The Coming Era of Assistant-Centered Design

Siri Shortcuts is inarguably one of the most exciting aspects of Apple’s rollout of iOS 12, promising to inject a much-needed boost of utility into the industry’s once-pioneering digital assistant. But as exciting as this moment is for Apple and its users, it speaks to a larger trend that will be sweeping all digital assistants: to be truly helpful, our assistants must be as conversant with other digital services as they are with us.

Advancements in robotics and AI have surfaced the notion of “robot-centered design,” which is critical as robotic automation swallows more and more tasks people used to be paid to do. Robots, although frequently anthropomorphized and sometimes imbued with human-like intelligences, will navigate the physical environment far differently than we do. Designers need to account for robot-specific environments, and mixed-use environments where a diversity of robots and humans, of varied abilities and dispositions, work and thrive alongside each other. This imperative applies equally to the factory floor as to our vast road system if we want autonomous vehicles to ever see mass adoption.

But perhaps even more important than re-thinking our physical environs for our robot friends is re-imagining our digital universe for use by our AI-enabled assistants. The battle for AI supremacy continues to escalate, and the consumer-facing assistants that sit in our pockets, in our cars and throughout our homes will soon undergo several major transformations that will upend how we think about, use and even pay for them.

Assistants will become proactive, highly trained and personalized extensions of us

Smart assistants today are now largely passive responders, only rarely providing a glimpse of their potential as truly predictive utilities. But the pace of innovation in AI, paired with our insatiable consumer demand for simplicity and convenience, will soon render them something entirely different: our proactive agents, emissaries that act on our behalf in both semi- and fully-autonomous modes. They will monitor our households, replenish our goods, manage our stock portfolios, schedule and reschedule our appointments (as Google Duplex has already hinted at), arrange our social occasions, and discover our content. They will know us so well that they will not only save us time and effort, but open us up to new experiences we may never have found by ourselves.

Users will stop thinking of “assistants” generically. Their primary identity will be a reflection of an individual and their needs, and only secondarily will they be known by their tech platform enabler. Google or Amazon may power my assistant and ensure it benefits from the technological leaps that come from scaled machine learning, but it will be first and foremost my service.

So how, exactly, will our assistants do all this?

Knowing us is only half the battle. The other half is intelligently crawling the web, interfacing with a host of content and systems, a great many humans, and yes, other AI-enabled assistants. What will they choose for us? What will they look for as the digital hallmarks of appropriateness to our tastes and specifications? In what circumstances will they escalate a decision to us, and how will that decision be presented?

These are hard questions to answer given the algorithmic opacity surrounding big tech’s AI. A lack of transparency is understandable for competitive reasons, but also due to the nature of neural networks, where it’s often far harder to know how an AI learns than to teach it something. So how will we design our digital universe to index for assistants whose inner workings remain somewhat of a mystery, even to their creators?

Tech companies will release Assistant Optimization Guidelines

“If you want Siri, Alexa, Google Assistant, Cortana or Bixby to find you, here’s what you need to do.” Expect that data labeling guidance to come very soon. We’re already seeing advancements like speakable markup for search on Google Assistant to enhance appropriateness and discoverability of content for voice. This is just a voice-centric tip of the assistant optimization iceberg. The addition of visual search capabilities, and the ability to respond to information and design with human-like perceptual tendencies, will be required to return better results and recommendations to us. Beyond the algorithmic coldness of today’s search engines, the assistants of tomorrow will reflect an understanding of personal taste, tone and nuance.

Assistant-to-Assistant (A2A) communications will become the norm

Communication between assistants is already happening on a platform level — as Alexa and Cortana forged a symbiotic partnership that serves the mutual aims of Microsoft and Amazon. But this is during the era where assistants are defined as the platforms we all use, not the personalized tool of the individual. My assistant will naturally correspond with those of my friends and family. But businesses will also get on board, and my company’s AI assistant may well iterate with my client’s AI assistant on a draft of a document before it reaches the attention of either me or my client. Will individuals and businesses create different paths (e.g., human-to-human vs. assistant-to-assistant vs. assistant-to-human) for different sorts of services? The debate surrounding Google Duplex and disclosures around when an AI is in use certainly indicate this route-parsing will be needed, and will need to be made explicit.

Assistants will be monetized, in a big way

The sheer scope of activities that will be potentially automated by our personal assistants, including the complexity of B2B tasks in an A2A context, will necessitate tiered service offerings. Performing a vast array of functions reliably on our behalf won’t be cheap, and will likely require frequent human intervention in a hybrid service model — when something goes wrong with my assistant’s actions, I’ll need someone to complain to and ensure it is a mistake not repeated. The willingness to pay for premium assistants will likely be high, and will depend on just how powerful the highest-echelon offerings are for personal and business use. The economic implications may be further stratification of an already widely-gapped society, but may also include their own job creation opportunities.

There is no certainty as to if and when these developments come to pass, but it’s not a stretch to believe that consumers and businesses will increasingly want to — to co-opt and genericize Google’s current Assistant tagline — “make an assistant do it.” I also wouldn’t bet against the pace of advancement of the AI behind consumer-facing and enterprise-level applications. Considered together, the only missing piece — albeit a big, complex one — is the evolution of the digital fabric that connects us, and our assistants, together.

Originally published on LinkedIn · September 2018