Artificial intelligence is infiltrating our everyday lives, whether we realize it or not. It's powering smart systems and solving business problems -- think about using machine learning and big data to create new services and enhance existing ones. In some domains, AI systems are already more skilled than humans, and in the coming years AI will recast strategies across many industries, from healthcare (AI systems are helping detect cancer now) to harnessing the power of crowds to solve big problems (like gamers helping with genome research). And in most cases, it will all seem dead simple. The complexity's on the back end. So how can typical companies tap into the power of AI now? There are three areas I think bear watching.
E-commerce and product recommendations: Amazon incorporates collaborative filtering -- the first generation of online recommendations. You're undoubtedly familiar with these "if you bought product A, you will probably like product B" pitches. However, this method has limitations, the biggest being that collaborative filtering works well only for products driven by taste and bought in large quantities; if you purchase lots of B.B. King tracks and Elmore Leonard novels, Amazon probably does pretty well at suggesting new options. However, higher-priced items or goods that aren't bought in large quantities (and thus generate little interaction data) aren't conducive to being purchased by recommendations via collaborative filtering.
AI-powered second-generation recommendations expand this reach to new product areas, including ones that present the user with constraints and needs while setting the foundation for a knowledge-based structure. For example, someone could say he wants a tablet that runs Android, offers 4G LTE and costs less than $800. This also sets the stage for long-tail recommendations: Now you know a customer is an Android user and values always-on, fast connectivity.
How to get an advantage: Retailers need to get better at processing product databases and extracting data from internal and external sources to gain superior knowledge about products. Example classification and extractions, along with image and pattern recognition, help you build better insights. Continually work to pull in new information sources. I'm seeing AI used to gather information from multiple data streams across e-commerce sites, in order to recommend truly relevant products to consumers. In addition, no longer must recommendation systems rely solely on statistical correlations between products consumers are buying; they're delving into deeper information sources to understand why, when, where, how -- and most importantly, what -- customers are buying. Processing data in new ways means taking into account a deep understanding of such things as gender, time of purchase, payment preferences, amount of time spent on a certain page, click trails -- all that data is needed for AI systems to help the business understand consumers' preferences and behaviors.
Vertical search: Vertical, or topic-specific, search is gaining popularity in areas from travel and electronics to books, films and more; it's where a person goes directly to a site such as IMDB or Kayak rather than through a general engine like Google or Bing. Vertical search uses semantics, text classification, feature extraction and advanced big data analytics mixed with other AI algorithms. Google is waking up to the threat and lately has put a lot of effort behind vertical search in an effort to deliver more relevant, tailored results. However, vertical search presents an opportunity to get an edge on Google by incorporating human logic and knowledge into specific verticals or apps. A good example is Zite, an app that offers relevant news articles by learning behavior. It uses several AI technologies to make news delivery much more intelligent than a simple Google News feed.
How to get an advantage: Consider making vertical search pools plentiful, narrowly focused and well-stocked with content. Deep verticals can cooperate and solve problems without a loss of precision.
Virtual assistants: The app explosion that began when smartphones hit the market isn't slowing. Now, the trick is standing out via intelligent apps that go beyond just performing basic commands, becoming more intelligent and seeking to be more predictive of end user behavior. While Siri and Google Now aren't quite there yet, the next-gen virtual assistant will not only deliver meaningful information but will do so in a curated, predigested, presentable way that creates a seamless user experience. The big dogs are racing to build this "next-generation virtual assistant." Last month Apple acquired Cue, a virtual assistant platform, in a move seen as an indication that it's trying to make Siri smarter, more predictive and more interactive. While it's hard to imagine the current iterations of Siri or Google Now being a staple in most people's everyday lives, we as humans will adapt and come to expect on-target recommendations and intelligent answers.
How to get an advantage: While enterprises won't be building their own virtual assistants anytime soon, expect customers to begin to rely on virtual assistants as they become able to provide real-time, meaningful information. Watch the space closely, and as you design APIs, for example, keep in mind the goal of developing long-term, intimate relationships with customers via the apps with which they interact.
Overall, the smarter an app or platform is, the more predictive it can be, therefore the better it performs for the end user -- and the less likely it is to get deleted when someone needs more space for cat videos. As AI capabilities continue to expand, we will see intelligent systems that actually understand and predict human behavior. Yes, to get there, we'll need to jump a lot of technology hurdles and harness incredible amounts of processing power. But make no mistake: Machine learning and artificial intelligence will be key to building exciting, compelling products and services.