Even the most cursory glance at the 2017 SXSW Interactive schedule tells you that a strong theme this year is the rise of the bot and the increasing role Artificial and Machine Intelligence will play in society.
Agencies have been quick to capitalize on the potential of AI and the role "conversational interfaces" can play in delivering branded services to customers. We are always looking for new ways to help our clients engage with their customers better, which is why we’ve seen a slew of campaigns recently that feature some form of automated chat-component.
But, this is really just the beginning for a fully-fledged conversational marketing revolution. Soon we’ll start to see Machine Intelligence used to both persuade us to do something as much as it’s starting to help us do things. When I say persuasive, I’m not talking about programmatic ads but rather AI-driven conversational commerce.
Not for much longer will online shopping sessions start by following a piece of social content that caught the eye, to a (relatively dumb) branded destination, and then through a succession of menu clicks to filter you down to a product suggestion.
Soon we’ll start to see Machine Intelligence used to both persuade us to do something as much as it’s starting to help us do things.
As Alexa is rapidly teaching us, the new shopping trip starts with a spoken instruction along the lines of "Find me a flattering white cotton shirt for less than £100 that I can have delivered tomorrow."
Most of that request is easy-peasy for systems to parse, but the key word in that statement is flattering. That request for something that will complement my (ever challenging) body shape can’t be mapped onto discrete product attribute tags. It’s a highly complex and personal set of interdependent opinions and values.
Now sure, if a machine was given access to a load of personal data – say my Instagram account – then it could rapidly scan through those pictures and build up an opinion of my body shape from all the selfies and make a recommendation of which shirt is likely to be most flattering for me.
However, my Instagram account is mainly made up of pictures of all the epic dinners I’ve been eating (which is probably why I need help finding a flattering shirt style – but hey) and I suspect I’m not alone in this fact.
Equally, it could scrape all the messenger banter between friends discussing outfits to glean clues on what’s deemed flattering, but without context it’s pretty hard to be sure.
To date, ecommerce has largely ignored these subjective shopping challenges by smothering it in customer service extensions to make it easier to buy – offering more and bigger pictures of garments, extreme close-ups of fabrics and the ultimate backstop fail-safe of offering no-quibble free returns.
As helpful and convenient as those services are, they aren’t really advancing the art of online retail, merely compensating for its deficiencies.
Ask anyone who works in traditional high street fashion retailing, and a subjective request for a "flattering fit" is far less daunting. A skilled sales assistant is adept at casting a critical eye over a customer, asking questions, listening to the replies, before making suggestions to quickly assemble a rail of possibilities for the customer.
Poke’s work with Ted Baker over the past five years has taught us the vital importance of delivering great customer experience through the line, and we’ve wrestled a few times with how to convert Ted’s legendary in-store staff expertise into the online space.
Modern retailers are starting to work out how online behaviors and patterns can be interpreted into better product recommendations, with platforms like Propulse Analytics leading the way, however this only improves the accuracy of recommendations – it doesn’t mean that shirt will be a perfect fit when it arrives and go on to become a treasured and over-worn favorite.
We can now digitise body measurements and with 3D product measurements, match a garment to your body shape, and make sure the right size is chosen. You might even be able to visualize that garment on an approximation of your body-shape if the retailer has invested in 3D models of their garments, but that’s impractical for low-priced fast fashion brands.
Surely, in order to look ahead we first need to look back, and recognize what we’re lost in the race to online trading.
That expertise, opinion (and honesty) from a qualified sales assistant is what’s missing, and we should augment those unique human abilities by training AI with in-store customer interactions to enrich recommendations.
So, back to SXSW. What I’ll be looking for is genuine cognitive intelligence, as opposed to clever multiple-choice branched logic patterns. I’ll be looking for AI systems which can understand more than voice, and parse other customer signals, such as facial expressions. That changing room mirror could learn a lot from the expressions we pull when we try things on.
Combine that level of Machine Intelligence, with the experience of trained human assistants, and you have the beginnings of a new benchmark for retail experience.
—Tom Hostler is a founding partner at Poke.