Schedule a Demo
Does My Company Subscribe?
Josh SuttonGlobal Head, Artificial Intelligence PracticePublicis.Sapient
The hype around artificial intelligence (AI) is ramping up, especially as big tech companies like Amazon, Google, Facebook, IBM and Microsoft attempt to commercialize its use. Agencies are also starting to figure out how they can leverage AI to make their clients’ marketing and advertising efforts more effective. eMarketer’s Bryan Yeager spoke with Josh Sutton, global head of the artificial intelligence practice at Publicis.Sapient, to demystify how AI is used for marketing and what its future may hold.
eMarketer: At a high level, what role does the artificial intelligence practice play within Publicis.Sapient?
Josh Sutton: We’ve been looking at AI for the better part of four years now with the real focus of getting an understanding as to what it really is that people mean when they say “AI.”
Within the practice, we primarily do two things. One is that we help companies look across the landscape of all the different big data, machine learning and other artificial intelligence platforms that are available to understand which ones can help them best achieve their business objectives. We’re really being the engineers that help people figure out how to put together the best engine to meet their specific objectives.
The other focus area is looking at how we can apply artificial intelligence to the advertising space. Being part of Publicis Groupe, we’re obviously looking at what we can do to transform our own industry. We want to leverage various AI platforms to push the boundaries on what can be both to improve things that are done today, as well as to innovate and create new models that hadn’t previously been available.
eMarketer: You said that AI is widely used but widely misunderstood. Give us a sense of how you view the AI landscape as it stands today.
Sutton: We look at the whole host of companies that are out there that are claiming some degree of artificial intelligence and group it into three primary categories.
The first isn’t AI in the traditional sense. It’s more in the big data category: the platforms that enable aggregation of large datasets with which one can apply machine learning. These range from the bespoke systems people have built in-house, to the Palantirs of the world. It’s an area where there’s been a lot of focus and a lot of noise, but quite frankly, probably not the business results that some people expected over the past few years because there haven’t been the tools that extract real value out of that set of information yet.
That leads to that second tier of tools, which are the ones that you’re hearing all about today: machine learning platforms. Machine learning platforms are, at the end of the day, a very straightforward and simple concept. They take large amounts of data, apply algorithms to them to identify trends or common occurrences and effectively predict or identify insights or common responses to areas so one can better learn the likelihood of that occurring again.
Within that space, there are a few of the leading players, and I think the market is driving towards more of a consolidation. We view IBM, Microsoft and Google as the big three that are openly commercializing their platform to varying degrees. IBM has probably been the most aggressive with Watson, with Google dancing a fine line between open-sourcing some of their platforms and keeping some in-house. Microsoft is right there in the middle. Amazon and Facebook are also both making very substantial investments, but doing it in more of an in-house black box model [vs. the other three players].
The third tier of platforms are what I call human interface tools. Within these include causal AI tools—platforms that understand the world the same way that a human does. The main players in the market are the platforms that came out of MIT [Open Mind Common Sense] and Stanford [Cyc]. At this stage, they’re not creating insights that are particularly compelling above and beyond what a person would.
But what they are doing is communicating and acting and thinking the same way a person would, which makes them very powerful from an interface point of view and for understanding things like social media. I would group natural language processing and generation tools here, as well.
This suite of applications is in its infancy right now, but in the coming years, it will provide a more direct retail user interface that takes AI from being a cumbersome business tool to a seamless tool that’s used by audiences across the world.
eMarketer: How are these different AI technologies being applied in marketing and advertising today?
Sutton: One of the largest sets of data that we’ve started to pull together is aggregating all of the behaviors that can be purchased at the individual level—things like geopositioning of where everybody is throughout the day, what were people purchasing from point-of-purchase providers or what were people looking at via search engines.
You can pile that together and get a very good picture of behavior of people across the world, or country by country. Within a given country, you can tell to a pretty high degree of specificity how the population is behaving, and how they’re reacting to certain stimulus.
We initially built it as a tool to look at how can we start to draw insights from data. How can we let our data scientists generate insights for our customers to better understand their consumers and what matters to them? What do they actually want based on behavior?
Figuring out how to get the right message to the right person on the right platform at the right time is not a simple task, but what we can do is start to look at what works and what doesn’t. That’s where machine learning really starts to shine. It’s taking the legacy of A/B testing to a much grander scale than it could ever be envisioned by people. You can conduct true systemic testing of what’s working, what’s not and what’s really driving behavior to enable you to adapt in real time to maximize your results as you put your message out to consumers.
There’s always been the adage that I know half my marketing is working, I just don’t know which half. What we’re able to start to do is really identify which half is driving results by looking at the real-time interactions between the messages that we’re putting forward and the behaviors that we’re observing to draw real insights about what works and what doesn’t at a scale that hasn’t been done before.
What we’re able to do in terms of targeting, identification and generating results is dramatically better than anything that we could’ve done even a year or two ago.
eMarketer: What might the future hold for the intersection of marketing, media and artificial intelligence?
Sutton: When you layer on that human level of reasoning at scale, you can do things with a level of efficiency that traditionally had required a large human staff, like media planning and buying. You can start to use a combination of your data lakes and your machine-learning capabilities overlaid with a causal AI platform that thinks and mirrors the way that people would think to start to perform the basic real functions and accelerate the activities that would be done by a planner.
Separately, we can start to look at predictive analytics in a fundamentally different way. We can overlay both machine learning and causal, semantic AI to predict who’s looking at making large premeditated purchases: cars, engagement rings, mortgages and more. Once that is determined, we can figure out what factors matter to that person and customize the message that’s delivered.
That can be done for a wide range of activities. There’s the obvious ones around targeting your media spend and how you allocate dollars. There’s also the much softer side, but I think potentially an even more important side: How do you create a great, personal experience for people on digital channels?
One of the things that we’re investing in right now is building an artificial intelligence interface between a number of content management platforms and the actual user interface—the actual screen. If a person is on a content-rich website, how do I look at what I know about that person and based on what I know about them, look through the 10,000 or 20,000 different pieces of content that I could show them to put forward the ones that are going to be relevant and meaningful to them?
One of the common challenges we’ve heard with a lot of people is, “We treat everybody like they’re a new customer, where 99% of the people that come to our site are actually people we know a lot about.” How do we take that knowledge of them and translate that into an experience that feels very bespoke and personal to them? That requires AI.
eMarketer: You mentioned the ability to further automate and scale media buying functions currently performed by humans. What do you think the potential impact of AI could be on the workforce in the marketing and advertising industries?
Sutton: I think you’re going to see a material shift in how people do their day-to-day jobs. Certain things today that are performed that are either rote functions or require more straightforward analytics of information are going to be accelerated by technology. That’s not to say that people will be replaced by technology, because the combination of AI plus human is a much more powerful platform than AI alone.
Nonetheless, I think you’re going to see [that] people are able to perform much more than they historically have been able to as individuals through leveraging technology. Productivity is going to go up dramatically.
People are going to be able to focus on higher-value activities. Another change is probably some overall decrease in market size. And a third thing is an increase in the demand for technologists within the marketing world.
You've never experienced research like this.
Nearly all Fortune 500 companies rely on us.
Inquire about corporate subscriptions today.