Amazon called re:Invent 2022 amid “mass layoffs” at its global digital division, as it is on track to lose $10 billion this year. The news has rattled a crowd interested in voice-based services and conversational AI, who are now questioning whether the lack of interest and use of Alexa reflects an overall disenchantment with voice services, including integrating Alexa into customer service or digital commerce initiatives Major brands and global corporations in the foreground.
Most postmortems can be traced back to an article by Eugene Kim in Business Insider (after the paywall, of course). Alexa was featured as the pet project of founder and former CEO Jeff Bezos, who gave it a huge runway when it launched in 2014. However, many attempts to “monetize” the operation have failed. Instead, the roughly 70 million Echo and other Alexa-enabled devices respond only to a short set of commands, such as “tell me the news.” “What’s the weather” or “Turn on the kitchen lights and start the coffee maker.” It never developed into a full-fledged medium for ordering goods (from Amazon), entertainment, or trying to incorporate it into e-commerce strategies like ordering pizza or summoning uber. Both proved unpopular.
This leads Ron Amadeo, author of the aforementioned article on Ars Technics, to ask a rhetorical question: “Are all voice assistants doomed?”, noting that Google Assistant suffers from the same dilemma as Alexa. His implicit answer was “yes.” Faced with mounting losses, both Google and Amazon are cutting resources in their respective divisions.
don’t have to
A year ago at re:Invent 2021, Jeff Blankenberg, Alexa’s chief technical evangelist, painted a very rosy picture for voice assistants in a session titled “What’s New in Amazon Alexa.” He noted that its developer community has developed more than 150,000 Alexa skills, noting that a certain percentage of them “make seven figures” from their skills. He also noted that “developer revenue” from “skills procurement” doubled year-on-year. Of particular note in Blankenberg’s presentation were initiatives to address issues such as the “voice interoperability problem,” where steps are being taken to enable individuals or businesses to build skills that use their own wake words and connect with a variety of other voice assistants. Other innovations include a “Transit” app that lets Alexa users better plan their commutes; a “Find My” app that uses Alexa to locate devices or people important to device owners; Shopping Actions serve as a mechanism for merchants to sell goods or services through Alexa.
All these initiatives are still in place. For example, it was recently reported that more than 130,000 third-party sellers worldwide used Shopping Actions to make more than $100,000 in sales on Amazon during the 2021 holiday season. But they’re significantly lesser known in all of AWS’ offerings. When looking at the agenda on the virtual re:Invent 2022 website, I couldn’t find “What’s New with Amazon Alexa” that offered updated pictures. This made me doomed to scroll through Google, Twitter, and LinkedIn for further mentions of Alexa’s grim fate, and by extension Siri, Google Assistant, Cortana, Bixby, and other voice assistants. It’s clear that every virtual assistant trying to tap the prospect of reuse and monetization has an Achilles’ heel. They don’t talk to each other, and they don’t engage well with the customer service infrastructure that all brands have integrated into their contact centers, websites, mobile apps, and digital messaging platforms.
In the 8+ years since Alexa launched, thousands of companies have hired or contracted developers to build their own Alexa Skills. Few of these initiatives are designed to integrate with existing customer service contact centers. That’s not surprising, because even though it’s a voice assistant, Alexa doesn’t share any real-time voice streams its end users create. Instead, Amazon captures and transcribes spoken language in near real time, and shares transcriptions with third parties, including skill developers. Gone are the rich content represented by pitch, tempo, and timbre, which are often used in contact centers to gauge agent performance or user sentiment. Companies cannot use an individual voice to authenticate each end user. As a result, brands lack sufficient confidence to pull customer profiles and provide personalized service.
Competitive sees voice as an asset and is making it a key component of “conversational intelligence,” informing future conversations between customers and agents, or between customers and bots. Amazon has all this information. It just doesn’t share. This is the first red flag for enterprise implementers and a major obstacle to monetization.
You can now contact an agent via Alexa, but it won’t be easy
Amazon’s Echo device was originally positioned as a “smart speaker”, and Alexa was its companion voice assistant. They were never phones, though Alexa has also been instantiated as an app for smartphones, and voice assistants for a handful of car brands like BMW and Lamborghini. It’s also often used to understand commands for smart appliances, home controllers, and lighting systems.
Making a call or contacting a customer service agent in a contact center involves an entirely different set of skills for Alexa. Two notable examples reflect the fact that there may be some basic need for voice calls via Alexa, while showing that many activities must take place under well-known cover. One approach described here enables developers to use the Alexa Skills Kit (ASK) to enable Alexa users to initiate a chat or trigger a call with a customer service agent.
It is a combination of ASK and Amazon Connect, a proprietary Contact Center as a Service (CCaaS) offering. As noted in the tech guide, it appears to require customers to use the Alexa app on their smartphones, rather than talking to an Echo or the car’s message center. Authors Cuong Do Vu and Soumiya Mathur describe how Alexa users can start multiple steps by saying “connect me to an agent” when using a skill offered by their brand of choice. After authenticating users, it lets them initiate chat sessions immediately.
If a customer wants to talk to an agent, an instruction is sent to the brain in Amazon Connect to treat the session as a “callback.” In other words, instead of calling a business contact agent, the Alexa skill signals a resource in AWS to ask Amazon Connect to initiate a call to the customer, which then connects to the agent. Making a simple phone call seems to call a lot of processes. There is no feedback on whether any enterprise customers are using this function.
Another long-term effort to enable Alexa to make phone calls comes from the team that provides the Chime SDK. It’s called a “skill call”. It’s not limited to talking to an Amazon Connect contact center. Instead, it enables users of Alexa skills to place calls using simple voice commands. For example, after the obligatory “Alexa” and turning on a skill, users can simply say “call customer support” and be connected to existing support resources.
While the user interface is still cumbersome, users no longer need to initiate activities from the Alexa app on their smartphones. They also don’t trigger the chain of events that lead to Amazon Connect callbacks. The user is actually calling a customer service agent. Calls can be routed by SIP, phone number, or (optimistically) connecting to an Amazon Chime conference. What’s more, “Skill developers attach relevant data such as usernames, account IDs, and reasons for calls to communication sessions to help reduce handle times and increase customer satisfaction.” It’s a small tunnel outside Alexa’s walled garden.
room for improvement
Making Alexa financially successful was always a challenge. In addition to being unable to make phone calls, privacy concerns also put about 40% of the potential market away from purchasing and using it. VUI is not conducive to containing advertisements. Using “reminders” or “notifications” to tell Amazon shoppers when to reorder frequently purchased items (tissue, toilet paper, bottled water) is considered intrusive. However, the biggest factor is the underlying architecture and ecosystem, which sets Alexa apart from other voice assistants and makes the development of skills and apps incompatible with other platforms. The lack of standardization across platforms has been a huge hurdle for enterprise customer service professionals, who already have multiple messaging platforms and voice assistants to try and support.
The Amazon Connect and Chime SDK initiatives attempt to address the challenges of interconnecting with contact centers and other customer support infrastructure. This is a start. Outside of Amazon, the Open Voice Network has launched a number of projects and workshops aimed at facilitating interoperability between voice assistants and overcoming many of the barriers to enterprise adoption of conversational AI. With tens of millions of devices capable of understanding and responding to what their owners say, it’s not too late to open up, expand real-time communication capabilities, and promote enterprise-aware applications.
Categories: Smart Assistants, Articles