By Ian Monaghan
When it comes to transformational technologies, the irony that unites them all is how they're routinely hyped before proven, applied, or even understood among CMOs and marketers. By the time they change the world, people barely notice and are well onto shinier objects. Technological innovations morph and rebrand as they spread throughout society and the business sector. The information superhighway, grid computing, and personal digital assistants (PDAs) are just a few examples.
In a few decades, the ringing buzz of blockchain could be just another essential integration in the daily life of global business and industry processes — as commonplace as Bluetooth is today — rather than a mystical framework that few understand or can apply to their brand or organization.
Artificial intelligence (AI) is no different. It's been in the spotlight for several years, parading jaw-dropping outcomes of technology such as robots that can play the violin or voice assistants that can make a dinner reservation, while subfields like machine-learning algorithms, neural networks, natural language processing, and computational statistics do the heavy lifting behind the curtain.
These subfields are the building blocks that power AI functionality that have established an ecosystem in the last decade. According to Alok Kothari, director of data science at Adobe, growth in computing power and infrastructure has created production-ready AI frameworks across industries. These technologies used to require heavy do-it-yourself (DIY) components, but they've since made big data easily accessible for data science, in both research and production.
Technologies such as PyTorch, TensorFlow, Spark, and Hadoop are reducing the time from research to application dramatically. In the lightning-fast space of adtech, an increasingly agile approach to model generation and deployment is paramount for marketers. Advertising is an industry in which cutting-edge improvements can't wait two or three business quarters — let alone two or three years — to create efficient AI solutions.
Still, if people turn and see AI stamped on products and services that surround their daily lives, how do consumers know which intelligent services add or detract value — both personally and professionally?
One of the least understood use-cases of AI today is its place in digital advertising, even though it is one of the clearest. To truly demystify AI, it's important to examine the technology within four key categories and how each category fits into the advertising landscape. The intelligence below can help dispel the myths of what AI can and should do for marketers and clarify its value proposition now and for the future.
Some examples of AI in action are undeniably amazing. They hint at why some visionaries and technologists consider AI's invention in the same category as the discovery of fire.
Watson's win against two of Jeopardy!'s greatest champions in 2011 thrust AI into the mainstream.
Today, voice computing, self-driving cars, and the engines used by tech titans like Google affect the lives of millions and transform business models. Advertising can be found here as well. Uru, a startup acquired by Adobe last year, helps video publishers monetize content by enabling advertisers to buy ad "space" on the "surface" of the actual video footage and then virtually integrate logos, product placements, or promotional art.
With immersive mediums like virtual reality (VR) and augmented reality (AR) on the cusp of widespread adoption, brands will need to present inside these mediums in a seamless and harmonious way.
If revolutionary AI captures front-page headlines, efficient AI translates to gigantic contributions, in terms of productivity gains and scale. Google, for example, is leveraging its AI tools for its browser-based version of Inbox to answer people's emails. Grammarly, an editing crutch for many, powers its products' rules, patterns, and AI to improve the writing of millions. What's more, AI is streamlining work processes across the law profession, helping attorneys with due diligence and research, process automation, and legal insights.
Similarly, adtech is undergoing a massive change when it comes to efficiency and scale due to AI. From forecasting to performance optimization to ad rotation optimization, advertisers increasingly rely on intelligent machines to power massive search engine marketing and programmatic media operations. They work behind the scenes to help create personal experiences, optimize delivery in real time, and deeply analyze ad performance. Attribution AI can help marketers determine the impact by earned, paid, and owned media and how to spend their precious dollars across different channels to drive better results.
The next category of AI is supportive, which helps marketers make decisions but isn't necessarily going to raise the bar in terms of mass productivity gains or performance across an industry. Sephora customers, for example, can personalize their shopping experience by trying on makeup virtually using AR and matching their skin tone to a foundation with AI. Similar to Sephora, prescription glasses and sunglasses retailer Warby Parker's Virtual Try-On enables customers to try on virtual frames through AR with the help of Apple's Face ID.
Warby Parker also uses AI algorithms to personalize communications to make it easier for customers to find what they want. Meanwhile, Spotify's "Discover Weekly" playlist reaches tens of millions of listeners who return each week to check out their customized song lists. The data-driven music service uses machine learning to continually improve this feature.
Key advertising examples of supportive AI include analytics functions that automatically recognize unusual activity and determine whether the event is worthy of an alert. This is known as "anomaly detection" and is essentially a statistical method that monitors how a metric has changed in relation to previous data.
Advertisers use the algorithm to separate "true signals" from "noise" to identify which fluctuations matter and which don't, and then identify the root cause of a true anomaly, such as drops in landing page views, spikes or drops in trial registrations, and drastic drops in average order value. The algorithm can also consider the seasonality of data and factor in holidays. Ultimately, it helps companies to create reliable forecasts for such key performance indicators.
Machine learning for look-alike modeling is yet another example of Supportive AI. Advertisers aiming to extend their reach, by landing more customers who look like their best customers, use it to help build larger audiences stemming from smaller segments. The larger audience reflects the benchmark characteristics of the original audience.
Last, but certainly least, is hype AI. Where to start? Does the world need an electronic toothbrush with AI abilities? How many chatbots that are described as virtual assistants are essentially powered by humans?
Smart-home marketers sling AI to describe how their wares understand people and their needs. This may all materialize one day. But the jury is still out. The good news is that in advertising, AI hype can't last long since evidence of it working or not becomes quite apparent in an era of heightened accountability and when million-dollar media investments are at stake.
By separating fact from fiction for AI and analyzing applications for advertising across the categories identified above, marketers can better understand the role and value of the technology.
One day, software used by creators to design assets could potentially show that images with a certain color scheme have generated better results. Or campaign optimization tools could tell brand managers and marketers that a video with a certain aspect ratio is performing better in an in-flight campaign. This will show what videos recropped to the optimal would look like and ask marketers to accept or edit the drafts and substitute them into the campaign going forward.
As 2020 approaches, marketers should boost their expectations and imagine a world with heightened intelligence for advertisers across all relevant categories. Advertisers should hold tech companies accountable for creating AI programs for efficiency and support and lean on data scientists to help separate the reality from the hype.
Ian Monaghan is a product marketing manager of data and integrations at Adobe Advertising Cloud, a partner in the ANA Thought Leadership Program.