While the speech recognition in Apple’s Siri voice assistant has received the most attention, a key innovation in Apple’s Siri is its dealing with a user request as completely as possible. Siri shortcuts many navigation and data entry steps through tight integration with apps on the iPhone and with selected Web sites. Apple has the most elegant and integrated solution that uses this approach—in part because they tie it to a specific hardware platform and applications shipped with that platform—but other vendors such as Microsoft, Google, Nuance, and Vlingo have “voice assistant” models that move in this direction. If a text option were available (type what you would otherwise say when speaking isn’t an option), the “assistant” model can become the dominant user interface style, at least on mobile devices. (See “Where Apple missed the mark with Siri.”)
The evolution to an Assistant model is inevitable, and it could revolutionize mobile marketing and search. The change isn’t evolutionary—It’s a fundamental shift in control of user access to information and resources. The Assistant controls the results of a request much more than a classical web search. For many requests, for example, Siri doesn’t simply dump the user into a list of web sites generated by one of the search engines; it goes the next step, for example, responding to “Find me a great Greek restaurant in Palo Alto” by bringing up the Yelp web site showing the requested information. Web sites such as Yelp will aid this process by adding more filters and allowing saved searches; a hint of this is in the Yelp app announced by Yelp November 2 that will be pre-loaded on some T-Mobile phones. Adding resources for intelligent software agent access to web sites will become a major trend, making the Assistant model increasingly effective as a shortcut to a user’s stated goal.
Apple or other Assistant services have an option to capitalize upon that control to generate advertising revenues or referral fees to specific web sites. Eric Schmidt, Executive Chairman, Google, apparently sees this control, as embodied in Siri, as a threat to Google’s dominance in web advertising. In testimony before the Senate Judiciary Subcommittee on Antitrust, Competition Policy, and Consumer Rights on September 21, Schmidt commented, “Apple has launched an entirely new approach to search technology with Siri, its voice-activated search and task-completion service built into the iPhone 4S.” He quoted outside reports that characterized Siri as Apple’s entry point into the search business and an analyst’s characterization of Siri as intending to be a “Google killer.” While his comments should probably be taken with a grain of salt, since he is attempting to quash an antitrust investigation, his point is accurate—Siri does have the option of choosing sources for user requests or even becoming a search engine on its own.
The Assistant model is driven by natural language understanding (which can ultimately operate on speech or text). An additional aspect of the Assistant model is the use of dialog—back-and-forth clarification of a user request. Dialog adds to the power of natural-language interaction by allowing requests by the Assistant for clarification when the request could be interpreted in more than one way or when there is missing information needed to address the request. Some speech recognition solutions, by contrast, are once-and-out, e.g., “voice search” that allows speaking an entry for a search text box that then displays the usual visual list of web sites. Dialog will enhance both the effectiveness of the Assistant model and the “personality” of Assistants.
But the real story goes beyond the natural-language interpretation to what that technology allows—getting to a desired result or requested information in fewer steps. This reduction in steps has its greatest impact on small devices such as mobile phones, but is likely to spread to other platforms as well. (Expect microphone arrays–at least two microphones to allow reducing background noise–to become standard on pad computers and personal computers.) While Apple is providing leadership and credibility with its natural language interpretation in Siri, the companies previously mentioned (and others) also have similar underlying technology that we can expect to appear in similar Assistants.
The Assistant paradigm will fundamentally impact web and mobile advertising and marketing models. On smaller mobile devices where display ads or long lists of web sites are less effective, companies are struggling to find a good advertising mode. The effect of the Assistant model on marketing will be initially most pronounced on mobile devices.
The most obvious way that an assistant could affect marketing is in response to a sale opportunity, e.g., a question such as, “Do you have a discount coupon for jeans?” Much as in display advertising, a department chain could pay for an offering from that chain to be presented, perhaps with the address of the nearest store. Note that this avoids conventional web search, even though it may present similar results. This example also takes advantage of a mobile platform that can report location.
As a less obvious (and more complex) example, suppose you say to your assistant, “What’s my checking account balance?” When the Assistant paradigm is in full force, you would be prompted to enter your bank information the first time you did this, and as part of that enrollment, an app specific to your bank would be downloaded (provided by the bank). That app would include a way to verify your access to the account with a PIN or even using your voice characteristics.
When you asked a similar question relating to your bank later, the assistant would understand the context and drop you into the app, where you would continue interacting by voice. The company providing the Assistant would of course collect a fee for each time it launched the company-specific app, a form of advertising that is not obtrusive, and even regarded as a service. A company would benefit from this service by more automation of customer service requests that might otherwise go to a customer service line. A customer passing seamlessly from a friendly mobile assistant to a friendly automated company assistant would be less likely to expect to talk to an agent.
Lest this sound too far-fetched, note that Microsoft already supports voice-interactive customer service for many large companies through its Tellme operation. This service generates hosting revenues for Microsoft beyond any ad fees that the new model might generate.
There are other independent hosted customer service operations that feature speech recognition (for example, those of Nuance Communications and West Corporation). Many of the hosted operations, including Microsoft, already support “natural-language” call routing that leads with an open-ended prompt in personal-assistant style (“How can I help you?”), avoiding the annoying list of options typically presented by call center automation. The voice applications are generally developed with the aid of professional services groups within the hosting company or by independent organizations specializing in such development. Perhaps advertising agencies should also be involved and help give the Company Assistant a winning personality.
Effective customer service has been shown in multiple surveys to strongly affect a company’s image and repeat business, and is certainly an aspect of marketing. Automating that service in a way viewed as friendly through the Assistant model serves a marketing purpose. Automating customer service in a way that increases acceptance of automation can reduce call center costs, and thus, in addition to a marketing benefit, provide cost savings.
Both simple company apps and more complex apps and services will find a place in the Assistant paradigm. The evolution of the paradigm is likely to produce some very creative results. The Assistant model will inevitably change search advertising. Companies that are early to address this trend will reap the most benefit.