to do a fly-over of Artificial Intelligence in the marketing field. The goal – to get a bead on the overall maturity of the market, and take a look at some of the best products and services using AI. In short, if you are a marketer, we’ll catch you up and give you a foundation for AI in marketing. Of course, in our follow-on posts and podcasts this year we’ll be talking directly to the companies using and developing new products driven by AI. The goal – keep you – the strategic marketers – up to date with what’s new and working when it comes to AI and marketing.
So let’s take a look at the marketplace and see what’s going on.
eMARKETER
First, from eMarketer – you’ve heard of these guys – market research and analysis on all things digital marketing. They’ve put together a pretty killer webinar and white paper they call “Artificial Intelligence for Marketers: The Future is Already Here”
I guess we know what they’re headed with this one, eh? The Future is Already Here.
In any case, this one’s available free on the web. You can find the details on the report at eMarketer here.
In the meantime let’s hit a couple high points from the deck:
Point #1 – AI for marketing is still in its infancy.
It’s coming faster than many folks imagined though – that’s eMarketer’s point.
I’d say they are probably right if you consider all of the horizontal platforms that are deployed – think IBM Watson, Facebook’s FBLearner Flow Platform, Amazon’s (AI) Artificial Intelligence Services on AWS, Google’s DeepMind division and their now public and open source – DeepMind Lab along with their Google Cloud Machine Learning Platform, and of course the folks at OpenAI – the joint effort of Elon Musk, Peter Thiel and Sam Altman (startup accelerator Y Combinator founder) and their OpenAI Gym.
The IBM Watson platform is leading the way more broadly, with services for four areas: language, speech, vision and data. They claim the services can be used off the shelf by clients or developed into custom applications.
The Wall Street Journal reported last year (April 2016) that Watson had 550 organizations from 17 industries partnering with Watson. And in October 2016, IBM says they will reach 1 billion users on Watson by 2018. We’ll get the Watson guys on an upcoming podcast and hear from directly. They are one to watch given that kind of momentum.
Equals 3 – that’s the name of the company – Equals and the number 3, is rolling out a marketing focused platform. It’s amibitious – they call their product “a cognitive companion to the marketing professional.” Their software is named Lucy uses 10 of Watson’s computing services to answer marketing questions in a user friendly way. It’s got preloaded data from a variety of sources, think – Advertising Research Foundation, American Marketing Association, Twitter, Facebook, Kantar, Nielsen.
What’s interesting with this product is that they are using AI for the interface to the marketer – that is, marketers can converse with the Lucy via the natural language interface, and they are using AI to drive marketing tasks for the user, like segmentation, develop persona portraits, suggest messaging by segment, etc. It looks to me to be a massive undertaking but if they can do it, they’d have advanced the marketer’s ability to easily use the marketing tools, while also providing state of the art marketing programs. Pretty clever stuff.
In any case, the core ingredients for accelerated growth are falling into place.
Point #2 – Marketers in all industries are experimenting with key application areas including BI (business intelligence), customer acquisition, programmatic advertising, campaign optimization, and multi-channel optimization.
Ok, granted that pretty much covers most of the areas in marketing, but there really is some good stuff happening. To be fair, it’s not necessarily creating tremendous value just yet, but we’re out of the academic environment and squarely into practical use.
We’re going to touch on all of these areas over the coming weeks in follow-on posts here and podcasts – via interviews with a wide range of AI marketing solution companies.
For now, let’s take a look at one of these – AI for programmatic advertising. As a marketer you’re all too aware of how fast programmatic is taking over the advertising process associated with paid search, social, display and video advertising. Between the native platforms at Facebook, Google, Bing, AOL, and even Amazon, Walmart and others, and the horizontal solutions that manage the full funnel advertising efforts. Interestingly, though it’s embedded and therefore a whole lot less obvious, AI and especially Machine Learning and Deep Learning have played a crucial role in driving efficiency and effectiveness of real-time bidding engines used in programmatic advertising. To give you some sense of the size of this market, Jupiter Research estimates RTB-driven advertising will be worth $42 billion by 2021, while in 2016 it was already $3.5 billion.
What’s the big leap provided by Machine Learning for the real time bidding process?
According to Jupiter Senior Analyst Steffen Sorrell – adding machine learning to RTB allows the automated ads to target individuals, not just a demographic or segment.
In my world, that’s a crucial next step in making advertising work better. It creates a narrower, more targeted and personalized experience, and you do less shotgunning of a segment or demographic and more precise targeting of the best prospects.
eMarketer references RocketFuel – you’ve heard of these guys – a pioneer in programmatic advertising. They are using Deep Learning to better match ads and audiences. So let’s pull up for a second.
What is Deep Learning?
The short definition – it’s an extension of Machine Learning that models an increasingly more abstracted representation of a data via a series of neural net layers. A good example is in the machine vision world where the lowest level of data is the actual digital image – the millions of pixels that make up one image of a number printed on a page. Subsequent layers of the neural network abstract up from the pixels to, for example, edges and segments, and further up in the next layer to shapes, and eventually up to numbers. In this case, the deep learning network is “reading” numbers. The key here – similar to a human – we can move up and down the layers of abstraction to think about the data and the content. It’s a different application and involves some other really incredible AI concepts, but the Google Deep Mind platform – the guys who beat the world champ Go player, use this Deep Learning idea to abstract up from the current board configuration to a more strategic understanding of the game and reason about their strategy. In turn they drop back down to the board level when they choose a next move.
In any case, Rocketfuel uses a concept they call “moment scoring” by evaluating more than 2 trillion data points, making 5 billion ad requests a day. Big Data for sure, and it optimizes across 1st and 3rd party data and executes campaigns across display, mobile, video, social and email. Rocketfuel is pushing the technology to figure out how to abstract up even further to understand “the psychology of influence”. According to Rocketfuel’s Mark Torrance, CTO –
“… we can figure out what pushes people’s buttons – whether it’s time-pressure, aspirational or advertising with scenes of the outdoors.” One could imagine this concept cutting multiple ways – feedback loops to drive different creative that better matches a product audience, and alternately, learning what kind of ads a specific person responds to – across different products. Either way, you can see the abstraction here – moving up from the pure ad and response level to intelligent insights on the ads. Pretty interesting.
We’re not touching on a couple key categories – and I want to acknowledge that. With a limited podcast length, we’ll cue up some of the other important application areas for future podcasts. So stay tuned. In particular, we’ll tackle Recommender systems and Bots or Virtual Digital Assistants. Those two topics merit their own dedicated broadcasts.
Point #3 – The key product players include big tech companies – Google, Facebook, IBM, Amazon, etc. along with advertising agencies, and a whole host of marketing focused startups.
We touched on the big players a bit earlier – so let’s talk about some of the AI startups.
This is by no means a comprehensive list of the players, but it will provide a sense of the nature and breadth of the AI-driven solutions available to marketers.
First up, a company called Thoughtly, and their product Ellipse. It’s a natural language processing startup, and it’s good at processing large (really large) amounts of text from thousands of sources in real time and then create a visual topography and summary of the content. From that roll up the application, for example helped BBC Worldwide, identify over-represented and under-represented themes in worldwide content, and in turn map specific content with the appropriate audiences. There are a bunch of these NLP-centered companies who are really good at understanding a large collection of data – truly understanding it, not just structuring it for fast keyword search. The example that gets used frequently in this space- think about the topic of Jaguars. Obviously, that term can refer to the family of big cats or to the car company. A good NLP solution understands that similar terms, respectively would be panthers, and Porsche. Panthers are a form of big cat like the jaguar, and Porsche is a luxury performance car like the Jaguar. Searching using panther would bring up articles or content referencing the big cat – Jaguar therefore.
Narrative Science, and their produce Quill is one of the players in the natural language generation side of things. These companies are auto-generating content that looks like it is written by human authors – the sentence structure and grammar is right, but more importantly it is able to extend beyond the inherent keywords and phrases to create higher order summaries, and extensions of the underlying content in the way a human would. There are companies doing this in the financial space – collecting the pertinent financial data about a company and generating required legal financial documents. It’s happening in the B-to-B marketing space too – DemandBase, a B-to-B marketing platform, incorporated an AI-driven solution that surfs the web and related content outlets for information about companies, products and people, and uses the pertinent elements to enrich it’s prospecting lists for DemandBase b-to-b marketing customers. In short, it uses NLP to collect and understand what’s out there like an intern, sales ops or business analyst might, and then synthesizes the important stuff right into the prospect database. Pretty clever stuff, considering it’s doing this at scale, faster than a human, and updates more frequently.
In any case, the folks at Narrative Science have a horizontal platform for this type of application and cite examples a USAA (insurance), Deloitte, Credit Suisse, and Dominion Dealer – a software solution for car dealers. They claim to have 70 customers principally in financial services and consulting, along with insurance, government and e-commerce.
A totally different orientation to AI and natural language understanding, is an Israeli company called Beyond Verbal who use AI technology to understand consumer’s emotions and character based on spoken voice. A natural application here is the Customer Call Center context, where they listen in to agent and customer calls and track the two parties’ ‘state of mind’. They call themselves an “emotions analytics” company, and explain their technology as “analyzing emotions from vocal intonations”. The gist of it is that they are aiming to understand people’s moods, attitudes and emotional characteristics – what we think of as personality from their raw vocal intonations in real time as they speak. If you think about it, it would be incredibly helpful to have a steady flow of your customer’s mood as they interact with your product – online, in a retail store, at a kiosk, via a call center, virtually anywhere. As a marketing consultant, I think about how many times I’ve said to my clients – you’re not just informing them, you are touching them – looking for that emotional connection that signifies a deeper relationship with your brand. Theoretically these guys could help us know whether we are achieving that deeper level of affinity and connection with our clients.
OK, so let’s finish off with an even higher level overview of the AI market for Marketers.
To get started, let’s take a look at the big players and especially the most active corporate investors. The folks at CB Insights share a ranked list of investments by the big guys in AI generally (that is, not uniquely in marketing applications). Topping that list are the venture funding arms of a number of big tech players including Intel, at the #1 spot, Google Ventures, Ge Ventures, Samsung Ventures, Bloomberg Beta, and so on down the list. Check out our most recent blog post at DeepMarketer.AI for the complete list and selected investments by each major player.
Next up, CB Insights looks at it from the opposite perspective – who are the most well-funded AI startups (again across application areas, and in many cases here – for horizontal platform development using AI that cut across many application and industry sectors) or putting it differently, these are the folks that are developing the core technology or big bet players. And given that these guys are getting the largest investment – the horizontal platform players, and receiving much more funding than individual application solutions. In short, the race is on to win major market share for being the platform of choice for AI solutions. It’s analogous to the historical race around search, the browser, OS’s for the various areas – laptop, phone, cars, VR/AR, etc. You’ll want to keep an eye on these guys therefore – not only because one or a few will likely emerge as the next major platform, but also because as that race gets sorted out we’ll no doubt see an acceleration in the adoption of AI solutions around the selected platforms- with built in ecosystems, standardized development platforms driving faster, more efficient time to market.
In any case topping the list of highest funding for AI startups include Sentient Technologies at $144M investment to date, followed by Ayasdi, Vicarious Systems, Context Relevant, Cortica, Workfusion, RapidMiner, Digital Reasoning Systems, H20.ai, and Viv Labs.
Here's CB Insights AI100 - the top 100 companies redefining their industries:
