The two key nascent technologies being showcased at SXSW 17 in Austin Texas, are undoubtedly AR/VR and AI cognitive learning, the latter of which essentially involves talking to computers through conversational search, chatbots and natural language processing.
AR and VR are, of course, fascinating, though still feel a little gimmicky, but AI has the possibility to not only fully disrupt the marketing world over the next few years, but to fundamentally shift the way we interact not just with machines but with people.
Over the last few days, I have seen technologies and services from IBM, Google, Amazon, Facebook, Dell, and heard from companies either just starting their journey or well into AI, chatbots, personal assistants, and machine learning that either make you very concerned or very excited about the future, depending on your perspective. We heard about a fashion retailer who last year allowed an AI to design a number of skirts and tops based on current trends and styles, and a recipe website that has built an AI to take lots of recipes for the same thing (say chocolate chip cookies) and try and develop a super recipe based on all the attributes of the recipes on site.
Take a very simple task for a human: "I'd like to get a bottle of wine that goes with pasta, on the way to Charles' house". For your human brain it's easy to process. You can instantly picture the route to your friend's house, recall where the bottle shop is, allocate a suitable budget, and decide which wine (red or white at least) goes with the pasta dish you are taking. But if you were to ask an AI the same question then the data sets the computer would have to know are staggering. At the very least it would need to know:
- Your address
- Charles' address
- The route to Charles’ house, and the closest bottle shop to that route
- Your normal budget for a bottle of wine
- What wines go with pasta, and in fact what type of pasta dish you are making
- The inventory of the store closest to your route
That is a lot of information – address books, bank data, maps, wine database, food database, wine matching API, store inventory, etc. It’s especially complex because unlike search as it stands today that gives you a number of options based on keywords that you search for (in this case something like 'bottle shops near me'), in audio search you only want one response. The right answer. But this is exactly what companies like Google, Apple and Amazon are trying to achieve, through natural language processing and cognitive computing.
At the moment, the solution for AI to be able to understand what you need is through conversational search. Basically, the computer asks you questions that makes answering your main question easier, because—and this is really important—you only want one answer; not a list of choices. So, in this case, your AI Personal Assistant, may respond with, "what is Charles' address? Which pasta dish have you made? What is your budget? Are you happy with Glengarry on New North Road?" And like any AI worth its salt, it will remember every answer you give, so that over time it will make it smarter and smarter.
I went to a talk that included one of the leads on "conversational search" from Bing and his advice was, plan now, as it will be the single biggest shift in consumer marketing behaviour, since search itself. This is because the web as it stands today has been indexed and is searched through keywords and phrases, sifting through search results until you get what you are looking for. With AI, keywords become redundant in many respect and through conversational search the focus shifts from keywords to intent. Search engines, like Google and Bing, are monetised through brands buying keywords and then serving an ad to a consumer along with other topically relevant ads. So, what happens if rather than having multiple results, you only want one; how would that be decided, what is the role of SEO, versus SEM? In the example above, in which we're looking for a bottle of wine you wouldn't want a wine brand to be able to bid on your question and get their brand of wine recommended over others simply because they paid for it.
So what's the advice from SXSW? Ignorance is not a strategy; start thinking now about what tasks an AI, be it a chatbot or something more elaborate. Could take over and start building a case? Happy to point you in the direction of some of the vendors I've met here. And remember that the sooner you get started the sooner your AI will start learning.
In the world of search, now really is the time to start thinking about voice search and conversational search. Discuss with your SEO team a 12-month roadmap to be ready. Alexa, Google Home, Cortana, and Siri are here, and they are only going to get smarter. You don't want to be left behind.
For example, how powerful would it be for a consumer to ask Alexa on a Saturday morning "what's the best deal on Samsung TVs in New Zealand at the moment?" and immediately the response comes back: "Noel Leeming has 50 percent off all Samsung TVs this weekend. The St Luke’s store, which is closest to you opens at 9am. Would you like me to send directions to your phone..."
- Alex Radford is general manager of digital media for the Dentsu Aegis Network