An introduction to Intelligent Virtual Assistants in the current market
The challenge of today's Intelligent Virtual Assistants (IVAs) is that they have largely overlaid existing functionality without introducing truly novel capabilities. A virtual assistant may allow you to issue commands such as “send a text message to Brian Jones,” but that’s something a user could otherwise do directly using standard phone capabilities. Either way, the end result is the same and, after the novelty wears off, users become indifferent.
Not unlike the first-generation websites that replicated the hard copy brochures before them, IVAs must transition to something more meaningful; they must fundamentally change the capabilities of the devices and applications surrounding us. Before Steve Jobs’ death, he stated that he had cracked the secret to building a successful Apple TV set: "It will have the simplest user interface you could imagine." When this device enters the market, the key element will be a Siri-driven experience.
But it is not just about TVs. Looking down the road five to ten years, we will see the integration of everyday objects from toasters to cars. Imagine clean-bezel devices with no buttons or knobs – where hotel toasters take simple commands such as "toast bagel" and pull temperature and duration preferences from a phone or a similar digital token, delivering a bagel toasted to perfection. Washing machines will act on a single command such as “please bleach this load of whites.” Cars will adjust seat, mirror, and temperature preferences while polling your personal device for your most recent music playlist without the press of a single button. Or how about an alarm clock that adjusts your wake-time based on the next morning’s flight time or wakes you up 20 minutes early when snow falls.
The Emerging Role of IVA’s
Futurists have long predicted a day where task-based agents would perform chores on our behalf: researching purchase options, scheduling flights, or gathering news. Virtual assistants will have a significant role in bringing this about because they represent a common interface that will transform the way we interact with machines. Instead of requiring us to spell out, in detail, what we need, our devices will simply act upon what we mean. This shift away from needing to learn how to command machines and toward machines understanding and adapting to us will be a welcome one. It will usher-in the next generation of applications and services, extending the capabilities of devices with functionality that we only imagine today.
While it may have first appeared that there could be a single agent that we could employ to go out and scour the digital world on our behalf, Apple’s Siri has taught us that mastery of an unbounded solution space is a difficult problem to solve. Google’s new voice search assistant may improve upon this slightly, but at the end of the day, what these virtual assistants are starving for are additional data services.
The Three Layers of IVA’s
In terms of Siri, the richer experiences – ones that don't simply end in a Google search query – rely on third party sources. For example, if you ask a question related to measurement or mathematical equations, it is shuffled off to Wolfram Alpha.
The way to improve upon this model is to treat virtual assistants as having three distinct layers: the broker, the classifier, and a collection of task based services which would drive specialization in any of the three layers.
The Broker: This is the top level interface – the friendly concierge that may be implemented within any given device, application, or website. This provides the point of interaction with the end-user and is responsible for managing state and preference data while acting as a broker for the exchange of data with a broad array of data services.
The Classifier Engine: This layer is responsible for taking an input and making a determination as to what service can best fulfill the request. Going back to the Siri/Wolfram example, Siri is not responsible for the actual determination of the answer, only for determining that it does indeed involve an equation or measurement.
Task-based Services: Contrary to the “futurist vision” of task-based agents interpreting the world around us, this job will be delegated to a collection of "language aware services" that can interpret an intent and apply deep domain understanding to natural language queries, using any available preference and state data to establish context.
Open Standards And IVA Adoption
Next IT favors an open-standards based approach. This approach allows any device manufacturer to implement their own top-level interface, enabling them to distinguish their devices from those of competitors while tailoring their offering to specific situations such as cable boxes or automobiles. More importantly, such OEMs will not have to relinquish control over media, content, or advertising sources. For instance, if Apple owns a top-level interface, like Siri, and a user queries for a particular type of media, the results will most certainly be skewed towards iTunes assets.
The result of an open standards approach will be that disparate commercial entities will want to provide rich, language-aware interfaces that facilitate their own business’s objectives, and, as a result, there will be a sufficient business case to invest in such models.
While it would not be commercially viable for a company such as Apple or Google to codify the baggage policies of a hundred different airlines, it would be worthwhile for an entity such as United Airlines to facilitate a language-aware service. For example, if an IVA is aware that a user has purchased a ticket on United, and the user asks "How much would it cost to take another bag?" a custom IVA is being leveraged to create cost reductions (call deflection), loyalty (a superior customer experience and immediate resolution), and potential ancillary revenue streams (selling baggage or upgrades).
The result will be a deep collective understanding across diverse industries that facilitates a much richer IVA experience: an ecosystem where all parties are rewarded for their participation.
Stay tuned for my next post, ‘Next IT’s Role in the New Digital Ecosystem’. And, don't forget to subscribe to the Next IT Blog to receive email notifications of new posts.